Objective Laser-induced breakdown spectroscopy(LIBS)has emerged as a promising technique for rapid,in-situ characterization of rock-core samples owing to minimal sample preparation requirements and simultaneous multi-element detection capability.In geological exploration and resource evaluation,accurate quantification of major elements such as sodium(Na),magnesium(Mg),potassium(K),iron(Fe),calcium(Ca),and aluminum(Al)is essential for assessing mineral composition and economic potential.However,quantitative LIBS analysis of complex geological matrices faces significant challenges:strong pulse-to-pulse intensity fluctuations driven by transient plasma variability,pronounced matrix effects from heterogeneous mineral composition,and microscale surface non-uniformity of pressed powder pellets.Conventional single-shot approaches are susceptible to these disturbances,yielding unstable predictions and limited cross-matrix generalization.The present work aims to develop a practical,hardware-independent approach that exploits repeated-shot statistics to enhance repeatability,accuracy,and dynamic-range coverage of LIBS quantitative models for rock-core analysis. Methods Forty-three certified national reference rock powder samples spanning soils,sediments,and ores were pressed into pellets and measured using a 1064 nm pulsed Nd:YAG LIBS system with a side-axis fiber-coupled spectrometer covering 200-800 nm(Fig.2).For each sample,100 individual laser shots were acquired in a fixed-distance array pattern with 5 mm inter-point spacing;five pre-ablation shots were applied before each sequence to remove surface contamination.Key acquisition parameters are summarized in Tab.1.After outlier rejection,baseline correction,normalization,and z-score standardization,partial least squares(PLS)regression models were independently built for six target elements using 35 training-set samples.A multi-shot fusion strategy was implemented at the prediction stage by arithmetically averaging the 100 per-shot predictions for each test sample,without altering the learned spectrum-to-concentration mapping.To address wide dynamic concentration ranges,a binary high-low modeling scheme was introduced:a low-range sub-model trained in logarithmic concentration space for trace-level sensitivity,a high-range sub-model with concentration-proportional sample weighting for elevated-level robustness,and a tunable Sigmoid gating function for smooth blending of both sub-model outputs(Fig.3). Results and Discussions Single-shot predictions for all six elements exhibited considerable scatter,with large per-sample standard deviations reflecting stochastic plasma fluctuations and local surface non-uniformity(Fig.4(a)).After applying multi-shot fusion,predicted values shifted markedly toward the ideal 1∶1 reference line and residual error bands narrowed substantially across both training and test samples(Fig.4(b)).Quantitatively,test-set root mean square error(RMSE)decreased by 10.5％-66.7％and the coefficient of determination(R2)improved for all six elements(Tab.2).The largest gain was recorded for Fe,whose RMSE dropped from 2.903 wt％to 0.966 wt％(66.7％reduction),consistent with its exceptionally broad concentration span across the sample set;K also benefited markedly,with test-set R2 rising from 0.878 to 0.964 and RMSE falling by 45.8％(Tab.2).These improvements are attributed to statistical averaging of pulse-to-pulse random errors and the broader spatial coverage afforded by the 5 mm-pitch array pattern.For elements whose concentrations span multiple orders of magnitude,the binary high-low PLS framework provided further gains.Taking Fe as the representative example,linear-scale and logarithmic-scale scatter plots confirm that the low-range sub-model systematically underestimates at high concentrations,whereas the high-range sub-model loses precision in the dilute region;the dual-model blend eliminates both deficiencies(Fig.5,Fig.6).Quantitatively,the Sigmoid-gated dual model reduced Fe mean relative error(MRE)from 115.12％(high-range sub-model alone)to 20.11％,while simultaneously lowering RMSE from 0.958 wt％to 0.666 wt％(Tab.3).Consistent MRE and RMSE reductions were observed across all six elements under the binary scheme(Tab.3),confirming that the approach generalizes beyond Fe.The complete inference pipeline processed approximately 1674 spectra per second(0.597 ms per spectrum),confirming real-time suitability for on-site geological deployment. Conclusions A prediction-level multi-shot fusion strategy combined with a binary high-low PLS modeling scheme and a Sigmoid gating mechanism substantially enhances LIBS quantitative stability for complex rock-core matrices.Multi-shot fusion alone reduces RMSE by up to 66.7％,and the binary dual model further suppresses relative errors in the low-concentration region while preserving high-concentration accuracy.Compared with prior approaches based on single-shot spectral normalization or double-pulse hardware enhancement,the proposed framework extracts additional predictive value from repeated-shot data without increasing measurement complexity or hardware cost.Classical PLS regression and statistical aggregation form the sole computational basis,facilitating deployment on portable or industrial LIBS platforms.Millisecond-scale inference satisfies the real-time demands of field geological surveys,resource evaluation,and environmental monitoring.Future research may explore adaptive outlier-aware shot selection,cross-instrument domain adaptation,and more expressive nonlinear regression modules to further improve robustness across extreme matrices and ultra-trace concentration ranges.
Objective Mid-wavelength infrared band covers a key atmospheric window and contains characteristic absorption lines of numerous gas molecules,driving strong demand fortrace gas analysis,target detection,and free-space optical communication.Developing novel mid-infrared detector materials and heterostructures has become a major focus in optoelectronics.The quantum cascade detector(QCD),a photovoltaic device based on intersubband transitions,employs a multi-quantum-well structure.Electrons in the absorbing well are excited from the ground state to an excited state by infrared photons.Subsequently,through a staircase of subband levels formed by coupled quantum wells with energy spacings matching the longitudinal optical phonon energy,photogenerated carriers are directionally transported via LO phonon scattering.Under zero bias,vertical collection of photogenerated carriers is achieved,offering advantages such as zero-bias operation,low dark current noise,and flexible wavelength tunability.However,the low responsivity of QCD devices is mainly limited by two factors.First,the absorption region adopts a quantum-well structure with a small absorption coefficient,making it difficult to fully utilize the incident photons.Second,the multiperiod cascade structure makes the responsivity inversely proportional to the number of periods,requiring the consumption of multiple photons for each electron collected by the external circuit.In addition,thermal backfilling and scattering losses during photogenerated electron transport further reduce the carrier collection efficiency.This work focuses on the issue of low absorption coefficient and reports a mid-infrared QCD structure and device employing a double-well-coupled structure. Methods A double-well coupled structure was designed to enable multichannel transitions from a strongly coupled ground state to several near-degenerate excited states,and the corresponding device is fabricated.Theoretical calculations validate the feasibility of this structure.The QCD structure is grown on a InP substrate via metal-organic chemical vapor deposition,and device fabrication is completed using lithography and etching.The photocurrent spectrum is measured using a Fourier-transform infrared spectroscopy system.Responsivity is calibrated with a blackbody radiation source and a lock-in amplifier.Dark current and R0A are characterized by an Ⅳ source meter,and the detectivity is subsequently derived. Results and Discussions Theoretical calculations indicate that the device achieves an extraction efficiency of 86.5％,a significantly increased transition matrix element,and an absorption intensity of 1.8％,corresponding to a theoretical peak responsivity of 44.6 mA/W.Experimental results show that the fabricated double-well-coupled mid-infrared quantum cascade detector exhibits a peak response wavelength of 4.9 μm and a peak responsivity of 16 mA/W at 77 K(Fig.3).The R0A value is 6× 105 Ω·cm2,and the peak Johnson-noise-limited detectivity reaches 2×1011 Jones(Fig.5),which is consistent with the theoretical design values.The discrepancy between the experimental and theoretical results is likely attributable to inferior epitaxial growth quality compared with theoretical predictions. Conclusions A double-well-coupled quantum cascade detector with an operating wavelength of 4.9 μm at 77 K was designed.By optimizing the energy level structure,the theoretical responsivity was successfully increased,and the performance improvement was experimentally demonstrated.The detector exhibits high detectivity and responsivity at 77 K,while still maintaining a response at room temperature,indicating that the double-well-coupled structure offers advantages in terms of flexible energy level design and enhanced device performance.In the future,further improvements in epitaxial material quality will be pursued to achieve even higher performance.
