This paper presents a lightweight deep learning approach for detecting small leaks in buried gas pipelines using a phase-sensitive optical time domain reflectometry (Φ-OTDR) system. Vibration signals acquired from a distributed optical fiber vibration sensor (DOFVS) are converted into time–frequency images via continuous wavelet transform (CWT) to facilitate joint spatiotemporal feature extraction. An improved model that integrates a lightweight MobileNetV3 backbone with a bidirectional LSTM (BiLSTM) module is proposed to capture both spatial patterns from individual CWT images and temporal dependencies across consecutive frames. Experimental results demonstrate an overall accuracy of 95.96% in classifying four leak pressure levels, including the reliable detection of leaks as small as 1/16 inch under pressures as low as 0.1 MPa. Compared to conventional CNN models, an improved recognition accuracy of 8%–12% is demonstrated for the identification of low-pressure and small-aperture leakage conditions.

A novel broadband, high-resolution spectroradiometer based on a virtually imaged phased array (VIPA) has been developed to meet the stringent requirements for high-precision vertical profiles of key atmospheric constituents. The instrument operates within the spectral range of 7535-7680 cm-1, achieving a spectral resolution of 0.023 cm-1 (690 MHz) with an integration time as short as 400 ms. Using this system, high-resolution atmospheric transmittance spectra in the 7535-7680 cm-1 band were successfully measured, and the vertical profile of water vapor was retrieved using the optimal estimation method (OEM). The results demonstrate that the novel VIPA-based spectroradiometer enables rapid acquisition of high-resolution atmospheric transmittance spectra over a broad spectral range, providing a robust and effective new approach for remote sensing of key atmospheric constituents. To the best of our knowledge, this represents the first successful application of a VIPA-based spectroradiometer for retrieving vertical profiles of atmospheric constituents.

This paper proposes a fibre-optic Fabry-Perot (F-P) temperature sensor based on the thermo-mechanical deflection effect of a bimaterial helical beam. The sensor exploits the difference in thermal expansion coefficients between polyimide and aluminium layers to induce deflection at the free end of the cantilever, which is converted into cavity length variation. According to the F-P interference principle and demodulated using fast Fourier transform (FFT)-based spectral algorithm for temperature demodulation. Thermo-mechanical coupled finite element simulations under a uniform temperature range of 0 degrees C-150 degrees C shows a temperature sensitivity of 0.394 mu m degrees C-1, with the helical beam structure exhibiting superior stability compared with the maze-shaped beam and the maze-shaped perforated beam structure. An F-P displacement measurement platform was constructed, and calibration as well as stability tests were carried out within 26 degrees C-146 degrees C. The measured sensitivity reached 0.411 mu m degrees C-1, in close agreement with the simulated value (0.395 mu m degrees C-1). Temperature cycling experiments further demonstrated excellent repeatability of the sensor, with temperature fluctuations confined within +/- 0.35 degrees C (approximate to +/- 0.29% F.S.). Moreover, the FFT demodulation method achieved a temperature resolution better than +/- 0.005 degrees C. In summary, the helical beam temperature sensor method features inherent safety, passive operation, immunity to electromagnetic interference, high sensitivity, and rapid response. It can be easily networked through optical fibres and is well suited for high-precision monitoring of weak thermal signals in complex environments.

A single-sideband laser heterodyne spectrometer (LHS) was developed for the first time, to the best of our knowledge. A multi-channel configuration employing different band-pass filters and frequency-shifted LOs achieves single-sideband detection at the hardware level, thereby eliminating the central dip distortion inherent in conventional dual-sideband LHS. The operation of the single-sideband LHS was analyzed and experimentally validated through optimization of the frequency shift of acousto-optic modulators and the bandwidth of radio-frequency filters, yielding a two-fold enhancement in spectral resolution. The performance of the single-sideband LHS was evaluated and validated through measurements of methane absorption spectra. The results demonstrate a significant improvement in spectral fidelity, with the error reduced by 64% compared with conventional dual-sideband LHS. The single-sideband LHS reported in this paper is expected to provide a powerful tool with high spectral resolution and high accuracy for remote sensing of upper atmospheric gases and planetary atmospheric molecules.

