102results about How to "Reduce false detections" patented technology

The invention provides a rainfall monitoring method and system based on unmanned aerial vehicle multi-mode noise reduction in the technical field of environment monitoring and unmanned aerial vehicle application, and the method comprises the steps: S1, collecting a visible light image and a long-wave infrared image through a multispectral vision module after an unmanned aerial vehicle takes off, and collecting a rainfall audio from an acoustic cabin through a microphone; s2, carrying out noise reduction processing on the visible light image, the long-wave infrared image and the rainfall audio; s3, extracting a raindrop size distribution histogram and a spatial density thermodynamic diagram based on the visible light image and the long-wave infrared image, and extracting acoustic MFCC features based on rainfall audio; s4, performing feature fusion operation based on the raindrop size distribution histogram, the visual density thermodynamic diagram and the acoustic MFCC features to obtain rainfall intensity; and S5, carrying out cumulant dynamic compensation based on the rainfall intensity to obtain the cumulative rainfall. The rainfall monitoring method has the advantages that the precision, robustness and practical level of rainfall monitoring are greatly improved.

The invention discloses a pavement crack accurate extraction method and system based on multi-scale image segmentation, and belongs to the technical field of pavement detection, and the method comprises the steps: obtaining a to-be-detected pavement image, carrying out the brightness normalization and geometric correction of the pavement image, and constructing a multi-scale image set with different resolutions; generating a spatial texture channel and a spectral domain response channel for each scale image in the multi-scale image set, and fusing the spatial texture channel and the spectral domain response channel to obtain a spectral-space coupling input tensor, wherein the spectral domain response channel is obtained by extracting energy characteristics of a plurality of frequency bands after performing spectrum transformation on the scale image; a multi-scale image set is constructed for road surface images, and coupling expression of spatial texture information and spectral domain response information is introduced, so that feature consistency of fine cracks under different resolutions is enhanced, and scale drift and response distortion caused by shadows, light reflection and seam textures are inhibited at the same time; therefore, the separability and the stability of the crack in a complex scene are improved.

The present application relates to the field of digital image processing and information security, and discloses a document tampering detection method and system based on image data processing, comprising the following steps: first, extracting the noise residual and microscopic penetration characteristics of the document image, and constructing a physical potential energy field and a virtual viscous resistance field; then, using a Darcy law variant model for dynamic evolution, generating a virtual flow velocity vector field to simulate the sliding behavior of fluid in heterogeneous media; subsequently, constructing a heterogeneous graph based on the flow field divergence singular point and streamline trajectory, using a graph neural network to aggregate the node dynamics characteristics for deep reasoning, and finally generating a tampering positioning mask. The present application innovatively introduces fluid mechanics field theory, converts hidden static texture differences into significant dynamic flow field anomalies, solves the problem that the prior art is difficult to capture microscopic tampering traces, and significantly improves the detection accuracy and generalization ability in complex document scenarios.

The invention provides a directional ship detection method and system based on joint attention. The method comprises the following steps: acquiring image acquisition data; performing feature extraction by using a backbone network to obtain a multi-scale feature map, performing feature re-calibration on the multi-scale feature map, and outputting an enhanced target feature map; carrying out convolution operation on the enhanced target feature map by adopting self-adaptive geometric convolution to obtain ship direction characterization information; generating a candidate frame based on the ship direction representation information, calculating intersection-parallel-ratio statistical characteristics in combination with a ship real frame, and performing training sample adaptive distribution according to the intersection-parallel-ratio statistical characteristics; based on an Anchor-free detection framework, carrying out ship orientation detection by utilizing the training sample subjected to self-adaptive distribution to obtain a ship orientation detection result; through multi-scale feature re-calibration, adaptive geometric convolution and dynamic sample distribution, the ship direction and features in a maritime monitoring scene can be captured, and the detection precision and real-time performance are improved.

