Parallel compact artificial protozoa optimizer algorithm and its application in 3D coverage

Explaining explainability: A comprehensive survey on explainable artificial intelligence and relevant industry applications

Aiming at the problems of poor adaptability and low operational safety of agricultural machinery in hilly and mountainous areas, this study conducts the design and experimental research on a high-precision autonomous positioning system for the multi-crop combined harvester. The Beidou and Inertial Measurement Unit (IMU) high-precision positioning system based on Network Real-Time Kinematic (NRTK) was designed, and the hardware selection and software development of the autonomous positioning system were completed. The NRTK fixed base station in complex field environments was deployed Based on 4G communication, and dynamic differential calculations were performed with the random positions of the machinery to accurately obtain positioning data. Using the STM32L475 chip as the core information processor, efficient processing of real-time position information of combined harvester was realized based on coordinate conversion and attitude error correction of the autonomous positioning system. Performance tests were carried out using a self-developed the combined harvester. The results show that: the positioning error of the designed autonomous positioning system is less than 2 cm; when the harvester's speed ranges from 0.4 to 1.2 m/s, the maximum speed measurement error is less than 0.05 m/s, and the average speed measurement error is approximately 0.014 m/s, which meets the autonomous positioning accuracy requirements of the machinery.

High-fidelity neutronic analyses of advanced reactors require deterministic transport solvers capable of handling complex unstructured geometries while maintaining computational efficiency. This work presents the development and verification of three graphics processing unit (GPU)–accelerated deterministic solvers implemented within a unified framework, Neutronics using Deterministic Finite Element Algorithm (NuDEAL): the planar method of characteristics (MOC) coupled with the hybrid finite element method (HFEM), the discontinuous Galerkin method of characteristics (DGMOC), and the discontinuous finite element discrete ordinate method (DFEM-SN). These solvers provide complementary capabilities for consistently solving the multigroup transport equation and can be selectively employed to balance accuracy, computational cost, and memory requirements for a given problem. All the methods emphasize efficient GPU execution by leveraging memory alignment, compressed flux storage, and sequential azimuthal sweeps. The solvers are validated on the C5G7 benchmark and applied to advanced reactor problems, including the Advanced Burner Test Reactor (ABTR), Empire microreactor, and the Molten Salt Reactor Experiment. DFEM-SN achieved the highest accuracy, with eigenvalue errors below 50 pcm, while MOC/HFEM and DGMOC provided superior efficiency, with single-GPU run times comparable to those of large CPU clusters. The results demonstrate that deterministic GPU solvers on unstructured meshes can deliver both accuracy and scalability, enabling practical whole-core simulations for heterogeneous advanced reactors. The unified NuDEAL framework establishes a foundation for future extensions toward transient and multiphysics analyses on large-scale GPU architectures.

Coherent scattering, as the primary photoatomic scattering interaction in the low-energy region, is crucial for material structure analysis and nondestructive testing. Nuclear data processing codes typically employ the independent atomic form factor (IAFF) approximation to calculate the angular distribution of secondary photons from coherent scattering. Although this approximation meets accuracy requirements for most energy ranges, it becomes invalid when the incident photon energy approaches atomic absorption edges or falls below the kilo-electron-volt–energy range. This inaccuracy is due to the influence of anomalous scattering and molecular interference effects. In this paper, an improved model incorporating anomalous scattering and molecular interference effects is developed by introducing anomalous scattering factors (ASFs) and molecular interference functions (MIFs), respectively. These two effects are also implemented in the generation of the photoatomic ACE library for practical applications. For verification, the angular distribution is calculated with the IAFF approximation and the model incorporating ASFs and MIFs, respectively. The numerical results indicate that anomalous scattering results in a significant increase in forward scattering, while the inclusion of molecular interference effects yields angular distributions in better agreement with experimental measurements.