A near-infrared external-cavity laser heterodyne radiometer (EC-LHR) with balanced detection is developed for remote sensing of atmospheric water vapor/delta D and CO2/delta C-13. The tunable external-cavity laser serves as the local oscillator (LO). By optimizing the optical heterodyne balanced detection configuration, the EC-LHR achieves quasi-shot-noise-limited performance. High-resolution atmospheric transmission spectra of water vapor, HDO, CO2, and (CO2)-C-13 are simultaneously measured using the developed EC-LHR in ground-based solar occultation mode. Within the framework of the optimal estimation algorithm, three inversion strategies are employed: single-peak retrieval, multiple-peak joint retrieval, and isotopic-ratio-constrained retrieval, enabling the determination of column abundances and vertical profiles of CO2 and water vapor, as well as delta D and delta C-13. The reported EC-LHR has broad application potential in anthropogenic gas emission monitoring and water vapor transport research.

The performance of a near-infrared (NIR) high-resolution field deployable instrument line shape (ILS) calibratedlaser heterodyne radiometer (LHR) for accurate measurement of atmospheric CH4 column abundance is demonstrated. The ILS-calibrated LHR employs a distributed feedback (DFB) laser centered at 1653 nm to extract the CH4 absorption spectrum information in the atmospheric column from solar radiation. Based on the measured spectra and regularization deconvolution algorithm, the ILS of the LHR is accurately determined. Validation of the retrieved ILS is performed using low-pressure CH4 gas cell, and the accuracy is improved by 97 % compared to conventional ILS. Compared to traditional LHRs, the ILS-calibrated LHR leads to a 30 % reduction in residuals near the wings of the absorption lines and a 200 % reduction at the peak positions, resulting in a 10 % enhancement in the accuracy of the column abundance determination. The reported field deployable ILScalibrated LHR provides valuable insights for the accurate measurement of greenhouse gases in Earth's atmosphere.

Pedestrian detection in dense video surveillance scenes remains challenging due to severe occlusion, small-scale pedestrians, and the limitations of NMS-based post-processing and computationally intensive DETR models. To address these issues, this paper proposes LMS-DETR, a lightweight multi-scale pedestrian detection framework based on RT-DETR. Specifically, a lightweight multi-scale backbone network is designed to enhance feature representation across different pedestrian scales. In addition, a small-object enhanced feature fusion network is constructed to strengthen low-level semantic encoding and improve small-scale pedestrian perception. Furthermore, an efficient fusion-aware modeling module is introduced to effectively model fused multi-scale features, while reducing computational cost. Experimental results on the CrowdHuman dataset demonstrate that LMS-DETR achieves improved detection accuracy with significantly reduced model complexity compared with the baseline, indicating strong small-object perception capability and robustness in dense scenes.

Laser heterodyne spectroscopy is facilitating groundbreaking advances across multiple fields, including planetary atmospheric exploration, terrestrial greenhouse gas monitoring, wind field measurements, isotopic ratio analysis, and industrial gas emission monitoring. Nevertheless, Laser heterodyne radiometers (LHRs) lack effective wavelength calibration methods, which hinders their application in scenarios requiring miniaturization, high stability, and high precision. In this paper, what we believe to be a novel all-fiber LHR capable of real-time wavelength calibration is presented. The wavelength calibration scheme for this LHR was implemented using an all-fiber unbalanced Mach-Zehnder interferometer (MZI) combined with a corresponding wavelength calibration algorithm. Key performance-limiting factors of unbalanced MZI were experimentally analyzed. The criteria for determining the optimal optical path difference of the unbalanced MZI were established for the first time specifically for LHR applications. With a 22 cm unbalanced arm design, the system achieved a calibration resolution of 0.01623 cm -1 and demonstrated a calibration uncertainty of 2 × 10 −5 cm -1 at an averaging time of 1s. Its performance was validated through absorption spectra measurements conducted in a gas cell. Field measurements of atmospheric CO 2 absorption spectra were performed with the developed real-time wavelength-calibrated LHR. The unbalanced MZI wavelength calibration scheme not only provides a high-precision, environmentally robust frequency scale for laser heterodyne absorption spectroscopy but also exhibits great potential to serve as a reliable calibration solution for other high-resolution spectroscopic techniques.