The application relates to the technical field of motor assembly quality image detection, and discloses a visual-based motor wire harness connection part abnormality detection method and system, wherein the method comprises the following steps: acquiring a motor wire harness connection part image; constructing a connection part illumination state vector; performing adaptive brightness correction; suppressing reflection and pollution interference; determining an effective wire harness connection part area; and outputting a wire harness connection abnormality detection result. Compared with the method in the prior art which mainly relies on single detection confidence for judgment, especially under complex working conditions such as strong reflection of a metal shell, oil stain adhesion and wire harness shadow superposition, the technical problem that it is difficult to stably distinguish between real structure abnormality and illumination artifact interference is solved. Due to the introduction of adaptive brightness correction and the joint mechanism based on attention enhancement target detection and assembly geometric constraint, the accuracy of motor wire harness connection part abnormality detection is improved.

This invention discloses an industrial material inspection method and system based on feature extraction and contrast enhancement, belonging to the field of image inspection technology. It includes: S1, acquiring the transmitted signal using a digital flat panel detector and generating an original digital image based on the differences in X-ray absorption by different materials or defects; S2, performing uniformity correction and normalization enhancement on the original digital image, and achieving preliminary differentiation of different material regions; S3, dividing the original image data into processing domains based on the preliminary differentiation results, and performing differentiated enhancement processing based on the local feature attributes of each processing domain to generate a binary feature image. By integrating X-ray penetration imaging, multi-dimensional image enhancement, and a deep learning-driven intelligent rating system, this invention fundamentally solves the key technical bottlenecks in traditional industrial material inspection, such as inconsistent image quality, lack of material identification, and strong subjectivity in rating, achieving an automated, intelligent, and objective upgrade in industrial material defect detection.

The application discloses a kind of ball grid array package chip solder ball bubble defect automatic detection method and system, it is related to chip intelligent detection technical field, the method is guided by designing semantic cross-layer fusion framework, realizes multi-scale feature balanced fusion in neck network by introducing feature alignment and redistribution module, strengthens small target details and semantic information using local semantic enhancement module, and the positioning accuracy of bubble is optimized in combination with smooth geometry positioning loss function.The application significantly improves the detection capability of small, fuzzy and occluded bubbles, has strong robustness and generalization, and can be widely used in solder ball bubble, material porosity and surface defect industrial vision detection scene.

This invention belongs to the field of laser ultrasound imaging technology, specifically disclosing a laser ultrasound imaging method based on neural networks, comprising the following steps performed sequentially: S1, acquiring the A-scan signal; S2, decomposing and reconstructing the A-scan signal through empirical mode decomposition; S3, extracting the temporal features of the ultrasound signal using pooling; S4, establishing a neural network and training it multiple times to obtain multiple neural networks with different representation capabilities; S5, inputting the temporal features of the ultrasound signal of the sample to be tested into each trained neural network, and filling the recognition results into the corresponding positions according to the detection point order to form a pixel matrix; S6, superimposing all pixel matrices, setting a threshold to evaluate the state of each detection point, and generating the corresponding pixel image. The method of this invention is insensitive to noise and can be used to process low signal-to-noise ratio signals, improving the accuracy of material state identification. This invention is applicable to detecting the state of materials.

The application relates to the technical field of computer vision, in particular to a power safety monitoring image detection method and system, which comprises the following steps: inputting a to-be-detected power safety monitoring image and a generated image into a trained power safety monitoring image detection model respectively to obtain a detection result, the power safety monitoring image detection model comprises a feature extraction network structure, a feature distillation network structure and a multi-scale aggregation network structure in sequence, the feature extraction network structure is used for extracting local spatial features and global context information features from the to-be-detected power safety monitoring image and each generated image respectively, and fusing the two to obtain fused features corresponding to each image; the feature distillation network structure is used for extracting Value values and Key values from the fused features of each image, and obtaining splicing features based on the Value values and the Key values of each image; and the multi-scale aggregation network structure is used for processing the splicing features to obtain the detection result of the to-be-detected power safety monitoring image.