INTRODUCTION: With the emergence of new equipment and technologies, the difficulty of operation and maintenance (O&M) of power distribution equipment (PDE) has been continuously increasing. Traditional manual supervision and monitoring methods have been unable to meet the requirements of real-time performance and accuracy. OBJECTIVES: In order to effectively reduce operational safety risks, we propose an intelligent O&M violation detection method. METHODS: This paper optimizes the architecture of YOLOv12 and constructs three models: a security tool violation carrying recognition model, a general violation operation behavior recognition model, and a specific task violation operation behavior recognition model, this paper also uses the 3D electronic fence and real-time acquisition of each operator's 3D joint coordinates, and predicts the 3D joint coordinates of operation and maintenance personnel based on the Kalman filter. RESULTS: The method achievies accurate detection of O&M violations. In addition, this paper successfully establishes a 3D electronic fence for the O&M environment of PDE, and also achieves the recognition and early warning of violations related to spatial locations. CONCLUSION: The intelligent analysis and evaluation system for power distribution equipment operation and maintenance safety based on multimodal data fusion developed based on this method has been deployed and applied in the PDE O&M environment, achieving intelligent recognition of violations in power distribution equipment operation and maintenance and significantly improving the level of intelligence in on-site safety control.

As the primary load-bearing structure of deployable mesh antenna reflectors, the surface accuracy of cable-net structures directly determines the performance of cable-net antennas. To meet surface accuracy requirements, installed cable-net antennas must undergo surface adjustments, making the measurement of cable tension very important. However, constrained by measurement capabilities and conditions, large-scale cable tension measurement is highly challenging. To address this issue, this paper proposes a piezoelectric-integrated cable-net structure. By embedding piezoelectric patches at the nodes of the cable-net structure, the deformation of crimp terminals is converted into voltage signals via the direct piezoelectric effect. Furthermore, a cable force prediction method based on a BP neural network is introduced for piezoelectric-integrated cable-net structures. This method uses piezoelectric voltage values as the input layer and self-stress equilibrium factors of the cable-net as the output layer, thereby reducing the complexity of cable force prediction. Building on this, the influence of the quantity and placement of piezoelectric patches on the accuracy of the cable force prediction model is investigated. The study demonstrates that accurate prediction can be achieved when the number of piezoelectric patches is greater than or equal to the number of self-stress equilibrium factors. Additional piezoelectric patches and asymmetric placement can further enhance the prediction model’s accuracy. Finally, the predictive model was validated in triangular, quadrilateral, and tensegrity cable-net structures, demonstrating the validity of the cable force prediction method based on the backpropagation neural network. This work leverages neural networks to provide a new approach and solution for predicting cable forces in piezoelectric-integrated cable-net structures.

Widely used cone-beam computed tomography (CBCT)-guided irradiators struggle to localize soft-tissue targets due to low imaging contrast. While bioluminescence tomography (BLT) offers a promising functional imaging solution, its adoption in pre-clinical radiotherapy research has been limited. To address this, we developed MuriGlo, a novel BLT system compatible with CBCT-guided small animal irradiators to support high-precision radiation studies. We demonstrate MuriGlo's capabilities in supporting both in vitro and invivo experiments. MuriGlo consists of a detachable mouse bed, thermostatic control, mirrors, filters, and a charge-coupled device (CCD) camera, enabling multi-projection and multi-spectral bioluminescence imaging (BLI). The detachable bed facilitates animal transfer between MuriGlo and an irradiator for BLT-guided radiation study. We evaluated the thermostatic control's ability and demonstrated that it can maintain a consistent animal body temperature at 37°C throughout imaging. We also quantified detection sensitivity via signal-to-noise ratio (SNR) in detecting minimal cell quantities using glioblastoma (GL261) cells with Luc2 and AkaLuc reporters. The optical system can detect as few as 1173 GL261-Luc2 and 61 GL261-AkaLuc cells in vitro at SNR = 5. For image-guided capabilities, we present BLT-guided 5-arc, BLT-guided 2-field box, and BLI-guided single-field plans. The high conformal 5-arc plan fully covers gross tumor volume (GTV) at prescribed dose with minimal normal tissue exposure from moderate to high dose range, while the simplified, high-throughput BLT-guided 2-field box achieves 100% GTV coverage but results in larger normal tissue exposure. The choice of the planning strategy should therefore depend on the specific requirements for radiation accuracy and experimental throughput. Moreover, we compared MuriGlo's tumor localization accuracy for widely used irradiators, SARRP and SmART+. The localization accuracy of MuriGlo for both SARRP and SmART+ irradiators is < 1 mm with GTV coverage > 97%. This universal, BLT-guided platform enables plug-and-play integration with commercial irradiators, supporting functional image guidance and enhancing high-precision preclinical radiation research.

