Abstract:Optical frequency domain reflectometry (OFDR) technology has become one of the core technologies for fiber optic diagnosis and precision measurement. It converts the nonlinearity of optical frequency scanning into compensable phase errors and combines frequency domain signal processing to achieve centimeter level spatial resolution and long-distance measurement. However, the nonlinearity and phase noise of tunable laser frequency scanning can affect the equal frequency sampling of the system, causing reflection broadening and distance deviation. Therefore, nonlinear compensation becomes the key to improving system performance. The research on nonlinear compensation mainly focuses on two aspects: hardware and algorithm. On the one hand, by adjusting the hardware structure of the system, equal frequency interval sampling can be achieved to suppress sweep nonlinearity at the front end while considering range and resolution. On the other hand, at the level of digital signal processing, various algorithms are used to improve spatial resolution and signal-to-noise ratio without the need for additional complex optical paths. This article will introduce the basic principles of OFDR sensing and demodulation, analyze the factors that affect the performance of OFDR systems such as spatial resolution and sensing distance, and focus on summarizing the research progress of key technologies in improving the performance of OFDR systems. Finally, the problems existing in OFDR and the future development trends are proposed.

Abstract:To detect objects in remote sensing imagery remains highly challenging due to complex background clutter, extreme variations in object scale, and high sensitivity to arbitrary orientations. To address these issues, this paper proposes DMADet, a novel detection framework based on dual-domain feature extraction and multi-dimensional focused diffusion, aiming to enhance perception capability and robustness in complex remote sensing scenarios. The method first introduces a dual-domain collaborative network (DDCNet) that jointly exploits spatial and frequency domain information: edge-aware features in the spatial domain are strengthened using Scharr operators, while global contextual representations are captured in the frequency domain via Fast Fourier transform, enabling bidirectional interaction and complementary fusion between the two domains. Second, to alleviate semantic degradation across layers in conventional feature pyramids, a multi-dimensional focused diffusion pyramid (MFDPN) is developed, which employs a channel-block weighting strategy to adaptively integrate high- and low-level features and incorporates a lightweight dual-stream collaborative attention module (LDC-Attention) to enhance multi-scale contextual awareness. Finally, an adaptive rotation-aware detection head is designed, leveraging a dynamic routing mechanism to generate orientation-sensitive convolution kernels, thereby significantly improving rotational invariance and detection accuracy for arbitrarily oriented objects. Extensive experiments demonstrate that DMADet achieves mAP@0.5 scores of 75.9%, 96.8%, and 85.7% on the DOTA-v1.0, NWPU VHR-10, and RSOD benchmark datasets, respectively—consistently outperforming current state-of-the-art methods. These results validate the effectiveness and superiority of the proposed approach in mitigating performance degradation caused by background interference, scale diversity, and rotational variance, while substantially improving object localization accuracy and robustness in real-world remote sensing applications.

Abstract:Given the critical role of calcium gluconate in clinical detection, we have designed a detection method based on terahertz time-domain spectroscopy (THz-TDS), which enables rapid, non-destructive, and high-precision spectral analysis. Initially, calcium gluconate film samples were prepared using a compression method with a mass percentage ranging from 2.5% to 20%. Additionally, solutions with molar concentrations of 5 to 20 mmol/L were prepared using a water-glycerol mixture as the solvent. These samples were analyzed using a self-constructed THz-TDS transmission system under room temperature and vacuum conditions. Subsequently, Fourier transform was applied to convert the time-domain signals into frequency-domain spectra, from which the absorption coefficient and refractive index spectra were extracted. A quantitative relationship between concentration and spectral parameters was established through linear fitting. The experimental results indicate that the film samples exhibit distinct absorption peaks at 1.19, 1.39, 1.66, 1.87, and 2.29 THz. Data analysis reveals that the absorption coefficients at these peaks demonstrate a clear linear relationship with the sample’s mass percentage, particularly at 1.66 THz, where the linear fitting degree reaches 0.991. For solution samples, due to the strong absorption characteristics of the solvent, the refractive index spectrum analysis shows a good linear relationship between the refractive index and molar concentration at 1.66 THz, with a fitting degree of 0.987. These results suggest that this method not only allows for effective quantitative analysis of the absorption properties of samples but also provides accurate evaluation of the dispersion (refractive index) properties. When combined with miniaturized THz-TDS equipment and artificial intelligence algorithms, this technology demonstrates significant scientific research value and application potential in clinical blood calcium monitoring and drug authenticity identification.

