The spectral selectivity of underwater multiwavelength single-photon LiDAR offers a promising pathway to discriminate target materials beyond conventional geometric imaging. However, the complex interactions among wavelength-dependent water attenuation, target reflectance, and scattering-induced waveform distortion remain poorly quantified. This study establishes a comprehensive theoretical and experimental framework linking these factors, validated through controlled experiments across two water turbidity levels (attenuation coefficients of 0.1 m−1 and 2.0 m−1), six wavelengths (490–570 nm), and diverse target types. We demonstrate that target ranging bias exhibits a wavelength-dependent linear trend (8.3 ps/nm) in turbid waters. This phenomenon is fundamentally attributable to forward-scattering-induced centroid shifts rather than true spatial displacements, a mechanism we quantify through comparative peak-detection and Gaussian fitting analyses. Contrary to intuitive expectations, we reveal that spectral discrimination efficacy decouples from received photon counts. Principal component analysis confirms that a multidimensional spectral feature space enables accurate target clustering independent of absolute intensity, with specific bands (e.g., 510 nm and 550 nm) exhibiting heightened sensitivity to material signatures. These findings establish that underwater target recognition is primarily influenced by the spectral contrast between target reflectance and water transmission windows, rather than solely depending on received photon counts, providing a robust physical basis for next-generation underwater LiDAR optimization.

Fringe projection profilometry (FPP) suffers from heterogeneous phase unwrapping errors (PUEs) that are difficult to correct cooperatively. We propose a classification and adaptive correction method based on global structural priors. Building on spatial continuity and approximate low-rankness, the method classifies PUEs into three categories—structural gross, sparse gross, and random—and corrects them progressively through boundary preprocessing, structure-aware weighted low-rank optimization (SA-WLRO), and conservative median filtering. An optional adaptive backfilling step is then introduced, yielding a dual-mode approach: the core algorithm (without backfilling) and the backfilling mode (with backfilling enabled), which together enable a flexible trade-off between precision and data completeness. Systematic experiments on flat plate, composite geometric bodies, plaster sculptures, and stair model demonstrate that the proposed dual-mode approach achieves substantial improvements over four representative model-driven methods in all tested scenarios, which include planar surfaces, isolated discontinuities, complex curved surfaces, and steep height variations. Notably, on the stair model—the most challenging case— the core algorithm achieves an RMSE of 0.0437 rad at 97.55% VPP, while the backfilling mode raises the VPP to 99.99% with an RMSE of 0.3084 rad, still more accurate than the 0.4251 rad of the best competing method. Ablation studies confirm the necessity of each module. These results position the proposed dual-mode approach as an effective post-processing solution for high-precision FPP systems.

In offline signature verification, extracting effective features and enhancing identification accuracy remain critical challenges. Traditional feature extraction modules struggle to capture comprehensive, detailed characteristics from signatures alone and often suffer from overfitting issues. This paper introduces ResT, a novel approach that integrates the Residual Network (ResNet) and Transformer architectures for multi-scale feature extraction. The proposed method comprises two components: ResNet blocks and Transformer blocks, designed to extract local and global signature characteristics, respectively. Within the ResNet blocks, we integrate two modules-Spatial-to-Depth Convolution (SPD-Conv) and Spatial and Channel Reconstruction Convolution (SCConv)-to emphasize stroke-level features, thereby improving local feature extraction. For the Transformer blocks, Vision Transformer (ViT) is employed to analyze the signature's overall shape and stroke trends, capturing global features. To evaluate performance, we implement a writer-independent (WI) verification system and conduct extensive experiments on four public datasets: GPDS, CEDAR, BHSig-Bengali, and BHSig-Hindi. Results demonstrate that the proposed ResT model effectively distinguishes genuine from forged signatures and achieves competitive performance compared to existing methods.