Objective Semiconductor packaged devices often require failure diagnosis under the constraints of noncontact measurement,no package opening,and rapid hotspot localization.In practice,defect induced thermal signatures can be weak and may be obscured by direct current background radiation,slow temperature drift,and asynchronous disturbances.In addition,thermal diffusion in multilayer package structures reduces spatial contrast,which limits localization robustness.This work proposes an electrically excited infrared thermography method that uses internal Joule heating as a defect related heat source,enabling hotspot enhancement and localization without decapsulation. Methods A unipolar sinusoidal electrical excitation with a direct current bias is applied to the specimen,so that defect related conductive paths generate a periodically modulated Joule heating component at the fundamental excitation frequency.An infrared camera records the surface temperature sequence.Pixelwise dual channel orthogonal coherent demodulation is performed by discrete correlation with in phase and quadrature reference signals over an integer number of excitation periods,and the fundamental frequency amplitude plot and phase plot are reconstructed(Fig.1).This demodulation suppresses the direct current background,slow drift,and asynchronous noise.For objective comparison across different excitation conditions,a peak gradient based self calibrated spatial partition is introduced to determine a characteristic hotspot scale,and a coupled metric that jointly reflects peak significance and spatial focusing is constructed for parameter selection;the defect cluster center is obtained using a lightweight unsupervised clustering step(Fig.5,Fig.6). Results and Discussions Experiments were conducted on an electrically overstressed Microcontroller Unit(MCU)specimen in a quad flat package,and the reconstructed amplitude plots exhibit a single peak distribution that is close to Gaussian around the defect;the peak position remains consistent across excitation conditions,demonstrating stable localization(Fig.4).Frequency sweep results show that increasing the frequency markedly reduces the core hotspot area,indicating enhanced spatial focusing(Fig.7(a)),while the peak significance decreases at higher frequencies(Fig.7(c)).Nevertheless,the coupled metric increases overall with frequency(Fig.7(b)),indicating that the gain from spatial focusing exceeds the loss in peak magnitude under the tested conditions.Voltage sweep results indicate that the core area varies weakly with the excitation amplitude and shows larger scatter(Fig.8(a)),whereas the coupled metric exhibits a clear nonlinear trend across the tested range(Fig.8(b)),consistent with the combined influence of peak magnitude and spatial diffusion(Fig.8(c)).Phase plots show a stable phase anomaly region across multiple frequencies(Fig.9).The phase increases with frequency in a nonlinear manner and tends to saturate at higher frequencies(Fig.10),and phase profiles provide repeatable extrema that can serve as an auxiliary and more robust cue for localization(Fig.11,Fig.12).Decapsulation inspection,bare die observation,and emission microscopy(EMMI)comparison confirm that the localized infrared hotspot corresponds to the electrically abnormal region(Fig.13). Conclusions An electrically excited infrared thermography approach combined with orthogonal coherent demodulation is developed for hotspot localization in semiconductor packaged devices without decapsulation.Joint amplitude and phase analysis supports objective parameter selection and improves localization robustness.Under the experimental conditions of this work,a lock in frequency range of 0.5 Hz to 1.25 Hz and an excitation amplitude range of 1.0 V to 1.4 V are more favorable for stable defect detection and localization.
Objective The time delay integration(TDI)CMOS image sensor improves the signal-to-noise ratio(SNR)by accumulating the same image signal over multiple exposures,making it widely applicable in high-SNR-demanding fields such as aerospace,satellite imaging,and semiconductor testing.To further enhance the circuit's SNR,the number of accumulation levels in the analog domain accumulators is increased.However,parasitic capacitance and circuit noise also increase with the higher accumulation stages,which limits the improvement of SNR.Traditional analog accumulators can no longer meet the requirements of such applications.To reduce the impact of parasitic effects on accumulation accuracy,a compensation method utilizing polarity inversion switches and adaptive positive feedback capacitors is introduced.Furthermore,in response to the core technical bottleneck caused by front-end circuit noise,which limits SNR enhancement in high-stage analog accumulators,an innovative architecture for the TDI CMOS image sensor's analog domain accumulator based on fully parallel correlated multi-sampling(CMS)technology is proposed. Methods CMS technology is integrated with the analog domain accumulator using shared capacitors,resulting in a fully parallel CMS-based architecture for the analog domain accumulator(Fig.4)and its circuit timing sequence(Fig.5).This architecture allows the input signal to be sampled and averaged multiple times while enabling signal transfer and accumulation based on its structural characteristics,without the need for additional buffers.The architecture is designed to effectively suppress circuit noise within a more compact chip area,while ensuring that the TDI accumulation line rate remains unaffected,thereby enhancing the SNR performance of the TDI CMOS image sensor analog domain accumulator.By abstracting the mathematical model of CMS technology(Fig.2)and deeply coupling it with the TDI SNR model,the circuit's SNR enhancement is optimized,and a systematic performance evaluation is conducted. Results and Discussions To evaluate the SNR improvement of the proposed architecture after its actual design,an analog domain accumulator based on fully parallel CMS technology was implemented in a 2048×128 TDI CMOS image sensor using a 55 nm CMOS process.Simulations comparing the ideal(Fig.9)and actual(Fig.10)SNR improvements show that the actual SNR improvement closely matches the ideal values,demonstrating a significant enhancement in the circuit's SNR.Simulation of the circuit's effective accumulation order(Fig.11)reveals that the proposed circuit achieves a 128-order accumulation effect,equivalent to the 625.461-order effective accumulation result of the circuit in[10],significantly enhancing the circuit's effective accumulation order.The circuit's quality factor calculation indicates that it performs excellently in terms of power consumption,SNR,and line frequency(Tab.1).Additionally,image processing tests confirm that the circuit effectively suppresses low-frequency noise from the front-end circuit,leading to a noticeable improvement in SNR.This results in fewer imaging noise spots and clearer image quality. Conclusions An analog domain accumulator architecture based on fully parallel correlated multi-sampling technology is designed,utilizing shared capacitors to synchronize correlated multi-sampling and analog accumulation.This approach effectively suppresses circuit noise while maintaining the TDI accumulator's line frequency,achieving a significant improvement in circuit SNR and enabling a higher effective accumulation stage with a more compact chip area.The proposed method,implemented using a 55 nm standard CMOS process,is successfully applied in a 2048-column,128-stage TDI image sensor.Post-layout simulation results show that,at a line frequency of 10.85 kHz,the SNR of the 128-stage accumulation improves by 27.962 dB.Of this,the polarity inversion switch and adaptive positive feedback capacitor method contribute 20.9 dB,while the CMS technology provides an additional 7.062 dB.The power consumption per column is 301 μW,with an area of 0.361 mm2 for the 128-stage accumulator per column.The CMS circuit reduces the area by more than 40％compared to traditional designs,reducing the area of each column of the proposed accumulator to 98.5％.The circuit effectively reduces noise and enhances the SNR,significantly improving imaging accuracy under low-light conditions.
Objective The planning of lunar bases,represented by the International Lunar Research Station,is advancing lunar exploration into a new phase of"long-term utilization and in-situ resource transformation".In-situ resource utilization(ISRU)of lunar regolith is crucial for constructing sustainable lunar habitats.However,the extreme lunar environment and manufacturing processes may introduce internal defects into regolith-based components,posing risks to structural integrity.Thus,developing in-situ non-destructive detection techniques is imperative.While infrared thermography offers significant advantages,its conventional excitation sources(e.g.,halogen lamps,lasers)rely on external power supplies,presenting limitations for long-term lunar surface operations.To address this problem,this study proposes and investigates a novel in-situ infrared thermographic inspection method utilizing concentrated solar energy as the excitation source. Methods A three-dimensional finite element model coupling optical excitation with transient heat transfer was established.Four groups of lunar regolith simulant specimens with standardized dimensions(20 mm× 15 mm× 5 mm)were designed,each containing an array of flat-bottom hole defects at their surface center.These defects varied systematically in diameter(1-4 mm)and burial depth(0.2-0.5 mm).Numerical simulations were conducted under three characteristic scenarios:earth ambient temperature,lunar high-temperature,and lunar low-temperature environment.First,the detection process of a specimen with 1-mm diameter defects was simulated and compared across all three environments.Subsequently,the influence of defect size was analyzed by simulating specimens with defect diameters of 2-4 mm under the lunar low-temperature condition,using a pulsed heating protocol. Results and Discussions From the temperature distribution at typical times in earth ambient temperature scenario(Fig.5),lunar high-temperature scenario(Fig.6),lunar low-temperature scenario(Fig.7),it can be seen that the optimal detection window universally occurs during the late heating to early cooling phases across all scenarios,as the thermal excitation is sufficiently established while background thermal noise remains non-uniform,maximizing the temperature contrast between defects and the substrate.In the lunar high-temperature scenario,the elevated initial temperature combined with intense heating accelerates thermal diffusion,allowing deeper defects to manifest earlier.However,this advantage is offset by a significantly reduced thermal contrast due to the overall high background temperature,which compromises defect contour clarity and identifiable duration.As shown in temperature distribution of specimens with different defect diameters at typical times(Fig.8),under identical excitation conditions,larger defects produce thermal signals with greater robustness and a longer observable time window. Conclusions This study establishes a coupled"optical-thermal"3D finite element simulation model,systematically demonstrating the feasibility of using concentrated solar excited infrared thermography for in-situ defect detection of lunar regolith molten structures.The temperature field evolution trends are generally similar across the three scenarios,with the optimal detection window commonly appearing during the late heating phase to the early cooling phase.The distinction lies in the lunar high-temperature environment,where deeper defects manifest earlier but the excessively high background temperature reduces the recognizability of defect contours.Under identical excitation conditions,larger-sized defects produce thermal signals with stronger robustness and a longer observable time window.