Abstract In response to the issue of low color contrast against similar backgrounds and occlusion from clustered growth during oblate jujube harvesting, this paper proposes a novel network architecture, termed BEA-Net, which is designed from the ground up for real-time, robust detection in agronomic environments. The proposed method integrated the Bottleneck Transformer (BoT-CTR3), which incorporated multi-head self-attention, enhancing the local feature extraction ability and improving the detection baseline. Subsequently, the hybrid attention mechanism of Efficient Multi-scale Channel Attention (EMCA) was embedded into the backbone network, which improved color feature discrimination under similarly colored backgrounds. Finally, the Adaptively Spatial Feature Fusion (ASFF) module was integrated into the detection head, strengthening the model’s contextual information extraction to compensate for information loss in occluded regions. Experimental results demonstrated that the improved network achieved precision, recall, F1-score, and mean average precision (mAP) of 94%, 92.2%, 93%, and 97.8% respectively—surpassed the baseline network by 3.2, 3.6, 3.0, and 2.1 percentage points. Comprehensive ablation studies validated the contribution of each component, while comparative evaluations against 14 state-of-the-art single-stage detectors confirmed BEA-Net's superior performance. To verify generalizability, the algorithm was tested on our self-built crisp persimmon dataset, achieving 96.9% mAP and 97% F1-score—represented improvements of 2.3 and 4.0 percentage points respectively. Furthermore, deployment on three distinct Android mobile devices achieved an optimal inference time of 50 ms. Benchmarking against YOLOv5 variants (n, s, m) confirmed this performance fully satisfies the accuracy and real-time requirements of automated picking systems, significantly enhancing the algorithm's practical application value in real-world scenarios.

Uneven seed spacing, skewed stalk posture, and inconsistent planting depth remain major challenges in horizontal sugarcane planting. To address these issues, a semi-automatic transverse sugarcane planter integrating a supply–buffer–discharge seeder and multiple soil-engaging components was developed. The seed placement process and the interaction between stalk discharge and soil disturbance were investigated through Discrete Element Method (DEM) simulations and experiments. First, the working principle and key component parameters of the whole machine were determined. It integrated the processes of soil crushing, furrowing, seeding, ridge covering. In addition, a dynamic analysis was conducted on the inter-particle disengagement effect during the two-step seed filling process of lifting and discharging. Secondly, a discrete element simulation model for the entire process of soil-engaging seed arrangement operations was established for the machine. The effects of forward speed and seed outlet position were studied using a discrete element method (DEM) simulation model that coupled soil disturbance flow with stalk-seed discharge behaviour. Furthermore, a response surface methodology (RSM) experiment was performed on the seeding test bench to quantify the effects of guiding parameters on seed placement uniformity. The determination coefficient (R2) of the established regression model exceeded 0.9, indicating high prediction accuracy. The optimal collaborative parameter combination was optimized as follows: forward speed of 1.2 m·s−1, buffer inclination angle of 55°and supply roller speed of 26 r·min−1. After verification, the seed placement uniformity coefficient of the seeder reached 91.8 ± 1.4%, which met the expected accuracy requirements for horizontal planting.

The precise segmentation of small-sized defects on wood surfaces is critical for the quality grading of glued laminated timber (GLT). Existing semantic segmentation models face core bottlenecks in this context: high miss rates, blurred boundary localization, and excessive size measurement errors. To address these issues, this paper proposes an improved Defect-Mask2Former model that integrates an Attention-Guided Pyramid Enhancement (AGPE) module and a Defect Boundary Calibration and Correction (DBCC) module. Through synergistic optimization, the model achieved pixel-level precise segmentation. To support model training and validation, a custom image acquisition device was designed, and the PlankDefSeg dataset was constructed, comprising 3500 pixel-level annotated images covering five defect types across six industrial wood species. Experimental results demonstrate that on the PlankDefSeg dataset, Defect-Mask2Former achieved a mean Intersection over Union (mIoU) of 85.34% for small-sized defects, a 17.84% improvement over the baseline Mask2Former. The miss rate was reduced from 20.78% to 5.83%, and the size measurement error was only 2.86%, strictly meeting the ≤3% accuracy requirement of the GB/T26899-2022 standard. The model achieved an inference speed of 27.6 FPS, satisfying real-time detection needs. By integrating the model into the GLT grading workflow, a grading accuracy of 94.3% was achieved, and the processing time per timber was reduced from 30 s to 1.5 s, a 20-fold efficiency improvement. This study provides reliable technical support for intelligent GLT quality grading and offers a reference solution for other industrial surface defect segmentation tasks.