Abstract:To address the issues of high missed detection rates for small targets, low detection accuracy, and insufficient algorithm robustness in fiber rope surface defect detection under scenarios with fine textures and complex backgrounds, this paper proposes an improved fiber rope surface defect detection algorithm based on the unsupervised model a discriminatively trained reconstruction embedding for surface anomaly detection （DRAEM）. In the preprocessing stage, the GrabCut segmentation algorithm is incorporated to extract masks for each image, reducing background interference through mask-constrained anomaly generation and mitigating false positives and missed detections caused by complex backgrounds. In the reconstruction network, skip connections are introduced to capture high-dimensional image space normal data distributions, and a dual channel-spatial attention module is added to enhance the reconstruction capability for anomalous regions, to improve the reconstruction quality of fine-scale textures, thereby avoiding texture loss and preventing the missed detection of small-scale defects. In the segmentation network, Transformer modules are integrated after the last two layers of the encoder to optimize the capture of cascaded global features between normal and abnormal regions. Additionally, atrous spatial pyramid pooling （ASPP） is employed to capture multi-scale contextual information for global feature aggregation, providing sufficient semantic differentiation for segmentation and enhancing the model’s segmentation accuracy and the detection accuracy of small surface defects. Experimental results demonstrate that, compared to the original DRAEM, the proposed method achieves a 4.4% improvement in image-level AUROC, a 4.43% increase in pixel-level AUROC, and a 21.86% boost in pixel-level average precision. These enhancements significantly improve the model’s recognition accuracy and robustness, making it more effective for fiber rope defect detection applications.

Abstract:To address the current limitation in polarization image fusion technology—where the focus is predominantly on the visual quality and statistical metrics of the fused output while neglecting its applicability to subsequent high-level vision tasks—this paper proposes a dual-branch feature extraction architecture for polarization image fusion, guided by semantic segmentation. The fusion network comprises an encoder, a fusion layer, and a decoder. In the encoder, a dual-branch feature extractor—composed of GRDB and Swin Transformers—is constructed to extract local polarization features and global intensity information from the source images. Within the fusion layer, an INN is employed to model the inter-feature correlations, enabling lossless enhancement and effective fusion of the polarization characteristics. In the decoder, Restormer serves as the core building block to reconstruct and preserve high-frequency details and structural scene information from the fused features, thereby enhancing image clarity and generating the final fused result. To enrich the fused output with task-relevant semantics, the fusion network is cascaded with a segmentation network during training. The semantic segmentation loss is leveraged to guide the backpropagation of high-level semantic information, thereby optimizing the fusion network and improving the utility of the fused images for advanced vision tasks. Experimental results demonstrate that the proposed network achieves superior performance in both subjective visual assessment and downstream semantic segmentation tasks. Moreover, it outperforms existing fusion methods in objective metrics, with notable improvements of 27% in EN and 16.8% in SSIM.

Abstract:To address the limitations of traditional calibration methods in non-contact 3D reconstruction—specifically the frequent need for manual intervention, significant error accumulation caused by multi-step calibration, and low levels of automation— this paper proposes an automatic synchronous calibration method based on cross-structured light. By utilizing the geometric invariance of the intersection points of dual laser beams, a spatial geometric model of the cross-structured light is constructed. An algorithm is then derived to synchronously solve the camera intrinsic parameters, extrinsic parameters, and light plane parameters using a single set of images. This process eliminates the reliance on repeated target repositioning required by traditional methods and achieves parameter unification directly through the geometric transformation of feature point clouds, realizing the integration of calibration and measurement. Validated through scanning experiments on standard aluminum oxide plates and non-standard aluminum alloy components, the results demonstrate that under standard industrial conditions, the system’s dimensional measurement error is controlled within ±0.5 mm, the average 3D reconstruction accuracy reaches ±0.25 mm, and the peak error rate is below 2.67%. The proposed method effectively overcomes the reliance on manual assistance, significantly reducing operational complexity and improving inspection efficiency while ensuring high precision. It provides an efficient, low-cost, and fully automated solution for industrial measurement, reverse engineering, and surface morphology analysis.

Abstract:To address the challenges of color cast, blurriness, and detail loss in underwater images caused by light absorption and scattering, this study proposes a KFT-GAN for underwater image enhancement. The paper introduces a KFT module that facilitates the efficient transmission of critical features such as color, edges, and texture within the network. By integrating depthwise separable convolution, a lightweight network model is constructed, reducing the parameter count by 60.6% compared to models utilizing conventional convolution, thereby enhancing the model’s learning efficiency. This module enables the generator’s encoding stage to extract essential key features from the input image and transmit these features to the decoding stage through downsampling and skip connections, improving the quality of the reconstructed image. Additionally, this study proposes a hybrid loss function based on perceptual loss principles, emphasizing multiple key attributes of the image to achieve superior visual quality. The proposed model demonstrates excellent performance on both the EUVP and UIEB datasets, achieving PSNR values of 21.384 2 and 18.025 6, and SSIM values of 0.741 3 and 0.688 9, respectively. Qualitative and quantitative comparisons with traditional algorithms and deep learning methods validate the effectiveness and superiority of the proposed model.