Objective Hyperspectral imaging technology plays a crucial role in earth observation and quantitative remote sensing,where data quality fundamentally depends on precise spectral and radiometric calibration.However,mid-wave infrared hyperspectral imaging systems often suffer from spectral response drift and radiometric characteristic changes in complex field environments.The coupling between these two factors leads to cross-propagation of spectral-radiometric errors,which severely limits the accuracy of traditional stepwise calibration methods.This study aims to address this challenge by proposing a novel joint calibration method capable of simultaneously retrieving spectral and radiometric parameters to enhance the overall accuracy and efficiency of field calibration. Methods A spectral-radiometric joint calibration model based on atmospheric absorption feature weighting was established.Firstly,the stable carbon dioxide(CO2)absorption feature near 4.3 μm was selected as a natural reference.An objective function was constructed to minimize the difference between simulated radiance and sensor-observed radiance.The model innovatively integrated the center wavelength shift(Aλ),the Full Width at Half Maximum(FWHM)scaling factor(s),and the radiometric calibration coefficients(gain k and offset b)into a unified iterative variable space.An adaptive weight vector was built by combining spectral gradient,curvature,absorption depth,and CO2 prior information to strengthen constraints in key spectral regions.An improved Levenberg-Marquardt algorithm was then employed to efficiently and robustly solve this weighted nonlinear least squares problem through multi-parameter synchronous optimization. Results and Discussions Validation using multi-scene field-measured data,including artificial targets and natural terrain,demonstrated that the joint calibration method significantly outperforms traditional stepwise methods.Specifically:1)Joint calibration reduced the spectral center wavelength calibration error to below 10％(with an optimal shift of-0.6 nm),while accurately correcting the FWHM from an initial laboratory value of 20 nm to 21.88 nm.2)For radiometric calibration,the joint method achieved a remarkably low root mean square error(RMSE)of 0.0792 W·m-2·sr-1·μm-1.3)Comparative experiments indicated that the RMSE of the traditional stepwise calibration method was 0.1130 W·m-2·sr-1·μm-1.This demonstrates that the joint calibration reduced the comprehensive error by 29.9％,effectively suppressing parameter coupling effects. Conclusions This study proposes an absorption-feature-weighted method for the spectral-radiometric joint calibration of mid-wave infrared hyperspectral imagers.The constructed joint calibration model integrates the construction of a full-channel weight vector,adjustment of spectral calibration parameters,and optimization of radiometric calibration coefficients.Validation and comparative analysis using multi-scenario field-measured data demonstrate that the joint calibration method significantly enhances overall calibration performance.Specifically,the spectral center wavelength shift was optimized to-0.6 nm,the FWHM was adjusted to 21.88 nm,and the radiometric calibration achieved the lowest reconstruction error(RMSE=0.0792 W·m-2·sr-1·μm-1)with excellent fitting consistency.All these metrics are markedly superior to the results obtained from traditional stepwise calibration.This confirms the effectiveness of the joint calibration in suppressing inter-parameter coupling errors,providing a reliable technical pathway to address the accuracy limitations inherent in traditional stepwise calibration methods.
Objective Trackside inspection of train wheel treads is essential for railway safety.In real scenarios,wheel-tread defects often exhibit large-scale variations and weak visual saliency,while the acquired images are frequently degraded by complex background textures,uneven illumination,and noise.These factors increase missed detections and false alarms,particularly for small and subtle defects.To address these issues,this paper proposes an improved detector,termed MSCF-YOLO,for robust wheel-tread defect detection via multi-frequency feature perception and adaptive noise suppression.Specifically,the proposed method enhances backbone representation learning by integrating star operations into the C3k2 module,strengthens the joint aggregation of low-and high-frequency cues using a multi-frequency feature-space pyramid module,and introduces an SOSim module to amplify responses to small defects while adaptively suppressing background noise and clutter-induced interference. Methods To achieve a favorable trade-off between detection accuracy and real-time performance for typical wheel-tread defects(pitting,spalling,and wear),MSCF-YOLO(Fig.1)is developed based on the lightweight YOLOv11n baseline and incorporates three targeted enhancements.1)C3k2-SCAA(Fig.4)for multi-scale representation.The C3k2 block is restructured by embedding a StarNet-inspired feature mapping together with Context Anchor Attention(CAA)(Fig.2),forming Star-CAA-Block(Fig.3)to replace the original internal bottleneck.This design strengthens cross-layer semantic aggregation and mitigates feature attenuation during multi-scale fusion.2)MBFSPPF(Fig.5)for multi-frequency feature fusion.A Multi-Band Feature Spatial Pyramid Pooling-Fast(MBFSPPF)module is designed by replacing fixed max-pooling with multi-dilation convolutions,which reduces spatial detail loss,enhances scale-aware context modeling,and enables complementary integration of low-frequency semantics and high-frequency details,thereby improving localization stability.3)SOSim(Fig.9)for noise-robust small-defect enhancement.A Small-Object SimAM(SOSim)module is introduced to perform self-adaptive noise suppression,enhancing defect-related high-frequency responses while attenuating background clutter and noise interference,thus improving the effective signal-to-noise ratio for subtle defects. Results and Discussions Experimental comparisons were conducted on the wheel-tread defect dataset under identical settings to demonstrate the effectiveness of the proposed method.The improved MSCF-YOLO was evaluated against representative detectors,including YOLOv5n,YOLOv7-tiny,YOLOv8n,YOLOv10n,YOLOv11n,RT-DETR,Mamba,LSKNet,EfficientViT,RepViT,ConvtextV2 and Swin-Tiny.As summarized in Tab.5,MSCF-YOLO achieves the best overall performance,attaining an mAP@50 of 87.5％,which ranks first among all compared methods.Notably,while delivering high detection accuracy,MSCF-YOLO maintains a compact model size and moderate computational cost,with 3.01 M parameters and 7.9 G FLOPs,which is comparable to YOLOv8n in magnitude.Moreover,MSCF-YOLO surpasses YOLOv8n in per-class detection accuracy and improves precision and recall by 3％and 2％,respectively.Although YOLOv5n provides the most lightweight configuration(1.76 M parameters and 4.9 G FLOPs),MSCF-YOLO leverages the collaborative module design to boost mAP@50 by 4.6％,with per-class gains of 0.3％,13.2％,and 7.1％,respectively.To visually demonstrate the improvement over the baseline,qualitative comparisons before and after the proposed enhancements are presented in Fig.16.The baseline YOLOv11n tends to miss small defects,resulting in frequent false negatives,whereas MSCF-YOLO can reliably detect subtle defects that are missed by the baseline and yields more consistent localization and recognition,particularly for spalling and wear,thereby reducing missed detections and improving overall robustness. Conclusions This paper presents an improved YOLOv11n-based detection framework for railway wheel-tread defect inspection.The proposed method integrates three efficient enhancement modules to improve detection accuracy.To strengthen cross-scale feature acquisition across network layers,StarNet-inspired star computation and Context Anchor Attention are incorporated into the C3k2 block,yielding an enhanced C3k2-SCAA module.In addition,the proposed MBFSPPF replaces the fixed max-pooling operations in the conventional SPPF module with multi-dilation convolutions,and stacks CBAM and CAFM attention mechanisms to reinforce channel-wise fusion of multi-dimensional feature maps while reducing spatial information loss during feature aggregation.Furthermore,an SOSim module is inserted before the detection head to enhance small-defect sensitivity and suppress the adverse impact of background noise.Compared with the baseline YOLOv11n,MSCF-YOLO achieves mAP@50=87.5％and mAP@50-95=50.1％,corresponding to improvements of 3.7％and 4.0％,respectively.It further attains per-class AP@50 scores of 95.7％,78.0％,and 89.0％for the three defect categories,with corresponding gains of 0.8％,7.5％,and 3.0％,respectively.Although MSCF-YOLO increases the parameter count by 16％relative to YOLOv11n,it still maintains a high inference throughput of 131.8 FPS,satisfying real-time requirements.These results demonstrate the practicality and effectiveness of MSCF-YOLO for railway wheel-tread defect detection.Future work will explore knowledge distillation and model pruning to remove redundant weights and further improve detection efficiency while preserving accuracy,thereby reducing missed and false detections in real-world applications.