This paper presents two automated correspondence-search algorithms for stereo laser triangulation in multimedia (refractive) environments, enforcing coplanarity through forward ray tracing without iterative back-projection. By shifting computations from image to object space, the methods directly minimize either the skew distance between refracted rays or the distance between their intersections with a laser plane, yielding strict refractive geometry within a single optimization loop. The methods are validated on an automated wood conservation monitoring system, where a stereo camera system with a line laser operates over a water-filled conservation tank. This application provides the environment for both algorithmic efficiency and accuracy requirements, demanding sub-millimeter precision over extended monitoring periods. Both algorithms reconstruct timber structures with high quality with the added planar constraint substantially reducing noise and edge outliers while slightly lowering point density. Water-surface estimation achieves plane-fit RMS of about 0.2 mm with a similar ground sample distance (GSD) and agrees with independent ruler measurements within 2 mm, enabling water-level monitoring over 14 epochs and cross-validation with calibration-derived plane parameters. Reference measurements on a 230 × 230 mm plane and three 50 mm spheres yield sub-millimeter residuals, with the plane constraint providing higher precision and fewer outliers. Finally, the computational efficiency is evaluated, showing favorable results for one of the algorithms compared to a standard procedure for correspondence search. Results demonstrate efficient, accurate, and transferable stereo laser triangulation through water and straightforward extensibility to multi-camera systems or non-perpendicular laser–water incidence angles. • Two novel correspondence search approaches for refractive stereo laser triangulation. • Application of the approaches on an automated wood conservation monitoring system. • Evaluation of the proposed approach against reference shapes. • Measurement of water planes with a stereo laser triangulation system. • Validation of water plane parameters with independent measurements.

Visual object counting has recently shifted towards class-agnostic counting (CAC), which addresses the challenge of counting objects across arbitrary categories—a crucial capability for flexible and generalizable counting systems. Unlike humans, who effortlessly identify and count objects from diverse categories without prior knowledge, most existing counting methods are restricted to enumerating instances of known classes, requiring extensive labeled datasets for training and struggling in open-vocabulary settings. In contrast, CAC aims to count objects belonging to classes never seen during training, operating in a few-shot setting. In this paper, we present the first comprehensive review of CAC methodologies. We propose a taxonomy to categorize CAC approaches into three paradigms based on how target object classes can be specified: reference-based, reference-less, and open-world text-guided. Reference-based approaches achieve state-of-the-art performance by relying on exemplar-guided mechanisms. Reference-less methods eliminate exemplar dependency by leveraging inherent image patterns. Finally, open-world text-guided methods use vision-language models, enabling object class descriptions via textual prompts, offering a flexible and promising solution. Based on this taxonomy, we provide an overview of 30 CAC architectures and report their performance on gold-standard benchmarks, discussing key strengths and limitations. Specifically, we present results on the FSC-147 dataset, setting a leaderboard using gold-standard metrics, and on the CARPK dataset to assess generalization capabilities. Finally, we offer a critical discussion of persistent challenges, such as annotation dependency and generalization, alongside future directions. We believe this survey offers a valuable resource for researchers, capturing the evolution of CAC and providing insights to guide future developments in the field. • Provides the first comprehensive survey on class-agnostic object counting (CAC). • Proposes a taxonomy: reference-based, reference-less, and open-world text-guided. • Benchmarks 30 CAC methods on FSC-147 and CARPK, setting a leaderboard. • Discusses trade-offs between accuracy, flexibility, and annotation requirements. • Identifies open challenges and future directions for generalizable counting.