Abstract:In computer vision tasks, dust and fog environments have a severe impact on the visibility and detailed features of images, which restricts the performance of downstream vision tasks. To restore and enhance the details of images degraded by adverse weather conditions, a spatio-temporal frequency domain image restoration and enhancement method is proposed. This method studies the mathematical model of light diffusion under dust and fog atmospheric conditions, uses Gaussian filtering to simulate the diffusion and attenuation effect of the atmosphere on light propagation, constructs a pseudo-time image sequence from the degraded input, and obtains the spatio-temporal frequency domain features of the sequence through Fourier transform in the spatio-temporal dimension. Inspired by the restored pseudo heat flux (RPHF) theory, a frequency domain deconvolution kernel is designed to weight the high-frequency information of the sequence to counteract the degradation effect of atmospheric diffusion on the image detail information. The inverse Fourier transform is performed on the weighted frequency features to reconstruct and enhance the image. To verify this method, a weather dataset containing dust and fog scenes with different degradation intensities is established for experiments. The experimental results show that compared with traditional algorithms, this method performs excellently in medium and severe degradation scenarios (such as the visible edge ratio eof severely degraded fog: 78.990) and can effectively restore images. However, in mild degradation scenarios, due to the large amount of high-frequency information in the images, the indiscriminate amplification of high-frequency information by the method has a negative effect on image quality restoration. Overall, this method is more suitable for the restoration and enhancement of moderately to severely degraded images.

Abstract:To address the problems of low target segmentation accuracy and poor mask quality in traffic scenes, an improved YOLOv11n efficient traffic instance segmentation algorithm, ETIS-YOLO, is proposed. Firstly, the C3k2-WTConv module is constructed by fusing the Wavelet Transform Convolution into the C3k2 module of the backbone network to efficiently expand the receptive field and enhance low-frequency feature extraction; Secondly, the feature interaction enhancement AIFI-LA module is designed to reduce the multi-scale computational redundancy of spatial pyramid pooling-fast (SPPF) and improve its ability to handle long sequences and preserve key feature information; Additionally, the feature recalibration EMCSA module is proposed and embedded into the up-sampling operator content aware reassembly of features （CARAFE） to form a CARAFE-EMCSA module, which reconstructs the up-sampling process to enhance the capture of contextual features and the overall discriminability of feature maps; Finally, Soft-NMS and DIoU-NMS are fused and replaced with the original non-maximum suppression （NMS）, which further optimizes the selection and improves the accuracy of the bounding boxes by utilizing relative position information while retaining more high-quality bounding boxes. The experimental results show that on the cityscapes dataset, the bounding box accuracy mAP@0.5 and mAP@0.5:0.95 values are improved by 9.2% and 8.5%, and the segmentation mask accuracy mAP@0.5 and mAP@0.5:0.95 values are improved by 10.6% and 8.8%, respectively, compared with the YOLOv11n model; on the BDD100K dataset, the bounding box accuracy mAP@0.5 and mAP@0.5:0.95 values are improved by 5.1% and 7.4%, and the segmentation mask accuracy mAP@0.5 and mAP@0.5:0.95 values are improved by 4.5% and 6.6%, respectively. It can be seen that the proposed method is effective in traffic scene segmentation.

Abstract:In complex environments where satellite signals are obstructed or denied, vehicle localization systems that rely solely on GNSS often fail to provide stable and high-precision position estimates. Visual landmarks therefore serve as essential auxiliary information for enhancing the robustness of autonomous navigation. However, traditional visual map construction methods typically suffer from long processing times, dependence on continuous tracking, and strong sensitivity to illumination variations and dynamic objects, resulting in insufficient landmark accuracy and spatial consistency. To address these limitations, this paper proposes a visual landmark construction method incorporating RTK constraints. The method first detects near-straight road segments based on the vehicle’s heading-angle sequence and selects representative image frames via linear interpolation as candidate visual landmarks, while introducing semantic priors to enhance their long-term stability and re-identifiability. Subsequently, a neighborhood image set is constructed around each landmark frame, and the initial camera poses are corrected using globally referenced RTK measurements to improve the robustness and reliability of local map initialization. Furthermore, an adaptive RTK constraint is integrated into both local sparse reconstruction and global optimization, enabling scale-consistent and geographically aligned high-precision landmark construction. A degraded smoothing constraint is activated when RTK signals are disturbed or temporarily unavailable to maintain system stability. Experimental results demonstrate that, compared with conventional approaches, the proposed method achieves notable improvements in map accuracy, reconstruction efficiency, and localization robustness under varying illumination, seasonal changes, and dynamic scenes, yielding an 82.2% increase in mapping accuracy, an 81.3% reduction in processing time, and up to a 70.1% improvement in localization precision.