Objective In the post-Moore's law era,the continuously growing demand for high-performance and low-power information processing has promoted the development of novel computing paradigms.Optical computing features outstanding advantages in parallel processing and energy efficiency,and has been regarded as one of the most competitive technical solutions to break through the bottlenecks of traditional electronic computing.At present,mainstream optical computing systems are mainly based on optoelectronic integrated architectures,which take the advantages of the multi-dimensional and nonlinear response of photons for highly parallel computing tasks,and rely on mature electronic chips for stable and precise regulation.As a critical interface component bridging all-optical computing and all-electronic computing,optoelectronic logic gates are indispensable for the implementation of hybrid computing systems.The development of optoelectronic logic gates with multi-dimensional input compatibility and unified device structure has become an important trend for highly integrated and efficient optoelectronic computing architectures. Methods This paper establishes a photoelectric logic gate with multi-mode input.A few-layer black phosphorus is selected as the optical detection channel(Fig.1),and a symmetric Schottky structure with metal electrodes is used to form the logic gate device.The device undergoes photoelectric testing to verify its logic functions under different input conditions(Fig.3 and Fig.4).The device's anti-interference ability and its potential applications in communication transmission are demonstrated(Fig.5). Results and Discussions The logic performances of the black phosphorus-based optoelectronic logic device under various input modes are systematically studied.Multiple basic logic functions are realized in a single device structure through different input strategies.XOR logic is achieved by all-optical input through high and low light intensity combinations,with an on-off ratio up to 260.AND logic is realized via polarization-programmed input by modulating light intensity and linear polarization angles of 0° and 90°,presenting an on-off ratio of 9.8.XOR and XNOR logic functions are obtained through electro-optical combined input by matching polarized light illumination or light-receiving area with source-drain bias voltage,where the XNOR gate shows an on-off ratio of 41.4.NOT logic is implemented under constant direct-current bias voltage by adjusting the polarization angle,achieving an on-off ratio of 144.OR logic is realized through specific bias and wavelength input by utilizing the wavelength-dependent absorption of black phosphorus at 520 nm and 1550 nm,with an on-off ratio of 14.4.Key operating conditions including zero bias for all-optical Exclusive-OR logic and 0.1 mV working voltage for AND logic are determined.The output current distinction between logic"1"and logic"0"is clearly verified.Detailed measurement procedures,current-voltage curves and logic truth tables are provided in Fig.3 and Fig.4. Conclusions A single-device platform was designed based on few-layer black phosphorus,utilizing its inherent selective absorption characteristics for light intensity,polarization,and wavelength,to achieve optoelectronic logic operations compatible with multidimensional optical inputs.On this platform,five basic logic gates—XOR,OR,AND,XNOR,and NOT—were successfully integrated.Performance tests confirmed its excellent operational capability:the XOR gate achieved an on-off ratio of up to 260,while the relatively more challenging wavelength-based logic computation still reached an on-off ratio of 14.4.Anti-interference tests in optical communication decryption scenarios showed that even with 40％external interference current,the device could still maintain high distinguishability of the output current signal.This work provides a feasible approach to resolving the integration bottleneck of existing optoelectronic logic gates,effectively meeting the demands for highly integrated and reliable optical logic devices in the optical communication field.
Objective The linear motor-driven electromagnetic launch drop tower represents a new generation of microgravity simulation facilities.It overcomes the limitations of free-fall drop towers,such as low microgravity quality and limited experiment cycles,thereby enhancing experimental efficiency and data accumulation.With the increasing demand for longer microgravity durations,it is essential to build taller drop towers has become an inevitable trend.However,engineering implementation faces increasing challenges,including deformation and stability of ultra-tall tower structures and straightness deviations in linear guide rail installation,which can induce vibrations in the drop cabin and adversely affect experimental results.Therefore,establishing high-precision three-dimensional(3D)measuring technology suitable for ultra-tall spatial structures is a fundamental prerequisite to ensure precise guide rail installation and the long-term stable operation of the facility. Methods Conventional structural measurement techniques,such as plumb lines,surveying robots,and terrestrial laser scanners,typically maintain millimeter-level accuracy,which falls out of the stringent sub-millimeter accuracy requirements.To address it,this paper proposes a multi-station laser tracker measurement method that combines high accuracy with wide spatial coverage.First,the measurement principle of the multi-station laser tracker system is introduced based on the coordinate transformation.Next,taking a 20-meter drop tower project as an example,repeatability tests of the laser tracker in vertical space were conducted to verify its reliability under such actual working space.Simultaneously,a 24-hour monitoring of temperature and 3D coordinates was performed using multiple laser tracker stations and weather stations to analyze the correlation between structural point deformation and ambient temperature.Finally,based on the repeatability error of the measured points mentioned above,the measurement accuracy of the multi-station laser tracker method was simulated within the 3D space of a 150 m high microgravity drop tower. Results and Discussions The experimental results demonstrated that the laser tracker achieved a repeatability better than±0.14 mm in single-station measurements over a 20 m vertical range(Tab.2).Deformation observed at the tower's measurement points was primarily driven by temperature fluctuations,with the top of the tower exhibiting a maximum deformation of up to 0.91 mm(Fig.5).Based on the established correlation between relative deformation displacement and temperature variation,it is recommended the ambient temperature fluctuations are limited to±0.25 ℃ during high-precision measurement and installation of the drop tower's precision guide rails(Tab.3).In the simulated multi-station measurement for a 150 m drop tower,the global point error RMS of the common points was 0.11 mm when temperature effects were neglected,while the average and maximum measurement uncertainty were 0.2 mm and 0.56 mm,respectively(Tab.4). Conclusions To meet the demand of high-precision 3D point measurement in ultra-tall vertical structures such as microgravity drop towers,this paper proposes a multi laser tracker station measurement method.Field measurements conducted on a 20-meter drop tower verified the reliability of laser trackers for vertical space metrology.Pre-engineering multi-station tests were also performed to evaluate actual structural deformation and its constraints on precision alignment.Through the integration of experimental data and simulation,the design scheme is subsequently optimized.The findings demonstrate that the multi-station measurement method is suitable for ultra-tall vertical structures—enabling sub-millimeter level alignment and stability control for drop tower facilities.Moreover,the work provides reliable practical experience for precision surveying and alignment in similar high-precision projects,such as ultra-tall buildings,large-scale scientific installations,and aerospace launch structures.For future research,the deployment scheme will be optimized to address line-of-sight obstructions by integrating site-specific conditions,such as spatial constraints,structural occlusions,and environmental vibrations.Furthermore,a practical and implementable measurement protocol will be developed to realize the optimal measurement strategy for real-world engineering applications.
Objective This study investigates the transmission characteristics of Laguerre-Gaussian(LG)vortex beams carrying orbital angular momentum(OAM)in underwater environments,with a particular focus on the effects of different turbulence intensities on beam quality and stability.By employing an improved Nikishov model that incorporates low-frequency subharmonic compensation,this research validates the transmission characteristics of LG beams with various topological charges(l=1,2,3,4)over a transmission distance of 0-40 meters.The analysis of light intensity,phase distribution,and spot drift index aims to provide theoretical support and experimental guidance for optimizing underwater optical communication systems. Methods An improved Nikishov model incorporating low-frequency subharmonic compensation is employed to generate accurate oceanic phase screens.The multi-phase-screen method,combined with angular spectrum propagation,is utilized to simulate the propagation of LG vortex beams through turbulent ocean conditions.The study evaluates the impact of turbulence intensity on beam drift and peak signal-to-noise ratio(PSNR)by comparing the results under weak,moderate,and strong turbulence conditions. Results and Discussions The results indicate that the beam drift and transmission instability increase with higher turbulence intensity and topological charge.Under weak turbulence conditions(Cn2=9.0×10-13,σR2=0.1),the root mean square(RMS)value of the spot centroid drift for l=1 is 0.098 mm,while under strong turbulence(Cn2=1.3×10-11,σR2=1.5),the RMS value for l=4 increases to 0.473 mm.The PSNR of the LG beams decreases linearly with transmission distance,with a higher rate of decline in stronger turbulence.For instance,under weak turbulence,the PSNR for l=1 drops from 48.96 dB to 27.39 dB over 40 meters,whereas under strong turbulence,it decreases from 37.68 dB to 14.87 dB.Research indicates that low-order LG beams are more suitable for short-range underwater optical communication. Conclusions This study refines the Nikishov phase screen model via low-frequency compensation,reducing the phase structure function deviation from 12.6％to 3.8％.Simulations demonstrate that LG beams with topological charge l=1 exhibit optimal turbulence resilience over 40 m coastal propagation(PSNR:28-45 dB,RMS drift:0.098 mm),whereas l=4 beams destabilize under strong turbulence(σR2=1.5).These findings validate low-order OAM modes as preferred candidates for robust short-to-medium range underwater optical communication links.