In the field of forensic document examination, accurately determining the chronological sequence of intersecting lines between seal ink and handwriting is a crucial technical step for verifying document authenticity, identifying contract tampering, and detecting forged signatures. This technique analyzes the physical superimposition relationship formed by the deposition of the two media on the paper substrate to provide objective scientific evidence for judicial practice. Although traditional methods such as microscopic imaging and mass spectrometry analysis have achieved some progress, they still suffer from common limitations including high equipment costs, complex operation, and potential damage to samples. This study proposes and validates an innovative non-destructive determination method that integrates structured light 3D reconstruction technology with deep learning algorithms. The research captures the microscopic 3D morphological features of the ink intersection area using a high-precision structured light scanning system and effectively eliminates noise interference caused by paper substrate undulation through Gaussian flattening technology. Subsequently, a multimodal fusion strategy combines 2D texture images with 3D depth information to construct a dataset rich in features. On this basis, a deep learning model based on an improved Residual Neural Network (ResNet) is designed, incorporating the ELU activation function and an EMA mechanism to enhance the model’s feature extraction capability and convergence stability. Experimental results demonstrate that the proposed method achieves a recognition accuracy of 94.39% on the test set, fully validating its effectiveness and application potential in the non-destructive determination of ink stroke sequencing.

Semantic segmentation is a key technology for autonomous vehicles to understand the surrounding scenes. Multi-branch network architectures have demonstrated their efficiency and effectiveness in real-time semantic segmentation tasks. Although PIDNet achieves a balance between performance and efficiency, it is inadequate in fine-grained segmentation and multi-scale feature integration. In this paper, we propose a real-time semantic segmentation network, called ARFNet, which employs a three-branch structure. To enhance the perception and fusion of multi-scale features, we introduce a Hierarchical Dense Atrous Pyramid Module (HDAPM). Additionally, we propose a novel Trans-Dimensional Interaction Module (TDIM) to systematically enhance feature representation through cross-dimensional attention mechanisms. The effectiveness of our method is demonstrated by extensive experiments on the Cityscapes and CamVid datasets, showing that it achieves a promising trade-off between inference speed and segmentation accuracy. Specifically, ARFNet achieves 79.8% mIoU at 94.8 FPS on the Cityscapes dataset and 80.9% mIoU at 154.1 FPS on the CamVid dataset.

The marine lifting arm system is a novel type of lifting equipment for the dismantling of large offshore platforms. In practical application, the marine lifting arm system is commonly subject to unexpected disturbances such as persistent vessel-induced disturbances, uncertain dynamics, and physical constraints, posing significant challenges for controller system design. This paper proposes a neural network (NN) based adaptive tracking control method with state constraints for the marine lifting arm system to effectively compensate for vessel motions induced by wave disturbances. First, the dynamic model of a marine lifting system is established by employing Lagrange's method. Subsequently, a nonlinear controller is developed based on the dynamic model to compensate for 3-DOF vessel motions, including heave, roll, and pitch motion. By constructing constrained terms, the nonlinear controller effectively restricts the lifting arm's motion within the predefined range, ensuring the safety of lifting operations even under adverse sea conditions. The neural network is employed to estimate parameter/structure uncertainties and the nonlinear input dead zones for lifting arm systems. In particular, the designed update law of the neural network considers physical constraints, thereby further ensuring that the lifting arm operates within the safe range. The theoretical stability of the proposed control system is rigorously proven using the Lyapunov technique. Finally, the practicality and effectiveness of the proposed method are verified by the hardware experiments on a self-built offshore lifting system.

In the field of fringe projection profilometry, phase sensitivity is a critical factor influencing the precision of object measurements. Traditional techniques that employ basic horizontal or vertical fringe projection often do not achieve optimal levels of phase sensitivity. The identification of the fringe angle that exhibits optimal phase sensitivity has been a significant area of research. The present study introduces a novel method for determining the optimal fringe angle, facilitating 3D reconstruction without the need for equipment adjustments. Initially, the optimal fringe is derived through mathematical analysis, and the system’s position within each coordinate system is standardized, leading to the determination of the optimal fringe angle in the world coordinate system. Subsequently, an optimal fringe pattern, akin to that produced by a rotating projector, is generated based on the concept of rotation around a central point, with corresponding adjustments made to the calibration parameters. Finally, the optimal fringe is projected onto the target object for 3D reconstruction, thereby validating the proposed method. The experimental results demonstrate that this approach accurately identifies the optimal fringe angle, significantly enhancing both phase sensitivity and measurement accuracy. The accuracy of the measurement is significantly greater, by an order of magnitude, compared to the traditional method, with the error being approximately 50% of that associated with the currently established improved method.