Abstract:To address the challenge of low hit rate in thunderstorm cloud identification, deep learning techniques are employed to enhance recognition accuracy. The research focuses on developing a novel thunderstorm cloud identification model, LkaUNet, which integrates a large kernel attention (LKA) mechanism with the U-Net architecture. This design enhances the model’s ability to capture global morphological features and long-range spatial dependencies of thunderstorm clouds. The study utilizes S-band radar base data and lightning observation data from Hunan Province (2022~2023), employing multi-stage quality control to synchronize radar composite reflectivity and lightning data while suppressing noise. This process generates high-quality datasets of lightning probability and radar composite reflectivity mosaics as input. The LkaUNet model builds upon the U-Net framework and incorporates large kernel attention modules to expand the receptive field, thereby improving long-range dependency modeling and feature perception. Experimental results demonstrate: When trained with regression loss functions and a threshold exceeding 0.4, the LkaUNet achieves higher critical success index (CSI) and negative alarm probability (NAP), along with a lower false alarm rate (FAR), compared to the baseline U-Net model; Under classification-based loss training, LkaUNet achieved a CSI of 0.730 1, with corresponding detection metrics of 86.27% hit rate, 13.73% miss rate, and 18.45% false alarm rate. The study concludes that LkaUNet effectively models long-range spatial correlations in thunderstorm clouds, providing a robust deep learning solution for monitoring severe convective weather. The approach highlights the potential of attention mechanisms in meteorological applications.

Abstract:The short-time Fourier transform (STFT) is essential for non-stationary signal analysis in areas such as audio processing, communication systems, and real-time spectrum analysis (RTSA). However, in practical RTSA instruments, conventional hardware implementations of the STFT are typically restricted to fixed window functions and fixed hop sizes, and they consume excessive logic cells and multiplier resources to meet high-throughput, low-latency streaming requirements, making them difficult to deploy on resource-constrained or portable platforms. This paper proposes a novel parallel windowed short-time Fourier transform (PW-STFT) architecture that leverages FFT-based processing techniques. By integrating parallel window multipliers and runtime parameterizable multipliers based on canonical signed digital (CSD) encoding, the design flexibly supports arbitrary window and hop sizes while minimizing hardware overhead. Experiments on a 32-point STFT with a hop size of 8 show that the proposed PW-STFT architecture significantly reduces resource usage (658 slices and 24 DSPs) compared to previous approaches while still maintaining an acceptable signal-to-noise and distortion ratio (SINAD) of 40.31 dB. This balance between hardware savings and output fidelity makes the design well suited for real-time STFT applications across a wide range of window types and signal conditions. Therefore, the PW-STFT architecture provides a flexible and resource-efficient solution for real-time spectrum analysis, enabling high-throughput and precise time-frequency analysis of non-stationary signals.

Abstract:Traditional energy harvesters have problems such as low energy conversion efficiency and narrow operating frequency bands. To obtain a high-performance and low-cost piezoelectric energy harvester, a variable vibration angle dual-end resonance wide-body piezoelectric energy harvester is proposed. This harvester uses a centrally symmetric widened cantilever beam as the vibration carrier, with the clamping end located at the central axis, and works in a mode of mid-span clamping with dual vibration end excitation. The vibration angle is changed by moving the mass block on the wide beam surface, and the dynamic response characteristics of the system are optimized through this design. In this study, a vibration exciter experimental platform was built and an experimental prototype was made to carry out controlled variable experiments. The experimental results show that the harvester has the optimal output power under a certain external resistance; the change of vibration angle can improve the output performance of the harvester; the system output power increases when the mass block increases; at a clamping angle of 20°, excitation acceleration of 0.4g, and frequency of 17.8 Hz, the harvester reaches the optimal output power of 4.66 mW; the output performance and resonant frequency can be changed by altering the length of the force arm, and the operating frequency band can be widened by paralleling multiple devices with different force arms. This study realizes the optimization of dynamic response under the composite deformation of the piezoelectric beam through the centrally symmetric cantilever beam structure and variable vibration angle design, providing a new theoretical basis for vibration energy harvesting.

Abstract:Aiming the serious distortion problem caused by the measurement error of measurement while drilling (MWD) magnetometer, an online calibration method of MWD magnetometer based on federated recursion extended Kalman filter (FREKF) is proposed. Firstly, the federated fusion framework of drilling tool geomagnetic information is established. Based on the error calibration model of magnetometer while drilling, the nonlinear measurement equation is constructed as the main filter, and the sub-filter 1 is formed with the observed geomagnetic vector model of accelerometer and the sub-filter 2 is formed with the recurrent geomagnetic vector model of gyroscope. Secondly, during the sub-filter’s measurement update, we use state estimation results and its covariance matrix as one-step prediction results and covariance matrix to improve state estimation accuracy. Then the subsystem of cyclic filtering is federated to realize the high precision calibration of magnetometer. Finally, the simulation experiment and real drilling experiment are designed. The experimental results show that the FREKF can effectively calibrate the magnetometer error while drilling. Following calibration, the azimuth mean error can be controlled within 0.8°.