Significance The multi-layer and multi-interface composite structures widely adopted in aircraft thermal protection and load-bearing components are prone to various hidden defects including structural folds,tiny pores,material inhomogeneity,interlayer delamination and interface disbond during manufacturing and service operation.These subtle defects seriously affect the structural stability and service safety of aircraft equipment.Single non-destructive testing technologies exhibit prominent limitations in defect identification accuracy,applicable material scope and visualization capability,which fail to achieve full-scale and high-precision detection of multi-type defects in complex aircraft structures.Existing multi-technology fusion detection methods mostly rely on dual-technology combination modes,lacking systematic three-dimensional(3D)reconstruction and global integration strategies for multi-source defect data.Therefore,it is urgent to develop an efficient and accurate multi-physical field fusion imaging detection technology to realize precise identification and intuitive visualization of multi-layer structural defects,which is of great engineering significance for ensuring aircraft structural integrity. Progress A novel 3D fusion imaging technology based on the multi-physical field response mechanisms of laser ultrasonics,X-ray detection and infrared thermography is developed for aircraft multi-layer structure defect detection(Fig.1).According to the differentiated material characteristics of aluminum alloy,composite material and thermal protection structure,as well as the morphological characteristics of typical structural defects,the optimal application scope of each single detection technology is quantitatively divided(Tab.l).Laser ultrasonic technology is applied to identify internal folds and tiny pore defects of aluminum alloy structural layers,X-ray detection is adapted to capture inhomogeneous defects of thermal protection structures,and infrared thermography is utilized to detect interlayer delamination and interface disbond defects of composite materials.On this basis,a complete technical system covering defect feature extraction,high-precision positioning and cross-source coordinate transformation is constructed.Combined with marker positioning,neural network intelligent recognition and point cloud generation algorithms,the unified global coordinate calibration of heterogeneous defect data from three detection approaches is realized,and a multi-source defect point cloud 3D integrated reconstruction model is established to complete the overall visual reconstruction of multi-layer and multi-interface structural defects. Conclusions and Prospects Experimental verification shows that the proposed technology can stably identify micro-defects with diameters ranging from 0.1 mm to 0.3 mm and depths of 0.1 mm to 0.15 mm,with the maximum spatial positioning error of all detected defects controlled within 0.2 mm.This technology effectively integrates the complementary advantages of three mainstream detection methods,thoroughly compensates for the single detection technology's defects of limited detection range and fuzzy feature identification,and solves the technical bottlenecks of cross-coordinate system matching and global visual reconstruction of multi-layer complex structural defects.The established 3D reconstruction model can intuitively and accurately present the spatial distribution rules of various hidden defects in aircraft multi-layer structures.Future research will focus on optimizing the synchronous acquisition efficiency of multi-source detection data and improving the environmental adaptability of the technology under complex working conditions and extreme service environments,so as to further expand its engineering application scope in aircraft full-life-cycle structural health monitoring and integrity assessment.
Objective With the rapid development of the new generation of infrared detector technology in China and the urgent demand for weak light signal detection and other fields,APDs with high sensitivity,high gain and low excess noise have become a research hotspot in the field of infrared technology.Compared with other materials,HgCdTe has a unique energy band advantage.The difference in collision ionization coefficients between holes and electrons can be very large.It is the only material to date that can achieve a collision ionization coefficient ratio of 0(or ∞)and single-carrier excitation by adjusting the Cd component.At present,the performance characterization of HgCdTe APD devices in China mainly focuses on gain characteristics,and the conclusion is consistent:a higher avalanche gain can be achieved by using a higher operating bias voltage or a narrower PN junction depletion region width.However,if the operating bias voltage of the device is blindly increased or the width of the depletion region is reduced,the noise current caused by the device's tunneling current will also increase significantly,ultimately leading to a decrease in the maximum useful gain,a severe deterioration in the signal-to-noise ratio,and a limitation on the detection sensitivity.Therefore,while achieving high gain,it is also necessary to take into account the impact of noise caused by gain on the signal-to-noise ratio of the device.This paper will study the avalanche gain,noise and other characteristics of linear HgCdTe APD devices with different injection areas,explore the mutual restraint relationship between the gain and noise characteristics of different injection areas,and comparatively analyze the gain and noise characteristics of APD unit devices and focal plane devices with the same injection area. Methods Firstly,a P-type Hg vacancy doped HgCdTe epitaxial layer was grown on a single-layer CdZnTe substrate.Secondly,a high-doping concentration N+region is formed on the surface of the epitaxial layer through B ion implantation;And during the annealing process after injection,an N-region with a relatively low doping concentration is formed;Finally,APD unit device chips are prepared through photolithography,etching,deposition,etc.Then,APD unit devices with different injection areas can be obtained by interconnecting and extracting the structure(Fig.1).In addition,a HgCdTe APD focal plane device with a pixel size of 320×256 and a pixel center distance of 30 μm was fabricated through the same preparation process.In addition,to meet the requirements of experimental testing,gain and noise test devices for HgCdTe APD unit devices,as well as performance test devices for HgCdTe APD focal plane devices,were respectively established(Fig.2,Fig.3,Fig.4). Results and Discussions The paper first studies the gain characteristics of HgCdTe APD unit devices with different injection areas.The research found that under the same reverse bias,as the injection area increases,the gain shows a decreasing trend(Fig.6).To verify the accuracy of the results,the gain of device No.4 was fitted respectively through the Kinch empirical formula and the Rothman empirical formula(Fig.7).Secondly,the noise characteristics of HgCdTe APD unit devices with different injection areas were studied.This section first studies the noise frequency curves of HgCdTe APD unit devices with different injection areas.Before the corner frequency,it is manifested as the 1/f noise of the device,and after that,it is mainly manifested as the shot noise of the device.The results show that as the injection area increases,both the 1/f noise and the shot noise will gradually increase,and the corner frequency will gradually shift backward(Figure 8);In addition,in order to study the noise frequency characteristics of devices under different reverse deflections,the No.4 device was investigated(Fig.9).It can be seen that as the reverse bias voltage of the device increases,the frequency-independent shot noise component of the device is also amplified with the internal gain.Finally,the variation relationship of the excess noise factor of a single component with reverse bias in different injection areas was studied(Fig.10).The results show that under the same reverse bias,as the injection area increases,the excess noise factor of the device also increases accordingly.Finally,the gain and noise characteristics of mercury cadmium telluride APD focal plane devices with an injection area of 15 μm×15 μm were studied in this paper,and the results were compared with those of unit devices with the same injection area(Fig.11,Fig.12,Fig.13). Conclusions In this paper,HgCdTe APD devices with different injection planes were fabricated by liquid-phase epitaxy technology,and the gain and noise characteristics of the fabricated unit devices and focal plane devices were compared and studied.On the one hand,under the high anti-bias conditions of 77 K and-12 V,the gain of device No.4 is greater than 2500.In the range of-10 V to 0 V,the low reflection is slightly lower,and the excess noise is between 1.0 and 1.5.In addition,the pixel size prepared in this paper is 320×256,and the pixel center distance is 30 μm.The average avalanche gain of the HgCdTe focal plane device at 77 K and-8 V is 108.86,the gain non-uniformity is 5.67％,and the average excess noise factor is better than 1.245.On the other hand,under the same bias voltage,HgCdTe APD devices with a larger injection area have a lower avalanche gain,a higher excess noise factor,and a larger 1/f noise corner frequency.Among them,the relatively low gain can be explained as the reduced contribution of edge gain,while the possible reasons for the relatively high noise are that after the injection area increases,the randomness of the photogenerated carrier collision ionization process increases.In addition,HgCdTe device itself is a defective semiconductor.After the area increases,more defects are included,and the traps assist tunneling current,surface leakage current,etc.will all increase.Therefore,for focal plane devices with a readout circuit to limit the maximum bias voltage,choosing a diode with a smaller injection area can achieve better performance.However,a compromise should also be made.If the injection area is too small,it will cause the pixel duty cycle to be relatively low,resulting in a smaller fill factor of the focal plane device and a reduction in the quantum efficiency of the device.