A bullet point cloud registration algorithm with a low overlap rate based on line feature detection was proposed to solve the problem of the difficulty and low efficiency of point cloud registration due to the low overlap rate among point clouds sampled by the bullet model. In this paper, voxel downsampling is used to remove some noise points and outliers from the bullet point cloud and applied to the specified resolution to reduce the calculation cost. The bullet point cloud is transformed to a better initial position by fitting the central axis with the geometrical features of the bullet. Then, the direction vector of the bullet linear features is obtained by using an icosahedral fitting discrete Hough transform to simplify the parameter space of the search transformation. Finally, the optimal rotation angle is searched for in the parameter space by using the improved Cuckoo algorithm to realize the registration of the bullet point cloud with a low overlap rate. Simulation and experimental results show that the proposed registration method can accurately register bullet point clouds of different densities with a low overlap rate. Compared with the commonly used ICP, GICP, and TRICP algorithms, the registration error of the proposed algorithm is reduced by 92.68% on average when the overlap rate is 52.85%. The registration error is reduced by 98.87% in the case of a 41.36% overlap rate, by 99.52% in the case of a 33.02% overlap rate, and by 98.89% in the case of a 22.75% overlap rate.

In order to improve the performance of lane detection algorithms under complex scenes like obstacles, we proposed a multi-lane detection method based on dual attention mechanism. Firstly, we designed a lane segmentation network based on a spatial and channel attention mechanism. With this, we obtained a binary image which shows lane pixels and the background region. Then, we introduced HNet which can output a perspective transformation matrix and transform the image to a bird's eye view. Next, we did curve fitting and transformed the result back to the original image. Finally, we defined the region between the two-lane lines near the middle of the image as the ego lane. Our algorithm achieves a 96.63% accuracy with real-time performance of 134 FPS on the Tusimple dataset. In addition, it obtains 77.32% of precision on the CULane dataset. The experiments show that our proposed lane detection algorithm can detect multi-lane lines under different scenarios including obstacles. Our proposed algorithm shows more excellent performance compared with the other traditional lane line detection algorithms.

In this work, we propose a 3D occlusion facial recognition network based on a multi-feature combination threshold (MFCT-3DOFRNet). First, we design and extract the depth information of the 3D face point cloud, the elevation, and the azimuth angle of the normal vector as new 3D facially distinctive features, so as to improve the differentiation between 3D faces. Next, we propose a multi-feature combinatorial threshold that will be embedded at the input of the backbone network to implement the removal of occlusion features in each channel image. To enhance the feature extraction capability of the neural network for missing faces, we also introduce a missing face data generation method that enhances the training samples of the network. Finally, we use a Focal-ArcFace loss function to increase the inter-class decision boundaries and improve network performance during the training process. The experimental results show that the method has excellent recognition performance for unoccluded faces and also effectively improves the performance of 3D occlusion face recognition. The average Top-1 recognition rate of the proposed MFCT-3DOFRNet for the Bosphorus database is 99.52%, including 98.94% for occluded faces and 100% for unoccluded faces. For the UMB-DB dataset, the average Top-1 recognition rate is 95.08%, including 93.41% for occluded faces and 100% for unoccluded faces. These 3D face recognition experiments show that the proposed method essentially meets the requirements of high accuracy and good robustness.

Fringe projection profilometry (FPP) is prone to phase unwrapping error (PUE) due to phase noise and measurement conditions. Most of the existing PUE-correction methods detect and correct PUE on a pixel-by-pixel or partitioned block basis and do not make full use of the correlation of all information in the unwrapped phase map. In this study, a new method for detecting and correcting PUE is proposed. First, according to the low rank of the unwrapped phase map, multiple linear regression analysis is used to obtain the regression plane of the unwrapped phase, and thick PUE positions are marked on the basis of the tolerance set according to the regression plane. Then, an improved median filter is used to mark random PUE positions and finally correct marked PUE. Experimental results show that the proposed method is effective and robust. In addition, this method is progressive in the treatment of highly abrupt or discontinuous regions.