Abstract:To address the path planning requirements for automated visual inspection of workpieces with complex surfaces—such as automotive steering wheels—conventional decoupled approaches that first select viewpoints and subsequently plan paths are prone to local optima and struggle to simultaneously achieve high coverage and computational efficiency. To overcome this limitation, a coverage path planning method featuring joint optimization of viewpoints and trajectories is proposed. First, the axis-aligned bounding box of the target point cloud model is spatially subdivided, and candidate viewpoint positions are generated by applying random offsets to the centroids of the resulting subregions. Viewpoint orientations are then determined through sampling in spherical coordinates, yielding a redundant set of candidate viewpoints. Second, a frustum-based coverage evaluation model is established, incorporating constraints on depth of field, field of view, surface visibility, and self-occlusion to quantitatively assess the effective coverage capability of each candidate viewpoint. Finally, an enhanced hybrid grey wolf optimizer-whale optimization algorithm (IGWO-WOA) is introduced, integrating chaotic map-based initialization, opposition-based learning, a dynamic convergence factor, and the spiral hunting mechanism from the whale optimization algorithm to enable multi-objective co-optimization of viewpoint selection and visiting sequence. Experimental results on two complex steering wheel models demonstrate that, compared with conventional swarm intelligence algorithms, the proposed method reduces path length by 20.6% and 11.5%, achieves coverage rates of 99.7% and 99.86%, respectively, and ensures collision-free execution throughout the inspection process, thereby delivering significantly superior trajectory quality.

Abstract:To address the limitations of traditional fall detection methods, which are prone to false detections under occlusion and illumination interference and struggle to balance lightweight design with detection accuracy, this paper proposes an improved YOLOv8-based algorithm for indoor elderly fall detection. Specifically, omni-dimensional dynamic convolution is integrated into the C2f module of the backbone network, enabling adaptive feature extraction and enhancing representational capability. In the neck network, the C2f module is further optimized with the FasterNet module to effectively reduce computational cost. In addition, a large selective kernel attention (LSKA) mechanism is embedded into the SPPF module to improve detection precision, while a spatial-enhanced attention module (SEAM) is introduced into the detection head to further strengthen discriminative ability. Comparative and ablation experiments were conducted, and the detection results were further visualized as heatmaps to validate the effectiveness of the proposed approach. Experimental results demonstrate that, compared with the baseline YOLOv8n model, the improved algorithm achieves an mAP@0.5 of 91.0% (an improvement of 1.1%), with a 20.6% reduction in parameters and a 42.7% decrease in GFLOPs, thereby confirming that the proposed method effectively balances detection accuracy and lightweight design in indoor elderly fall detection tasks.

Abstract:This paper presents the design of an optimized Cartographer-SLAM based autonomous navigation system for mobile robots. It aims to address the problems of mismatches and accumulated errors in loop closure detection encountered by the traditional Cartographer algorithm in long corridors, repetitive structures, and dynamic environments. The proposed method improves robustness and accuracy by enhancing loop closure detection and adopting adaptive optimization strategies.The main improvements involve using dynamic time warping for loop candidate selection and dynamic threshold adjustment to reduce computational redundancy. A Bayesian optimization mechanism is applied to fuse grid matching and temporal matching scores, with adaptive weight tuning according to environmental characteristics. In the back-end optimization, a confidence-propagation-based dynamic weighting scheme is introduced to suppress the impact of false matches on map consistency.Experiments are conducted in Gazebo simulation and real-world scenarios. In simulation, the loop closure error is reduced by 23% in “日”-shaped corridors and factory shelf environments. In the corridor of Teaching Building 9 at Nanjing Forestry University, the error decreases from 0.52 to 0.31 m. Tests on the ACES Building dataset show that the proposed algorithm outperforms mainstream methods with good generalization. The AGV navigation system also performs well in dynamic obstacle avoidance and path planning.This work provides an efficient and low-cost solution for autonomous navigation in complex environments and exhibits high engineering application value..

Abstract:X-ray computed tomography (CT) has become a fundamental tool in medical diagnosis, biological research, and industrial inspection due to its high resolution and non-destructive nature. However, conventional high-resolution CT imaging relies on dense projection data acquisition, resulting in prolonged imaging times and increased risk of radiation damage to samples. To balance image quality and radiation safety, sparse-sampling CT reduces the amount of projection data, thereby lowering radiation exposure and shortening imaging duration. Yet, this approach often introduces severe aliasing artifacts in reconstructed images, hindering further structural analysis. To address this issue, this study proposes an enhanced non-local means algorithm that incorporates an adaptive anisotropic field-of-view (FOV) kernel and a bilateral weighting function to effectively suppress aliasing artifacts in sparse-sampled CT images. The algorithm dynamically adjusts the FOV kernel to capture local structural features, significantly improving feature fidelity compared to traditional isotropic FOV kernels. Additionally, by employing a bilateral weighting strategy, the algorithm assigns higher weights to similar image patches, enhancing noise suppression while preserving critical details. Experimental results demonstrate that the proposed method substantially improves image quality on both simulated and experimental datasets, with minimal disruption to original structural details. Quantitative evaluation shows improvements of 15.9% in contrast-to-noise ratio (CNR) and 7.2% in structural similarity (SSIM) index compared to the classical non-local means algorithm, confirming its potential for sparse-sampled CT imaging. The proposed enhanced non-local means algorithm effectively improves reconstructed image quality and provides a viable solution for artifact suppression in sparse-sampling CT.