Objective Carbon Fiber Reinforced Polymer(CFRP)is susceptible to damage defects such as fiber breakage,matrix cracking,and delamination under low-velocity impact loading.Efficient and reliable non-destructive testing methods are crucial to ensuring the structural safety of CFRP components.Infrared thermal wave imaging detection technology has the advantages of non-contact operation,rapid imaging,and large-area inspection.However,the detection images are easily interfered by uneven heating,uneven emissivity,and background noise,resulting in problems such as low signal-to-noise ratio and blurred defect edge features in the detection images.To address the above issues,this study proposed an adaptive image enhancement method integrating Chirp correlation feature extraction and Block-Matching and 3D filtering(BM3D),which provides new insights for the research on adaptive enhancement technology of infrared thermal wave imaging for CFRP impact damage. Methods A Chirp-modulated infrared thermal wave detection system was constructed(Fig.1),and a drop-weight impact testing machine was adopted to prepare CFRP impact damage specimens with an impact energy of 23 J(Fig.2).The Chirp correlation algorithm was utilized to extract thermal wave phase features to suppress the interference from uneven emissivity.Based on the local noise statistical characteristics of the feature image,an adaptive noise reduction threshold function for the BM3D algorithm was constructed to realize dynamic adaptive enhancement for images with different noise intensities.Gaussian noise of different intensities was added to the normalized Chirp correlation phase feature images,and the defect signal-to-noise ratio(DSNR),peak signal-to-noise ratio(PSNR),and structural similarity index measure(SSIM)of the adaptive BM3D algorithm were compared and analyzed to quantitatively evaluate the comprehensive performance of the proposed adaptive enhancement algorithm. Results and Discussions The research results showed that compared with Fourier spectrum feature images,the Chirp correlation phase feature images exhibited superior imaging quality,and the uneven emissivity in the optimal defect signal-to-noise ratio image was effectively suppressed(Fig.3).Figure6 presented the noise reduction effects of median filtering,Wiener filtering,and the adaptive BM3D algorithm on Chirp correlation phase feature images under different noise intensities.As the noise intensity increased,the noise reduction capabilities of traditional median filtering and Wiener filtering decreased significantly,leading to blurred defect edge features.Figure7-Figure9 illustrated the comparison results of DSNR,PSNR,and SSIM of the three algorithms under different noise intensities.It could be observed from the experimental results that as the normalized noise intensity increased,the DSNR,PSNR,and SSIM of the median filtering and Wiener filtering algorithms showed a significant downward trend,while those of the adaptive BM3D algorithm did not present a significant decline,with the overall performance significantly outperforming the traditional median filtering and Wiener filtering algorithms.This indicates that the adaptive BM3D algorithm can achieve effective noise reduction for the Chirp correlation phase map of CFRP impact damage defects,and possesses excellent edge feature retention capability. Conclusions Aiming at the adaptive enhancement requirements of infrared thermal wave imaging detection images for CFRP impact damage defects,this paper proposed an adaptive image enhancement method integrating Chirp correlation feature extraction and BM3D filtering.The research showed that under the strong noise condition with normalized noise standard deviation σ=0.2,the defect signal-to-noise ratio(DSNR)of the adaptive BM3D enhancement method proposed in this paper reached 44.41,the peak signal-to-noise ratio(PSNR)was 32.25 dB,and the structural similarity(SSIM)was 0.87.All indicators were significantly better than those of the traditional median filtering and Wiener filtering algorithms,verifying the effectiveness and stability of the proposed method in noise reduction and defect edge feature preservation in strong noise environments.This research provides new ideas for the research on adaptive enhancement technology of infrared thermal wave imaging for CFRP impact damage,and can provide reliable technical support for the engineering application of this technology.
Objective Existing infrared polarization image enhancement methods have made some progress in cloud suppression,background removal,and image quality improvement.However,there's still room for further research in restoring infrared polarization images under complex cloud backgrounds.On one hand,some methods mainly rely on features like Stokes parameter combinations,degree of polarization,or polarization angle,with relatively limited consideration of the physical differences between cloud scattering background and target radiation.As a result,in cases with complex cloud textures,strong background non-uniformity,or low local contrast,there can still be issues with cloud residue or insufficient detail recovery.On the other hand,as infrared polarization imaging systems gradually expand to mobile platforms,embedded devices,and portable terminals,algorithms need to balance cloud removal effectiveness with computational complexity,storage overhead,and real-time processing capability.The focus here is on the problem that detecting small infrared targets under complex sky backgrounds is easily interfered with by cloud clutter,leading to high false alarm rates. Methods First,the concept of dark channel prior is introduced,and a pseudo-dark channel statistic suitable for infrared polarization single-channel images is constructed.A pseudo-dark channel statistic based on the local minima of a single channel is developed to estimate the cloud scattering component in infrared sky scenes,providing a basis for subsequent atmospheric scattering restoration.Then,a framework for infrared polarization cloud removal combining polarization decomposition and atmospheric scattering models is established.To address the issue of coupling between cloud scattering components and effective radiative information in infrared polarization images under complex cloud backgrounds,polarization imaging characteristics are integrated with an atmospheric scattering degradation model.Polarization decomposition is employed to separate and enhance polarization-related information in the images,obtaining an initial representation of effective information;the atmospheric scattering model is then used to estimate and suppress the residual cloud background.Finally,a guided filtering optimization method introducing a planarity constraint is designed.For complex cloud edge regions where transmission estimation is easily affected by local texture variations and sudden luminance changes,a planarity constraint parameter is incorporated into the regularization term of guided filtering to adaptively adjust the smoothing degree in different regions.This method helps improve the spatial consistency of transmission estimation in locally flat areas and reduces over-smoothing in areas with structural changes,thereby enhancing the balance between cloud suppression and image detail preservation to a certain extent. Results and Discussions In complex scenarios with overlapping cloud layers,the contrast(C)and local signal-to-noise ratio(LSCR)of the proposed algorithm are significantly superior to mainstream algorithms such as PFE and DPI.Meanwhile,the algorithm's single-frame runtime is only 0.036 s,providing high-quality input for subsequent spatial-domain target detection tasks. Conclusions To address the problem of infrared small targets being easily disturbed by cloud clutter in complex sky backgrounds,an infrared polarization cloud removal algorithm based on an atmospheric scattering physical model is proposed.This method combines polarization decomposition with atmospheric scattering modeling,utilizes a pseudo-dark channel prior to estimate the cloud scattering component,and optimizes the transmittance through improved guided filtering,thereby suppressing cloud clutter while preserving target edge information.Experimental results show that the proposed method demonstrates good target enhancement and background suppression capabilities under complex cloud backgrounds.Its contrast and local signal-to-noise ratio outperform methods such as PFE and DPI,while also exhibiting favorable real-time performance.This method can provide high-quality inputs for subsequent infrared small target detection.Future research will further investigate its adaptability under complex weather conditions such as fog and rain,and deployment verification will be conducted on embedded platforms.