Invalid points, such as shadow and background, in the captured fringe patterns of fringe projection profilometry (FPP) are often inevitable due to the limited field of view measurements of three-dimensional (3D) imaging equipment. To ensure the quality of 3D reconstruction data, these invalid points must be identified and removed. FPP captures co-frequency-based fringe pattern sequences and approximately distributes this data along an ideal cosine curve. We propose an invalid points removal method based on an error energy function. By analyzing the relationship between the pixel values of a series of captured co-frequency patterns and an ideal cosine curve, we quantize the error energy by using a Gaussian weighted Euclidean distance. An improved Gaussian filtering method based on modulation intensity is used to significantly increase the error energy difference between the valid points of the target and the invalid points of the shadow/background area. Finally, the points whose error energy is greater than a certain error threshold are removed as invalid points. Experimental results indicate that this method can efficiently remove invalid points in fringe patterns and performs better than existing traditional methods.

Aiming at the high error caused by the integral method for the traditional optical power meter, from the application point of view, a high-precision optical power meter with MSP430 is proposed. The photoelectric sensor converts the light signal into mA current signal, through a negative feedback current amplification, and then through a differential amplification and filtering circuit, and converts anolog signal into a digital signal through a 24-bit high-precision AD chip. The optical power is calculated using recursive and median digital filter. The measured value is displayed on the LCD. The optical power curve is also displayed after through the Savitzky-Golay filter in measurement phase. The experiment proves that the optical power meter has an error of less than 0.3%, low power consumption, stable system, friendly display interface, it can store measurement data and display optical power curve.

Tracking objects over time, i.e., identity (ID) consistency, is important when dealing with multiple object tracking (MOT). Especially in complex scenes with occlusion and interaction of objects this is challenging. Significant improvements in single object tracking (SOT) methods have inspired the introduction of SOT to MOT to improve the robustness, that is, maintaining object identities as long as possible, as well as helping alleviate the limitations from imperfect detections. SOT methods are constantly generalized to capture appearance changes of the object, and designed to efficiently distinguish the object from the background. Hence, simply extending SOT to a MOT scenario, which consists of a complex scene with spatially mixed, occluded, and similar objects, will encounter problems in computational efficiency and drifted results. To address this issue, we propose a binary-channel verification model that deeply excavates the potential of SOT in refining the representation while maintaining the identities of the object. In particular, we construct an integrated model that jointly processes the previous information of existing objects and new incoming detections, by using a unified correlation filter through the whole process to maintain consistency. A delay processing strategy consisting of the three parts—attaching, re-initialization, and re-claiming—is proposed to tackle drifted results caused by occlusion. Avoiding the fuzzy appearance features of complex scenes in MOT, this strategy can improve the ability to distinguish specific objects from each other without contaminating the fragile training space of a single object tracker, which is the main cause of the drift results. We demonstrate the effectiveness of our proposed approach on the MOT17 challenge benchmarks. Our approach shows better overall ID consistency performance in comparison with previous works.

Road detection is a crucial research topic in computer vision, especially in the framework of autonomous driving and driver assistance. Moreover, it is an invaluable step for other tasks such as collision warning, vehicle detection, and pedestrian detection. Nevertheless, road detection remains challenging due to the presence of continuously changing backgrounds, varying illumination (shadows and highlights), variability of road appearance (size, shape, and color), and differently shaped objects (lane markings, vehicles, and pedestrians). In this paper, we propose an algorithm fusing appearance and prior cues for road detection. Firstly, input images are preprocessed by simple linear iterative clustering (SLIC), morphological processing, and illuminant invariant transformation to get superpixels and remove lane markings, shadows, and highlights. Then, we design a novel seed superpixels selection method and model appearance cues using the Gaussian mixture model with the selected seed superpixels. Next, we propose to construct a road geometric prior model offline, which can provide statistical descriptions and relevant information to infer the location of the road surface. Finally, a Bayesian framework is used to fuse appearance and prior cues. Experiments are carried out on the Karlsruhe Institute of Technology and Toyota Technological Institute (KITTI) road benchmark where the proposed algorithm shows compelling performance and achieves state-of-the-art results among the model-based methods.