Abstract:To address the issue of reduced positioning accuracy caused by joint friction in Scara robots, an improved glowworm swarm optimization is proposed to identify the parameters of the friction model. Two optimizations are made on the basis of the traditional algorithm: by combining the Levy flight strategy and inertia factor, non-Gaussian random walks and adaptive inertia weights are utilized to randomly initialize fireflies trapped in local optima, enhancing the algorithm’s global search capability; the simulated annealing algorithm is introduced to perform local annealing operations on potential optimal solutions, improving the algorithm’s local optimization ability. Through performance analysis of test functions and parameter identification experiments, the results show that the improved artificial firefly algorithm has better optimization performance compared to other optimization algorithms. Finally, to further verify the effectiveness of the friction model identified through the algorithm, a fuzzy PID controller based on friction compensation is designed for the robot trajectory tracking control experiment. The experimental results indicate that the identified friction model has high accuracy, and the proposed control method can effectively suppress the adverse effects of joint friction on the trajectory tracking control of Scara robots compared to using only the fuzzy PID control method. The position tracking errors of the two joints of the robot are reduced by 76.1% and 81.9% respectively, further improving the positioning accuracy of the robot.

Abstract:The inertia value of the motion simulator of the docking performance testing platform (‘performance testing platform’ for short), applied for docking mechanism ground docking performance tests used in Lunar Exploration Program, can affect the accuracy of ground test data. As the inertia of the motion simulators for the performance test-bed must be exactly the same as those of the spacecrafts which the test-bed simulated, there is the great practical significance to measure the inertia of the motion simulators in-situ. In this paper, the motion vibration equation and in-situ calibration method of rotational inertia for the motion simulators is firstly investigated based on the additional mass spring oscillator method. Secondly, the mechanism motion models of the motion simulators are built based on SimMechanics Tollbox Software before and after attaching additional counterweights, the vibration periods of the two models are simulated, and the simulation inertia is calculated based on the theoretical model and the simulated vibration periods, verifying the correctness of the theoretical algorithm. Finally, a calibration device based on voice coil motor and limit mechanism is established for in-situ calibration tests of the motion simulator and analyze the main sources of errors in the tests. Experimental results show that the calibration results of the inertia of the three axes have a deviation of no more than 5% compared to the theoretical values, and the measurement error caused by the device is less than 2.5%, which is less than 1/4 of the maximum permissible error of the motion simulators. The results indicate that the calibration device can solve the problem of inertia calibration for the motion simulators of the performance testing platform.

Abstract:Polyethylene pipes are widely used in critical fields such as energy transmission and water supply and drainage. Accurate detection of their locations is of great significance for ensuring urban safety. In view of the problems of noise sensitivity and inaccurate time delay calculation in existing pipeline location technologies, this paper proposes an elastic wave imaging location method based on joint denoising and mutual information time delay. Firstly, an experimental platform is built and the signals received by the geophones are collected. Secondly, the signals are denoised by using the ensemble empirical mode decomposition combined with the wavelet threshold algorithm. Then, the time delay of the signals is calculated by the mutual information function method, and the pipeline location is determined by superimposing the data of multiple survey lines. Finally, the feasibility of the method is verified by COMSOL simulation, and field tests are conducted on PE pipes buried at depths of 0.5 and 1 m. The experimental results show that the average positioning errors of the pipelines obtained by the method in this paper are 0.042 and 0.085 m, respectively, which are reduced by 0.058 and 0.820 m compared with those without denoising. Compared with the cross-correlation function method and the time-domain superposition method, the average positioning errors of the method in this paper are reduced by 0.067 and 0.063 m at a depth of 0.5 m, and by 0.222 and 0.057 m at a depth of 1 m, respectively. The research results show that this method significantly improves the positioning accuracy of pipelines and has important engineering application value for the detection of buried PE pipes in complex environments.