Objective Accurate identification of the coal-rock interface is a critical step in achieving intelligent coal mining and ensuring safe and efficient production.During the coal mining process,the position of the coal-rock interface is directly related to the rationality of the cutting path and the operational safety of equipment.Inaccurate interface identification can easily lead to cutting into rock,increased equipment wear,and even safety accidents.At the same time,it can also reduce coal recovery rates and production efficiency.Therefore,achieving high-precision identification of the coal-rock interface is of great engineering significance.To address issues such as the inconspicuous differences in coal-rock characteristics,the poor real-time performance of traditional detection methods,and their limited recognition accuracy,a coal-rock interface perception and identification method integrating deep learning and thermal imaging technology is proposed.By exploiting differences in thermal responses between coal and rock,high-precision coal-rock interface identification is achieved,providing reliable technical support for autonomous decision-making and precise control in intelligent coal mining equipment. Methods Simulated coal-rock specimens were prepared(Fig.10),and an active infrared coal-rock sensing platform for the mining face was established based on the FLIR A655SC infrared thermal imager(Fig.13).A LabVIEW-based system and halogen lamps were used to regulate and excite the light source.Coal-rock thermal imaging and image acquisition were conducted under different excitation parameters.Composite data augmentation methods,such as flipping and noise addition,were applied to the collected data for expansion,thereby constructing a coal-rock thermal radiation image dataset(Fig.14).Meanwhile,an ITR-DeepLabV3+network model(Fig.2)is proposed for the characteristics of coal-rock thermal radiation images.A lightweight MobileNetV4 is adopted as the backbone,reducing computational complexity while maintaining high segmentation accuracy.The convolutional block attention module(CBAM)is introduced to enhance feature representation and improve the model's feature extraction capability.By combining depthwise separable convolution with the SE attention mechanism,a DS-ASPP module is proposed to reduce model complexity and strengthen multi-scale feature modeling.In addition,features with 4×,16×,and 32× downsampling from the backbone are utilized,and a shallow multi-scale feature fusion enhancement module is constructed by integrating the coordinate attention(CA)mechanism,enabling effective complementarity of multi-level feature information. Results and Discussions Ablation experiments were conducted on the proposed modules,and the results show that all the proposed modules effectively enhance model performance while balancing segmentation accuracy and model complexity(Tab.2,Tab.3,Tab.4).The proposed ITR-DeepLabV3+model achieves an Intersection over Union(IoU)of 91.87％,a Mean Intersection over Union(MIoU)of 94.98％,a Mean Pixel Accuracy(MPA)of 97.34％,and an F1-score(F1)of 95.76％,with 4.872 1M parameters(Params)and FLOPs of 26.270 7G.Compared with the DeepLabV3+model employing Xception as the backbone,the proposed model achieves improvements of 1.23％,0.76％,0.70％,and 0.67％in IoU,MIoU,MPA,and F1,respectively,while reducing the number of params from 29.024 3M to 4.872 1M and the FLOPs from 121.136 1G to 26.270 7G.Compared with the DeepLabV3+model employing MobileNetV2 as the backbone,the proposed model achieves improvements of 3.86％,2.38％,1.97％,and 2.14％in IoU,MIoU,MPA,and F1,respectively,while reducing the number of params from 5.813 3M to 4.872 1M,with FLOPs remaining nearly unchanged.Meanwhile,comparative experiments with classical semantic segmentation models were conducted.The results indicate that,considering both accuracy and efficiency,the ITR-DeepLabV3+model demonstrates superior overall performance compared to conventional network models(Tab.5). Conclusions Active infrared thermography is employed to capture the coal wall to be cut,a coal-rock dataset is constructed and augmented,an ITR-DeepLabV3+network model is proposed,and both ablation and comparative experiments are conducted.The experimental results demonstrate that the proposed modules yield significant gains in both accuracy and efficiency.Furthermore,the proposed ITR-DeepLabV3+model outperforms classical networks,including PSP-Net,U-Net,and HR-Net,thereby enabling stable and reliable identification of the coal-rock interface.In conclusion,the proposed method integrating deep learning and active infrared thermography provides effective technical support for accurate coal-rock interface perception under complex operating conditions and is of significant importance for enhancing the safety and operational efficiency of coal mining.
Objective Flip-chip technology is widely used in advanced microelectronic packaging due to its high signal transmission speed,low interconnection loss,and compact structure.However,with the continuous miniaturization of solder bumps and their operation under harsh conditions,internal defects such as missing solder bumps seriously threaten packaging reliability.Laser ultrasonic testing based on an optical microphone is an emerging non-contact nondestructive technique with high sensitivity and wide bandwidth.This study evaluates the feasibility of applying all-optical laser ultrasonic testing for high-resolution detection of solder bump defects in flip-chip structures. Methods A three-dimensional thermo-mechanical coupled finite element model containing missing solder bumps is established to investigate the effects of defects on temperature and displacement fields.Simulated flip-chip specimens with missing solder balls(600 μm diameter and 400 μm height)are fabricated and inspected using an all-optical laser ultrasonic system based on an optical microphone.Laser ultrasonic A-,and C-scan imaging experiments are conducted.A polynomial fitting coefficient method is proposed for C-scan data processing. Results and Discussions Simulation results show that missing solder bumps cause significant variations in temperature and displacement fields,which affect laser-induced ultrasonic responses.Experimental results demonstrate that the proposed system achieves high-resolution imaging of missing solder bumps with a diameter of 600 μm.Clear distinctions between normal solder bump regions and defect regions are observed in A-and C-scan images(Fig.6 and Fig.7).The proposed polynomial fitting method provides the highest signal-to-noise ratio,achieving a 5.26-fold improvement over the original signal and outperforming PCA and ICA methods(Fig.8). Conclusions Both simulation and experimental results demonstrate that all-optical laser ultrasonic testing based on an optical microphone enables non-contact,high-resolution detection of missing solder bump defects in flip-chip packaging,providing a promising approach for reliability inspection of advanced microelectronic interconnections.
Objective In unstructured scenarios such as nighttime operation,underground environments,disaster rescue,and dusty or smoky industrial sites,imaging-based vision is severely degraded by insufficient illumination and scattering media,leading to reduced recognition confidence or even failure.Tactile perception is illumination-independent and can provide reliable local mechanical cues,yet it is intrinsically limited in spatial coverage.Therefore,hardware-level visuo-tactile fusion with low power consumption is highly desirable for robust perception in low-light environments. Methods A self-powered biomimetic neuromorphic system for near-infrared(NIR)vision-tactile fusion perception was developed,in which a flexible triboelectric nanogenerator(TENG)was integrated with a floating-gate optoelectronic synaptic device(Fig.1).In the optoelectronic synaptic device,PbS quantum dots were introduced as an NIR-absorbing photosensitive charge-trapping layer,enabling NIR photoexcitation to be converted into retainable trapped charges in the floating gate.To realize self-powered tactile input,the TENG was fabricated based on a flexible PDMS elastomer and Ag nanowire electrodes,and a PDMS/MXene composite triboelectric layer was incorporated to generate high-amplitude voltage pulse.The TENG output pulses were directly used as gate-modulation signals for the synaptic device to enable event triggering and current-based weight updates. Results and Discussion The device performance was systematically compared under dark conditions and under 850nm near-infrared illumination,with particular emphasis on key neuromorphic metrics including bidirectional voltage-sweep hysteresis,excitatory postsynaptic current(EPSC)responses,and the linearity of current-based weight updates.The results show that NIR stimulation increases the on-state current by approximately 259％while reducing the off-state current by about 68％(Fig.3(a)),thereby significantly expanding the current dynamic range.Under multi-pulse excitation,the EPSC peak rises by about 81％and the nonlinearity decreases by about 50％(Fig.4(d)),demonstrating NIR-enhanced synaptic responses and improved linearity of conductance modulation.Meanwhile,upon tactile stimulation,the triboelectric nanogenerator(TENG)delivers a short-circuit current of 4.7 μA,an open-circuit voltage of 75 V,and a transferred charge of 37 nC(Fig.2(d)),which can reliably support repeated weight-update operations. Conclusions By co-designing a PbS-QD floating-gate infrared optoelectronic synaptic device and a flexible TENG,this work demonstrates a self-powered biomimetic multimodal neuromorphic system capable of NIR"visual"sensing,tactile self-powered input,and hardware-level spatiotemporal fusion with synaptic plasticity.The proposed SP-BMNS offers a feasible device-system paradigm for robust low-light perception and edge intelligence,with potential applications in low-illumination target recognition,robotic electronic skin,wearable human-machine interaction,and ambient interactive sensing nodes.