With the continuous increase in the speed of high-speed trains (HSTs), traditional friction braking no longer meets the requirement for rapid stopping over short distances. Wind brake devices (WBDs) serve as a crucial supplementary mechanism to enhance braking performance. Therefore, investigating methods to improve their aerodynamic braking force under multifactor influences holds significant importance. This study presents a hybrid surrogate model (HSM) with dynamic weighting based on local error evaluation, which integrates the modeling characteristics of multiple classical surrogate models. The proposed model fulfills the fitting accuracy and adaptability requirements of surrogate models in the optimization of WBDs. Initial sampling points are generated using Optimal Latin Hypercube Sampling (OLHS), and their corresponding geometries are created through the Morph mesh deformation method. The responses of these sampling points are obtained using computational fluid dynamics (CFD) simulations. The accuracy of the numerical calculation method is validated through scaled wind tunnel experiments. Then, the NSGA-II algorithm is employed for optimization. The results indicate that the drag of the head car equipped with WBDs increases by 11.2%, while the lift decreases by 21%. Analyses of the flow field and pressure distribution further demonstrate that the optimized WBDs attenuate the large windward vortex, expand and intensify the leeward low-pressure region, and modify vortex morphology to raise flow separation and reattachment, collectively enhancing braking force and reducing lift. The application of the HSM considerably improves computational accuracy and efficiency.

Delayed resonator (DR) is an active vibration absorber that can enable complete vibration suppression through proper tuning of the actuation forces based on delayed (past) system states. However, the tuning requires exact knowledge of system parameters, thus causing residual vibrations which are sensitive to parametric inaccuracies/uncertainties. To eliminate such residual vibrations, we generalize an adaptation strategy to online correct the parameters of a classical control law by equating the effects of the force output at the vibration frequency to the alteration of the absorber’s stiffness and damping, thereby accommodating the inaccurate estimations in all parameters involved in the tuning process. Furthermore, the equivalent model also allows the resulting adaptive DR to compensate for the inaccurate realization of the control parameters arising from the inaccurate hardware parameters. Simulations and experiments both verify the effectiveness and the efficacy of the general adaptation strategy, which significantly reduces the accuracy requirement for system parameter identifications and modeling for tuning DRs to achieve complete vibration suppression, while maintaining the simplicity of the delayed control logic.

This paper investigates the security of a reconfigurable intelligent surface (RIS)-aided unmanned aerial vehicle (UAV) system for integrated sensing and communications (ISAC). A multi-antenna UAV transmits ISAC waveforms to simultaneously detect an untrusted target and communicate with a ground Internet-of-Things (IoT) device in the presence of an eavesdropper (Eve). To reflect practical scenarios where Eve's channel state information (CSI) is unavailable or concealed, we adopt an imperfect-CSI model. Our goal is to maximize the average secrecy rate subject to a sensing quality-of-service constraint by jointly optimizing the UAV trajectory, RIS passive beamforming, transmit beamforming, and receive beamforming. The resulting problem is highly non-convex due to the coupled design variables. We address this challenge with an efficient block coordinate descent framework, in which each block is solved via successive convex approximation built on semidefinite relaxation. Numerical results demonstrate that the proposed algorithm reliably meets sensing accuracy requirements and yields substantial secrecy-rate gains for the IoT link, even under CSI uncertainty, compared with representative baselines.

The tractor road, as the core scene for autonomous driving of grain transport vehicles, is unstructured, complex, and obstacle-rich, leading to poor real-time performance and accuracy of joint road and obstacle detection with existing YOLOv5s. Furthermore, the reliability of passable area evaluation is low solely based on environmental factors. Therefore, YOLOv5s-C2S is proposed, fusing multi-scale features, attention mechanism, and dynamic features for joint detection. Firstly, YOLOv5s-CC is proposed for road detection by fusing context and spatial details and introducing Criss-Cross attention. Secondly, YOLOv5s-SGA is proposed for obstacle detection by grouped and spatial convolution, parameter-free attention, and adaptive feature fusion. By reusing YOLOv5s-CC weights, YOLOv5s-C2S shares low-level features and decouples high-level specificity. Based on the tractor road and obstacle information, combined with vehicle factors, a weighted scoring–based comprehensive method for passable area evaluation is proposed. Finally, the method was verified through experiments with an intelligent tracked grain transport vehicle using self-constructed datasets, including VOC_Road (11,927 images) and VOC_Obstacle (21,779 images). Compared with existing YOLOv5s, Deeplabv3+, FCN, Unet and SegNet, the mAP50 of road detection by YOLOv5s-CC increased by over 1.2%. Compared with existing YOLOv5s, R-CNN, YOLOv7, SSD and YOLOv8n, the mAP50 of obstacle detection by YOLOv5s-SGA increased by over 2%. Compared with YOLOv5s-SD, the mAP50 of joint detection by YOLOv5s-C2S increased by 9.3%, and the frame rate increased by 7.0 FPS. The proposed passable area evaluation method exhibits strong robustness and reliability in complex environments, meeting the accuracy and real-time requirements in autonomous driving of grain transport vehicles.