Gas leak monitoring represents a critical component in the production, transportation, and processing of highsulfur natural gas, playing a vital role in ensuring operational safety across all stages and enabling environmental impact assessment following potential leaks. This study addresses spectral interference challenges in midinfrared laser gas monitoring systems by developing a gas concentration inversion model based on a mixedLorentzian approach. Focusing on the two primary constituents of high-sulfur natural gas - methane (CH4) and hydrogen sulfide (H2S) - we established an 8.309 mu m central spectral line suitable for simultaneous detection of both gases and implemented a remote mid-infrared laser system.To resolve signal interference between CH4 and H2S during mixed-gas monitoring, we employed spectral line broadening techniques under simulated high-sulfur gas leakage conditions. This enabled effective deployment of the mixed-Lorentzian model for gas signal separation. The parameters derived from the separated Lorentzian components were subsequently integrated into our concentration inversion model, achieving successful decomposition of mixed infrared laser signals.System stability evaluations demonstrated that our mixed-Lorentzian separation model effectively resolves composite gas signals while preserving absorption feature integrity. The model achieved correlation coefficients of 0.9541 for CH4 and 0.9591 for H2S, both exceeding the 0.95 threshold. These results confirm the method's accuracy in simultaneous monitoring of CH4 and H2S concentrations within high-sulfur natural gas environments. This methodology shows significant potential for extension to similar challenges across the energy sector.

During the process of urbanization, the damage caused by illegal construction to urban infrastructure, such as underground pipe networks, is becoming increasingly severe. The mechanical vibration in the infrasound frequency band generated by it makes it difficult for the existing monitoring technologies to achieve real-time perception with a high sensitivity and a low false alarm rate. This article proposes an optical fiber Fabry-Perot (F-P) interferometric acceleration sensor based on a serpentine beam resonant structure (F-P SER-Sensor) for illegal construction monitoring. The finite element analysis software (ABAQUS) is used to carry out modal analysis, sensitivity simulation, and amplitude-frequency characteristic simulation on the serpentine beam sensitive structure, and the optimized design with four bending numbers and three serpentine beams is determined. The experimental results show that the sensor has a good response in the frequency range of 20-550 Hz, the measured resonant frequency is 1050 Hz, the sensitivity at the frequency of 300 Hz reaches 32.95 mV/g, and the acceleration resolution in the working frequency band is 6.25 mu g/Hz(1/2). When collecting hydraulic crushing and manual excavation signals in collaboration with an electronic sensor, it is shown that the sensor has significant response advantages in the low-frequency band (0-50 Hz), and is suitable for the low-frequency monitoring needs of construction scenes. This study provides a new idea for the design of an acceleration sensor, which has broad application prospects in the fields of urban lifeline safety engineering and so on, and the structure can be further optimized in the future to improve the long-term stability and anti-interference ability in complex environments.

Objective Pipeline networks are vital infrastructures for national energy supply, and their safe operation has garnered increasing attention in recent years. Distributed optical fiber sensing based on phase-sensitive optical time-domain reflectometry (5-OTDR) provides an effective solution for real-time intrusion event monitoring along pipelines, boasting advantages such as wide coverage, high sensitivity, and strong immunity to electromagnetic interference. However, 5-OTDR systems still face difficulties in accurately identifying weak and diverse intrusion signals under conditions of limited samples and environmental noise. This study aims to address these challenges by proposing a novel recognition framework that integrates synchrosqueezing transform (SST) for high-resolution time-frequency analysis with the advanced deep learning object detector YOLOv12, thereby enhancing the accuracy and robustness of pipeline intrusion event recognition. Methods A distributed optical fiber micro-vibration sensing system based on 5-OTDR was established to collect vibration signals generated by nine types of common intrusion events along a pipeline. These events include hoeing (HO), ramming (RA), entrenching shovel digging (ESH), hammering (HA), human stepping (HS), human jumping (JU), water flow (WF), shovel digging (SH), and excavator operation (EX). Each raw signal was segmented into 1-second windows and preprocessed for standardization. Three time-frequency transformation methods-short-time Fourier transform (STFT), continuous wavelet transform (CWT), and SST-were applied to convert the time-domain signals into two-dimensional time-frequency representations (spectrograms). SST was specifically employed for its superior capability to concentrate signal energy, particularly for low-frequency, non-stationary vibration patterns. During the SST process, generalized morse wavelets were used as the mother wavelet, and the synchrosqueezing operator threshold was set to 0.05 to enhance frequency localization. The resulting time-frequency images served as inputs to three deep learning models: YOLOv12, VGG, and ResNet. Each model was trained and evaluated on a balanced dataset comprising 639 training samples and 148 validation samples, with a detailed class-wise distribution provided in the manuscript. The models were evaluated based on classification accuracy, loss convergence, confusion matrices, and computational efficiency. Results and Discussions Comparative experimental results demonstrated that the proposed SST-YOLOv12 model significantly outperformed other method combinations across all evaluation metrics. The overall classification accuracy for the nine events reached 94.42 degrees o, with the accuracy for the six key intrusion events peaking at 95.75 degrees o. The SST-based time-frequency maps exhibited superior spectral focusing and clearer feature boundaries compared to those generated by STFT and CWT, making them more conducive to feature extraction by deep networks. Training and validation curves confirmed that the YOLOv12 model with SST input converged faster and achieved lower final loss values. Its validation accuracy remained consistently above 90 degrees o in later training stages and showed greater stability. The corresponding confusion matrix indicated minimal misclassification, with strong diagonal dominance across most event categories. In contrast, VGG and ResNet models performed adequately for interference event recognition but struggled with weak intrusion events, often yielding accuracy below 75 degrees o under STFT and CWT inputs. Moreover, their computational efficiency was lower, especially for VGG. The YOLOv12 model achieved an optimal balance between high accuracy and low computational cost. The synergy between the high-resolution spectral input provided by SST and the efficient detection backbone of YOLOv12 contributed to improved model generalization and practical applicability, showcasing strong potential for deployment in complex monitoring environments. Conclusions This study developed a novel pipeline intrusion recognition method by integrating synchrosqueezing transform with the YOLOv12 deep learning framework. This approach leverages the spectral compactness and high resolution of SST and combines it with a fast, high-performance detector capable of handling small samples and multiple categories. Experimental results validate that the proposed SST-YOLOv12 model achieves superior performance over traditional STFT and CWT-based approaches in terms of classification accuracy, computational efficiency, and model robustness. Furthermore, the proposed system exhibits strong potential for practical deployment in pipeline security monitoring, owing to its standardized image-based input format, adaptability to different environments, and extendability to new event categories through transfer learning. This work provides a promising and scalable solution for intelligent, real-time intrusion monitoring in critical infrastructure protection.