随着星敏感器探测灵敏度的提高,导航星表中恒星的数量急剧增加,导致星图识别的识别速度和识别率降低。因此,为了提高星图识别的识别速度和识别率,文中在三角形算法的基础上提出了一种基于多特征匹配的快速星图识别算法。首先,采用天球的内接正二十面体法对预处理后的星表进行分区。然后,将特征三角形的边长以及外接圆和内切圆半径的乘积作为特征值构建导航特征库,并根据后者的哈希函数对特征库进行分块。在识别的过程中,利用观测三角形外接圆和内切圆半径乘积的哈希函数实现导航特征库子块的快速定位,并在该子块内采用多特征匹配的方法得到观测三角形的识别结果,最后根据该结果确定星敏感器视场所包含的天球子区域,并在子区域内完成视场中其它导航星的识别。实验结果表明,文中算法的识别性能与导航特征库的分块数有关,在选择合适的分块数后,与常用三角形算法相比,算法在识别速度,识别率以及对星等误差和假目标的鲁棒性等方面具有明显的优势,算法的平均识别时间和识别率分别为17.161ms和98.58%,满足星敏感器对高识别速度和识别率的要求。

In order to solve the problem that the recognition rates of skilled forgery signatures is not satisfied, a personal signatures recognition scheme based on 3D depth feature of strokes is proposed. Firstly, the real signatures of different writers and the skilled forgery signatures are collected by high-precision stereo microscope, obtaining the surface 3D point cloud data of the signature strokes. Secondly, after filtering the noise of cloud data through the Gaussian filter, the statistical features such as the average depth of the stroke, the standard deviation of the depth along the stroke direction, and the entropy are calculated. To increase the number of data sets, data augmentation is implemented. Finally, the data set is divided into training set and test set, using classifier (including SVM, KNN, ANN) to classify the data set under different training ratios. The experimental results show that the best recognition accuracy of the algorithm on the local signature data set is 98.69%, which is better than most traditional algorithms and meets the requirements of practical applications.

Under the dynamic working conditions of a star sensor, motion blur of the star will appear due to its energy dispersion during imaging, leading to the degradation of the star centroid accuracy and attitude accuracy of the star sensor. To address this, a restoration method of a blurred star image for a star sensor under dynamic conditions is presented in this paper. First, a kinematic model of the star centroid and the degradation function of blurred star image under different conditions are analyzed. Then, an improved curvature filtering method based on energy function is proposed to remove the noise and improve the signal-to-noise ratio of the star image. Finally, the Richardson Lucy algorithm is used and the termination condition of the iterative equation is established by using the star centroid coordinates in three consecutive frames of restored images to ensure the restoration effect of the blurred star image and the accuracy of the star centroid coordinates. Under the dynamic condition of 0~4°/s, the proposed algorithm can effectively improve the signal-to-noise ratio of a blurred star image and maintain an error of the star centroid coordinates that is less than 0.1 pixels, which meets the requirement for high centroid accuracy.

针对基于深度学习的DeepLabV3+语义分割算法在编码特征提取阶段大量细节信息被丢失,导致其在物体边缘部分分割效果不佳的问题,本文提出了基于DeepLabV3+与超像素优化的语义分割算法。首先,使用DeepLabV3+模型提取图像语义特征并得到粗糙的语义分割结果;然后,使用SLIC超像素分割算法将输入图像分割成超像素图像;最后,融合高层抽象的语义特征和超像素的细节信息,得到边缘优化的语义分割结果。在PASCAL VOC 2O12数据集上的实验表明,相比较DeepLabV3+语义分割算法,本文算法在物体边缘等细节部分有着更好的语义分割性能,其mIoU值达到83.8%,性能得到显著提高并达到了目前领先的水平。

Under the dynamic working conditions for a star sensor, motion blur will appear in a star because of its energy dispersion in the process of imaging, which leads to a decrease in the signal to noise ratio (SNR) and makes the blurred region difficult to extract. Meanwhile, this causes a degradation in star centroid positon accuracy and attitude accuracy in the star sensor. Therefore, a restoration method for blurred star images based on region filters is presented in this paper, which simultaneously concentrates on the improvement of SNR and star centroid accuracy. Firstly, the kinematic models of a star centroid under different conditions are set up based on the characteristics of star sensors. Secondly, the motion trail of star centroid is determined based on the kinematic model, allowing the star blurred region to be extracted. The images inside and outside the star blurred region are then preprocessed by image processing algorithm respectively. Finally, the blurred star image is restored by an image restoration algorithm. The experiment results indicate that under the dynamic condition of 2 degrees/s, the region filter algorithm can effectively improve the SNR of a blurred star image. In restored images, the error of star centroid is less than 0.1 pixels, which can satisfy the requirements for star sensor of high centroid accuracy.