Accurate GNSS timing is fundamental to Power Time Synchronization Systems (PTSS). However, conventional substation infrastructures remain vulnerable to sophisticated spoofing attacks. In this research, a sensing-assisted array antenna-based spoofing mitigation method is proposed. The proposed architecture operates at the signal front-end and incorporates a dedicated spoofing sensing path to estimate the Direction-of-Arrival (DoA) of malicious signals, enabling adaptive null steering while preserving authentic satellite reception. To provide reliable spatial reference for DoA estimation, a unified high-precision attitude determination method is developed for compact 10 cm-scale array antennas under single-frequency and environmental error conditions. The method integrates the Constrained Least-squares AMBiguity Decorrelation Adjustment (C-LAMBDA)-based constrained ambiguity resolution, redundant antenna element-based vertical accuracy enhancement, and iterative refinement to mitigate centimeter-level environmental biases. Semi-simulated experiments demonstrate that the proposed method achieves baseline vector Root Mean Square Errors (RMSE) below 5 mm in horizontal components and approximately 10 mm in vertical components. The resulting attitude accuracies reach 2° in heading, 6° in pitch, and 4° in roll, while eliminating over 80% of systematic environmental phase errors with an average convergence within 6 iterations. These results satisfy the spatial accuracy requirements for effective spoofing suppression and front-end signal purification. Consequently, a robust technical approach is established for enhancing the anti-spoofing capabilities of PTSS without modifying existing infrastructure.

Abstract We present the second release of the GR-Athena++ waveform catalog, comprising four new quasi-circular, non-precessing, spinning binary black hole simulations. These simulations are performed at high resolutions and represent a step toward generating high-fidelity gravitational waveforms that can eventually meet the accuracy requirements of upcoming next-generation detectors, including LISA, Cosmic Explorer, and Einstein Telescope. Gravitational waves are extracted at future null infinity using both Cauchy characteristic extraction and finite-radius extraction. For each simulation, we provide strain data across multiple resolutions and analyze waveform accuracy via convergence studies and self-mismatch analyses. The absolute phase and relative amplitude differences reach their largest values near the merger, while the smallest errors are of order O(10 -2 ) and O(10 -3 ), respectively. A self-mismatch analysis of the dominant (2, 2) mode yields mismatches between O(10 -5 ) and O(10 -7 ) for a total binary mass of 10 6 M ⊙ over the frequency range 0.002 to 0.1 Hz using LISA's noise curve. All waveforms are publicly available via ScholarSphere.

Abstract The quality of metallic surface coatings is crucial to the protected structures, and defect detection of coatings is an indispensable component of maintenance work. However, traditional manual inspection methods suffer from low efficiency, significant time consumption, and overly subjective standards dependent on experience, making it difficult to meet the real-time and accuracy requirements of modern engineering maintenance. This paper proposes an MCPB (Multi-scale Coupled Pyramid Backbone) network model based on multi-scale feature fusion to address key challenges in coating defect detection, including brutal and indistinct defect characteristics, sample imbalance, and limited data availability. The network model employs a downsampling Focus method to improve the backbone of the baseline network, enabling lightweight feature extraction; designs RFocus and Focus modules to complete the upsampling and downsampling processes of the feature pyramid, enhancing multi-scale feature representation; introduces a Layer Coupling Attention Module (LCAM) to achieve cross-layer feature fusion, strengthening inter-layer coupling; designs a Parallel Reconstruction Module (PRM) for feature purification of deeply fused features, effectively suppressing background noise interference; and integrates Enhanced Spatial Attention Module (ESAM) and Enhanced Channel Attention Module (ECAM) to guide precise extraction of multi-scale features. Through experimental preparation and data collection, a coating surface defect dataset is constructed and processed. Experimental results on this dataset demonstrate that the MCPB network significantly outperforms YOLOv8, YOLOv11, and the improved method YOLOv8-CM in terms of defect detection accuracy, exhibiting extremely high precision for detecting critical defect types such as sagging and spalling. This method can efficiently accomplish target detection of important defects in on-site maintenance and shows great potential for application in the maintenance of hydraulic steel structure surface coating engineering.