This paper presents a lightweight deep learning approach for detecting small leaks in buried gas pipelines using a phase-sensitive optical time domain reflectometry (4-OTDR) system. Vibration signals acquired from a distributed optical fiber vibration sensor (DOFVS) are converted into time-frequency images via continuous wavelet transform (CWT) to facilitate joint spatiotemporal feature extraction. An improved model that integrates a lightweight MobileNetV3 backbone with a bidirectional LSTM (BiLSTM) module is proposed to capture both spatial patterns from individual CWT images and temporal dependencies across consecutive frames. Experimental results demonstrate an overall accuracy of 95.96% in classifying four leak pressure levels, including the reliable detection of leaks as small as 1/16 inch under pressures as low as 0.1 MPa. Compared to conventional CNN models, an improved recognition accuracy of 8%-12% is demonstrated for the identification of low-pressure and smallaperture leakage conditions. (c) 2026 Optica Publishing Group. All rights, including for text and data mining (TDM), Artificial Intelligence (AI) training, and similar technologies, are reserved.

In this work, an active coherent laser heterodyne sensor (ACLHS) with enhanced performance based on RF white noise perturbation and optical amplification is proposed. It enables simultaneous and precise measurement of methane concentration and accurate localization of leakage points. A near-infrared distributed feedback diode laser was used as the local oscillator, with a portion of its output frequency-shifted to serve as the signal source for detecting CH4 absorption line near 6046.96cm-1. The enhancement of ACLHS sensitivity was accomplished by radio-frequency white noise perturbation and optical amplification. The power of the radio-frequency white noise and the operating current of the optical amplifier were experimentally optimized to achieve the optimal signal-to-noise ratio of the spectral signal. Localization is realized by amplitude-modulated continuous-wave ranging technology. By coupling the ranging circuit into the heterodyne detection circuit, the system can perform concentration measurement and distance measurement simultaneously. This high-sensitivity ACLHS achieved a detection limit of 1.4 ppm·m for CH4 and a source localization accuracy of 15cm at a detection distance of 50m.

In this article, a quantum cascade laser (QCL) with a center wavelength of 8.309 mu m based on wavelength modulation spectroscopy (WMS) is proposed to measure the mixture of hydrogen sulfide (H2S) and methane (CH4) gases with a sensing distance of 30-50 m. The impact of different sensing distances at 15, 30, and 50 m on the detection signal was analyzed. Allan variance calculations demonstrate that the lower detection limits for H2S and CH4 were 0.593 and 1.160 ppb with integration times of 183 and 142 s, respectively. The results provide an effective method for remote measurement of highly sensitive H2S and CH4 multicomponent gases and ensure safe operation in a variety of industrial environments.