Abstract:Compared with conventional non-flowing particles, an increase in flow velocity of aerosol particles can exacerbate the ill posedness of the inversion equation and increase sensitivity to noise, which makes it difficult to obtain accurate regularization parameters. To improve the accuracy of selecting regularization parameters for flowing aerosol particles, an improved grey wolf algorithm based on weighted Morozov discrepancy (WMD-IGWO) is proposed to optimize the regularization parameters on the basis of the traditional Morozov discrepancy principle. This method obtains the noise component of the electric field ACF through wavelet packet decomposition, and establishes an objective function by weighting the deviation function based on the noise component, which can reduce the impact of noise on the data. The convergence factor of GWO is nonlinearly improved, and the objective function is incorporated into the IGWO for global optimization to obtain the optimal regularization parameter, thereby enhancing the accuracy of the inversion results. The inversion results of four simulated aerosol particles (292, 483, 167/575, 208/733 nm) at different flow velocity show that compared with the L-curve method, the WMD-IGWO inversion results in smaller distribution errors and peak position errors of particle size distribution (PSD), and higher accuracy of inversion results. The inversion results of 584 nm unimodal and 243/825 nm bimodal measured particles show that WMD-IGWO can reduce peak position errors by up to 0.041 and 0.116/0.087, respectively, which is superior to the inversion results of the L-curve method and verifies the conclusions of the simulation experiment.

Abstract:The analog time-register-based time-domain signal processing circuit (refer as analog time-domain circuits for short) is a current research hotspot, featuring lots of merits such as low power consumption and strong anti-interference ability. However, it suffers the time register-overflowing issue. This paper proposes and verifies an automatic calibration circuit that can effectively solve the problem of time register overflow caused by process-voltage-temperature (PVT) variations in analog time-domain circuits. The circuit mainly consists of a time register, a digital control algorithm, a time error detector, and a programmable current source. By using the digital control algorithm in combination with the time error detector, it can detect in real time whether the time register is at risk of overflow. When an overflow risk is detected, it will adaptively modify the programmable bias current to adjust the capacitor discharging speed in the time register therefore eliminate the risk. The calibration circuit is designed and fabricated using the SMIC 0.18 μm CMOS process. The chip test results show that, without using this calibration circuit, when the temperature and power supply voltage fluctuate, the time register bears the overflowing risk and a time deviation of the analog time-domain signal greater than 15% is observed. When using this calibration circuit, the time register runs stably, and the time deviation is less than 1%. This design can be integrated into kinds of analog time-domain signal processing circuits such as time-to-digital converters and all-digital phase-locked loops, to enhance their robustness.

Abstract:To address the problems of low computational efficiency, excessive node exploration, poor safety, and non-smooth path in large-scale continuous obstacle environments, this paper proposes an improved A* algorithm. Firstly, eight types of two-layer 5-neighborhood are proposed to improve computational efficiency and enhance path smoothness, and boundary expansion and full-obstacle expansion methods are designed to resolve the deadlock caused by small neighborhood search. Secondly, a hierarchical heuristic function strategy is proposed. The strategy divides the search space into multiple layers and assigns different weights to the heuristic functions of each layer based on predefined thresholds, thereby reducing the number of explored nodes and further improving computational efficiency. Finally, a safety detection method is proposed to ensure that the generated paths maintain a safe distance from obstacles. Compared with five algorithms in different environments, simulation experiments demonstrate that the improved A* algorithm reduces running time by an average of 30.9%, increases path safety by an average of 14.7%. In addition, the improved A* algorithm generates paths with moderate smoothness and achieves better overall performance than the five compared algorithms. The proposed improved A* algorithm not only satisfies the requirements for safe and efficient path planning for mobile robots in large-scale continuous obstacle environments but also shows greater robustness in different environments. Even without secondary path optimization, the improved A* algorithm generates paths with relatively high smoothness.

Abstract:This paper addresses the issue of image quality degradation in reconstructed holograms under noisy conditions through computational holography. By establishing a dual-metric evaluation system based on peak signal-to-noise ratio (PSNR) and structural similarity index measure(SSIM), we systematically investigate the influence of noise on the reconstruction quality of four typical types of computer-generated holograms. Based on MATLAB simulation platform, a comparative analysis is conducted on the performance of detour-phase holograms (Lohmann type Ⅲ), modified off-axis reference beam encoding (Burch’s and Lee’s encoding), kinoforms, and computer-generated holographic interferograms under different noise coefficients. The results indicate a clear correlation between the noise coefficient and the quality of reconstructed image, the PSNR values decrease with increasing noise levels, with a more pronounced decline observed in the low-noise range (ef1≤0.3). Moreover, the SSIM metric demonstrates greater overall stability, particularly excelling in modified off-axis reference beam encoding methods. Optimization of encoding strategies effectively enhances noise resistance—for instance, Lee’s encoding maintains a PSNR of approximately 19 dB at ef1=1.0 through orthogonal component decomposition and multi-level grayscale quantization, while Lohmann type Ⅲ encoding achieves a 94.2% improvement in PSNR by increasing the aperture width. Although continuous-tone holograms outperform binary holograms in detail retention, binary methods remain practically valuable in specific application scenarios. This study provides a theoretical basis for selecting computer-generated hologram (CGH) encoding schemes under various noise environments.