Objective The endoscope,a key instrument for target observation,information analysis,and operation execution in confined spaces,offers non-contact and non-invasive capabilities.It is widely used in clinical evaluations of the digestive tract and in the inspection and diagnosis of confined environments such as industrial pipelines via nondestructive testing.However,non-uniform illumination from the endoscopic light source,combined with the smooth surfaces of human organs or pipelines,often causes specular reflection.This leads to glare in the endoscopic field of view,which obscures target details and severely compromises imaging quality and diagnostic accuracy.Traditional glare suppression methods primarily rely on RGB three-channel cameras for image processing.However,these approaches are computationally intensive and limited to only three spectral bands,thereby constraining the available spectral information.As a result,they are inadequate for accurate online analysis of histochemical component differences at specific wavelengths.To overcome these limitations,this study proposes a novel glare elimination method that integrates endoscopic hyperspectral imaging with multi-exposure fusion. Methods To effectively suppress glare in hyperspectral images while preserving their rich spectral and spatial information,this study first analyzes the image and spectral characteristics of the central glare region and the peripheral non-glare region.Based on this analysis,a threshold mask is constructed to identify abnormal pixels in the hyperspectral data cube—such as those exceeding the camera's saturation threshold or exhibiting low overall signal intensity—for subsequent removal.Then,the maximum intensity of each pixel across the entire spectral band is calculated,and the spectrum of each pixel is normalized.Finally,the spectral intensity curves of the remaining pixels in each image are averaged and fused.For comparative purposes,three existing glare suppression methods were selected.Image quality was evaluated using the peak signal-to-noise ratio(PSNR)and structural similarity(SSIM),while spectral fidelity was assessed based on the mean absolute error(MAE),root mean square error(RMSE),and spectral angle mapper(SAM). Results and Discussions The endoscopic hyperspectral imaging system captured nine hyperspectral images at different exposure times.Three RGB bands(640,532,and 471 nm)were selected to generate pseudo-color images for visualization.At high exposure times,the central area reached the maximum intensity value of 4095,indicating overexposure and a consequent loss of detail in these bright regions.In contrast,the peripheral areas exhibited higher intensity values and retained more structural detail(Fig.2).As the exposure time decreased,the overexposed area in the center gradually diminished,allowing previously obscured details to be recovered.However,noise and burrs became more pronounced in the periphery(Fig.3),which could obscure fine details and required appropriate removal.After normalizing each pixel,the overall contrast of the hyperspectral images was significantly enhanced.Details that were previously lost in the glare region at high exposure times,as well as those obscured in the peripheral regions at low exposure times,were effectively restored(Fig.4).Finally,in comparison with other methods,the hyperspectral image obtained through average fusion demonstrated superior clarity and contrast in structural details across the entire field of view.It also achieved relatively high PSNR and SSIM scores(with the latter approaching 1,as shown in Tab.1),indicating that the structural information was highly consistent with the reference image.Moreover,the proposed method offered faster computational speed.In terms of spectral fidelity,our method exhibited a higher similarity to the original spectrum in both the central and peripheral regions(Fig.6).Quantitative evaluations based on different spectral similarity metrics(Tab.2)confirmed that our method achieved the lowest index values across different positions,demonstrating its superior spectral reconstruction accuracy.This approach not only effectively suppressed central glare but also preserved structural and spectral information at the edges. Conclusions This study tackles the challenge of glare in endoscopic imaging by introducing a novel elimination method based on hyperspectral and multi-exposure image fusion.Qualitative and quantitative comparisons with existing techniques demonstrate that the proposed approach not only effectively suppresses specular reflections but also enhances textural details across the entire field of view.Moreover,it exhibits superior performance in preserving spectral fidelity.This work provides a valuable tool for the spectral analysis of target surfaces in confined spaces and offers a promising solution for accurate,online tumor diagnosis.
Significance Pipelines are fundamental infrastructure for energy transportation,and their safety is directly related to industrial production and environmental protection.Corrosion remains a primary cause of pipeline failure,highlighting the importance of efficient and reliable non-destructive testing technologies.Infrared thermography is a representative non-destructive testing technique.Based on the presence or absence of an external excitation source,it can be categorized as passive or active thermography.Passive thermography relies on the self-emission of the monitored target,while active thermography introduces external excitation to create thermal contrast,making it more suitable for the early detection of corrosion and cracks in industrial pipelines.Among various active methods,eddy current pulsed thermography combines the deep penetration of electromagnetic induction with the high-resolution visualization of infrared imaging,demonstrating strong potential for pipeline corrosion detection.This technology possesses significant advantages such as non-contact operation,high detection speed,high sensitivity,and the ability to perform inspections through non-conductive insulation layers,becoming a key means of ensuring pipeline integrity. Progress The mechanism and components of the eddy current pulsed thermography testing system in pipeline applications are first introduced.The system primarily consists of a high-frequency induction heating power supply,an induction coil,an infrared thermal imager,and synchronization control and data processing units.During testing,the induction coil generates eddy currents in the metallic pipeline wall.The presence of corrosion or cracks distorts the eddy current distribution,resulting in localized heat accumulation.The technology is then systematically classified according to coil configuration and detection mode.To address the cylindrical structure of pipelines,researchers have developed arc array coils and flexible coils to enhance magnetic field coupling.Based on the excitation signal,the technique can be divided into pulsed excitation,lock-in modulation,and phase-based modes.Since this technology was applied to pipeline inspection,research has progressed from basic defect detection to precise characterization under complex conditions.To address challenges such as insulation layer interference,lift-off effects,and non-uniform surfaces,extensive innovation has taken place.For insulated pipelines,corrosion under insulation can now be effectively identified by optimizing excitation frequency and increasing power density.In quantitative evaluation,approaches such as principal component analysis,skewness analysis,and wavelet transform have been employed to extract thermal features,enabling accurate measurement of pipeline wall thinning.Additionally,convolutional neural networks integrated with deep learning have been utilized to automatically classify pitting and uniform corrosion,greatly improving the intelligence level of detection.Currently,this technology is widely used for the routine inspection of long-distance oil and gas pipelines,power plant boiler tubes,and urban heating networks. Conclusions and Prospects Currently,investigations on complex irregular sections,such as pipeline elbows and tees,remain limited,and interference from variations in surface emissivity still requires further mitigation.Most research is focused on qualitative identification in laboratory conditions;quantitative accuracy and three-dimensional morphology reconstruction in real engineering environments still require significant improvement.The size and power demands of testing equipment also restrict the automation of long-distance pipeline inspection using eddy current pulsed thermography.With the development of intelligent manufacturing and the Industrial Internet,pipeline corrosion detection is undergoing a transformation from manual interpretation to automated evaluation.Eddy current pulsed thermography technology is expanding from simple surface damage identification to the quantitative detection of internal wall thinning,and from two dimensional thermal image analysis to three dimensional defect reconstruction.In addition,the development of lightweight induction heating devices with high environmental adaptability and deep learning algorithms integrated with physical models will provide strong support for achieving high-precision quantitative evaluation and life prediction of pipeline corrosion.
Objective In recent years,as ground-based telescopes continue to advance toward larger apertures,wider fields of view,and higher imaging quality,the use of large-aperture corrector has become increasingly common.The MUST(MUltiplexed Survey Telescope)proposed by Tsinghua University is equipped with a wide field corrector(WFC),which consists of five lenses with a maximum aperture of 1.6 m.The WFC ranks among the most advanced internationally in terms of both size and complexity.To ensure the imaging quality of the telescope system,the alignment requirements for the WFC are extremely stringent:the decentering must be better than 20 μm and the tilt must be better than 9".These demands pose significant challenges for optical alignment of the WFC.Currently,there is no centering instrument of such large aperture and high precision to guide the optical alignment of WFC of this scale.Therefore,the characteristics of MUST corrector is analyzed and an ultra-large-aperture centering instrument is designed in this paper. Methods To meet the system precision requirements,a centering instrument based on reflective eccentricity is adopted in this paper(Fig.2).A phased alignment process based on this large-aperture centering instrument is proposed,which decouples the alignment degrees of freedom for individual lens,significantly reducing the complexity of system alignment(Fig.3).To fulfill the high-precision alignment demands of the WFC,the contributions of runout of turntable and straightness of guideway to measurement uncertainty are analyzed.In response to the large decentration caused by tilted mirrors with long radius of curvature,the requirements for the field of view of centering instrument are examined,and an alignment method for wedge prism pairs is proposed(Fig.5).Additionally,a selection strategy for a fixed-focus converging lens suitable for full-process imaging is presented(Tab.3). Results and Discussions An air-bearing turntable with aperture of 2 m is selected,with main error source of runout.Considering the difficulty of regrinding,flatness of the turntable should be better than 20 μm,while radial and axial runouts are required to be better than 0.5 μm each.A precision linear guide with a 3-meter range is chosen,which requires straightness better than 10 μm over its full travel range.The alignment process involves up to 22 different focal lengths of converging lenses.During actual alignment,adjusting the object distance via the guide allows clear imaging within a certain range,thereby reducing the number of lenses to 10(Tab.3).L3-B and L4-F are tilted spherical surfaces,which inducing decentrations of 29.24 mm and 21.49 mm,respectively.An industrial camera with a sensor size of 8.44 mm×7.06 mm is chosen to meet both the field-of-view and imaging quality.To ensure proper relative rotation of these wedge lenses,an alignment method is designed to mark the orientation of the tilt(Fig.5). Conclusions To meet the high-precision alignment requirements of the ultra-large WFC of MUST,this study investigates the measurement principles of reflective eccentricity and proposes a phased alignment process.This study provides specific engineering implementation recommendations to ensure that the measurement uncertainty of a 2-meter-class centering instrument remains better than 2 μm.This research forms a complete technical solution encompassing theoretical modeling,alignment processes,and error control,offering an implementable theoretical and engineering framework for the high-precision alignment of the MUST WFC.It demonstrates practical engineering application value and broader significance for potential adoption in related fields.