The airflow resulted from the leakage of buried gas pipeline would cause the surrounding soil to vibrate.By means of measuring acceleration, the attenuation of vibration signals in soil can be measured to judge whether there is leakage.In the paper, the leakage of buried gas pipeline was simulated and acceleration sensors were used to measure the acceleration signals in case of leakage at apertures of 2 mm, 4 mm and 6 mm, and pressures of 1.5 MPa, 2.5 MPa and 3.5 MPa, so that the propagation characteristics of acceleration signals in different soil media under different leakage pressures were analyzed.Then, combined with ABAQUS, the acceleration attenuation in different soil media was simulated.The results show that the peak acceleration signal exhibits exponential attenuation in soil, which agrees well with the test results.The research findings can help guide the selection and installation of leakage monitoring sensors for buried gas pipelines and the operation of gas pipelines to prevent accidents from occurring.

Focusing on the problem of unclear ray-traced spots and their distribution rules in the design process of the Herriott cell, first, the characteristics of long-optical-path gas absorption cells were analyzed, and the calculation method of basic cavity length and the effective optical path of Herriott gas absorber cells were studied. Second, according to the transmission characteristics of geometric optics, a physical model of light transmission in Herriott cells was established via the LightTools software. Finally, simulation analysis was performed on Herriott cells with 5- and 14.4-m optical paths separately, determining the quantitative relationship between d/f and the number of spots reflected on the concave mirror, and optimizing the effective optical path and output laser energy of the Herriott cells. Through research analysis, the sizes and distribution positions of concave mirror spots in the Herriott cells were identified, as well as the factors affecting the number of reflections. It was also found that the number of reflected spots gradually decreases as d/f increases, revealing the light-tracing results and its spot distribution rule on the mirror surface, as well as verifying the accuracy of the theory. The findings of this study provide a basis for the optical path system design and optimization for Herriott cells with different optical path lengths.

The research of hydrogen sulfide(H2S)and methane(CH4)concentration measurement tech-nology is of great significance to petroleum and petrochemical industry.In this paper,a Quantum Cascade Laser(QCL)with a center wavelength of 8.309 μm was selected as the detection light source based on Tunable Diode Laser Absorption Spectroscopy(TDLAS)technology.A 30 m long distance sensing sys-tem was established by using Wavelength Modulated Spectroscopy(WMS)technology to measure the mixture of H2S and CH4 gases.The experiment mixed H2S with 5%volume fraction of water vapor for measurement,and it showed excellent absorption characteristics in this band,with less cross-interference.Through sensing experiments,the impact of different sensing distances of 15 m and 30 m on the detection signal was analyzed.By increasing integration time and calculating the signal-to-noise ratio,a minimum sensing limit of 128.75×10-9 m was achieved.Finally,Allan variance calculations revealed that when in-tegration time was 183 s and 142 s,the lowest detection limits for H2S and CH4 were 0.593×10-9 and 1.160×10-9,respectively.The results provide an effective method for remote measurement of highly sen-sitive H2S and CH4 multi-component gases and ensure safe operation in a variety of industrial environ-ments.

This work utilizes the CEEMDAN algorithm to analyze the interference of Rayleigh back-scattering signals in standard communication optical fibers. The technology has several advantages, such as anti-electromagnetic interference, improved electrical insulation, corrosion resistance, higher sensitivity, and the capability for long-distance monitoring. In this study, in-situ monitoring data from a 53.2 km natural gas pipeline in a terrain area in Southwest China were analyzed. The results demonstrate that, using the CEEMDAN algorithm for a blind test conducted over fourteen days, a 100% recognition accuracy for mechanical tamping and a Nuisance Alarm Rate (NAR) of less than 1% were achieved.

Focused ion beam (FIB) machining has been demonstrated to be capable of fabricating nano- and micro-scale structures. In this paper we demonstrate techniques to design and fabricate 45° micro-mirrors into the end of multi-core fibres using FIB processing. The mirrors are fabricated by a two-step process: a scanning process which is used to make a rough cut followed by a polishing process to create an optical surface finish mirror. The machined 45° mirrors can be accurately aligned with optical fibre cores, which avoids issues associated with the alignment of external turning mirror components. Proof-of-concept demonstration shows that the fabricated structure is capable of measuring two-axis acceleration interferometrically with a linear response from 0.2 to 4 g and an rms. error of 0.03 g. Acceleration measurements of frequency response up to 700 Hz and cross-sensitivity of ∼4.3% are demonstrated.