Abstract:Because the measurement techniques using global navigation satellite system (GNSS) have high accuracies, GNSS users can obtain accurate zenith tropospheric delay (ZTD) products. To use GNSS-measured ZTD products, the separation between zenith hydrostatic delay (ZHD) and zenith wet delay (ZWD) is an important step. In this separation step, high accurate ZHD models are required. This study proposes a new ZHD model—the NNSZHD model, specifically designed for the separation between ZHD and ZWD. The development of the NNSZHD model is based on a novel modeling approach, which combines the ZHD values obtained from the direct calculation method and those from the indirect calculation method and is carried out by using an artificial intelligence algorithm. The direct calculation method refers to ZHD values calculated from ZHD model without measured meteorological parameters; while the indirect calculation method refers to ZHD values obtained by subtracting ZWD values (from ZWD model without measured meteorological parameters) from ZTD measurements. Then this study is based on this modeling framework and the GTrop model, which is a state-of-the-art ZTD model without measured meteorological parameters. Three key parameters (ZHDGTrop and ZWDGTrop obtained from the GTrop model, as well as ZTD measurements) are set as the modeling parameters. A multilayer feedforward neural network is used to develop the modeling framework of the NNSZHD model. The NNSZHD model is trained using data from 389 global sounding stations and the evaluation of its accuracy is carried out by using data from an additional 375 stations. The results show that for the region near the surface, the NNSZHD model has an accuracy of 12.35 mm, and its accuracy improves by 3.78 and 3.40 mm respectively compared with the GTrop and GPT2w models. For the tropospheric region with heights below 10 km, the model’s accuracy reaches 7.52 mm, and it shows improvements of respective 7.08 and 35.13 mm compared with the GTrop and GPT2w models. When we can only rely on ZTD models without measured meteorological parameters, the NNSZHD model has significant potential applications in GNSS tropospheric delay corrections and GNSS meteorology.

Abstract:Radar incoherent interference can cause an increase in noise floor and generate false plots. This exacerbates data association ambiguity in multi-extended target tracking, leading to track fragmentation and identity confusion, consequently causing errors in target state and shape estimation. Therefore, this paper proposes a joint spatio-temporal clustering and probability hypothesis density (PHD) based radar anti-interference method for multi-extended target tracking. Firstly, addressing the random time-varying characteristics of both the number and locations of extended target scattering points, the random finite set (RFS) theory is employed to model the multi-extended target state and measurement sets. High-quality partitioning of the dynamic, time-varying measurement sets is achieved by integrating spatio-temporal clustering. This approach not only avoids the complex explicit data association operations introduced by false targets but also resolves the partition explosion problem arising from increased measurement set dimensionality. Furthermore, by leveraging the probability hypothesis density (PHD) function, the interference from false targets is eliminated through the weighted summation of the Gaussian distributions and inverse Wishart distributions corresponding to different measurement subsets. This enables precise tracking of both the motion trajectories and shapes of multiple extended targets. Finally, experimental results under scenarios involving crossing trajectories, varying signal-to-noise ratios (SNR), and variable target numbers demonstrate that the proposed method achieves tracking points with optimal sub-pattern assignment (OSPA) distance below 0.5 m for over 95% of instances. It outperforms the K-means clustering-based PHD filter, the extended target generalized inverse Wishart PHD (ET-GIW-PHD) filter, and random matrix-based methods.

Abstract:The aircraft flat-tail honeycomb sandwich composite plate has complex structure, large area and many types of defects. The bonding state between honeycomb and skin can be intuitively analyzed by ultrasonic C-scan imaging. As a result, a large number of detection images need to be evaluated by the rich work experience of technicians, there are problems such as low evaluation efficiency and strong subjectivity. Therefore, an intelligent classification technology of ultrasonic C-scan image of honeycomb sandwich composite plate bonding layer based on deep learning is proposed. Firstly, an ultrasonic C-scan testing image of the interface between the adhesive layer and the honeycomb is acquired by an interface reflected wave tracking method, and the image quality is further improved by combining an image processing technology. Secondly, in order to construct the training database, C-scan image samples are intercepted by sliding window, and the samples are divided into three bonding states (target areas) according to the amplitude distribution of C-scan, and a small sample image data set expansion method is proposed. Finally, the 50-layer residual network (ResNet50) is constructed and trained, and the classification ability of the deep learning network for honeycomb bonding states is evaluated. The results show that the interface reflection wave tracking can overcome the shape change of the skin surface and form the C-scan image of the honeycomb bonding layer, and the ResNet50 network can identify the three types of target areas of the honeycomb sandwich composite panel structure with good stability and accuracy, and reflect the characteristics of “intelligence”.