The invention relates to the technical field of image recognition, and discloses a wind power fan blade defect recognition method and system based on image recognition, and the method comprises the steps: carrying out the blocking cutting, illumination normalization and image enhancement processing of an original image, constructing a defect-free sample and a defect sample, and dividing the samples into a training set and a test set; the method comprises the following steps: taking a GANopen network as a basic framework, fusing the lightweight design of Mamba-YOLO, constructing a joint loss function by adversarial loss, reconstruction loss and coding loss based on a training set, carrying out unsupervised training, optimizing network parameters until convergence, and obtaining a defect identification model; inputting the preprocessed to-be-detected fan blade image into the trained defect recognition model, performing defect recognition and positioning, and outputting the defect type and position; based on an output result of the defect identification model, dynamically adjusting an early warning level and response measures through a self-adaptive early warning mechanism; according to the invention, the efficiency and accuracy of wind power fan blade defect identification are improved.

This application discloses a method, device, and medium for training and applying a geoscientific big data feature extraction model, relating to the fields of geoscientific big data processing and artificial intelligence technology. The method includes: cropping geoscientific modal sample data to obtain several candidate cropping samples corresponding to each geoscientific modal sample data; extracting visual token sequences and text token sequences from the candidate cropping samples and text task prompts; calculating cross-modal entropy based on the visual token sequences and text token sequences, and determining the optimal cropping; inputting the geoscientific modal sample data, the optimal cropping, and the text task prompts into a large language model to obtain sample feature extraction results; and training a trained geoscientific big data feature extraction model using a total loss function, where the total loss function includes the cross-modal entropy based on the candidate cropping samples and the sample feature extraction results. This method avoids the occurrence of feature extraction fragmentation or missed detection problems, and also reduces background interference, thus reducing false positives or false negatives.

The invention discloses a fence structure-oriented small target double-stage detection and defect assessment method, which comprises the following steps of: inputting preprocessed RGB image data of a fence into a trunk convolution module of a defect risk assessment model to obtain first feature maps with different scales, fusing the first feature maps with polarization physical feature maps of the fence respectively, and obtaining a defect risk assessment result through a plurality of FPN modules; performing feature fusion on the fused feature map by adopting a cross-level feature fusion strategy and a cascade connection strategy, and sequentially outputting a fused second feature map to a channel attention module, a space attention module, a texture attention module and a region candidate module to obtain a candidate box set; when a candidate frame meeting the image reconstruction requirement exists, candidate frame reconstruction is carried out through an image super-resolution reconstruction module, RGB image data and a polarization physical feature map, then a bounding box, a defect category and a pixel mask are determined through an RoI feature processing module and a defect detection and segmentation module, and defect risk assessment information is obtained in combination with a geometric model of a fence.

The application belongs to the technical field of vehicle information anonymization detection, and particularly relates to a multi-type image anonymization annotation dataset construction and target coverage rate determination method. The method is based on a face and license plate image dataset with double annotation of theoretical anonymization region and anonymization features, introduces an anonymization feature extraction branch in an improved YOLOv5-L model and performs cross-modal feature fusion, combines a plurality of loss functions with theoretical anonymization region positioning loss as the core and a two-stage progressive training and difficult example mining mechanism, and realizes precise learning of the anonymization features. After normalizing the input anonymization image, the model inference obtains the theoretical anonymization region coordinates and maps them back to the original size, and through non-maximum suppression and matching of the IoU threshold, the region coverage rate is calculated to determine the positive detection, missed detection and statistical false detection rate. The application effectively overcomes the feature dependency failure and model robustness problem, and realizes high-precision, low-misjudgment anonymization detection and evaluation under various anonymization conditions.

The invention discloses a park weed accurate identification method and system based on an improved YOLOv11 deep convolutional network, and the method comprises the steps: 1, constructing a multi-scene park weed image data set, 2, carrying out the data enhancement and preprocessing, 3, constructing an improved YOLOv11 lightweight network model, 4, designing a self-adaptive loss function, and 5, carrying out the recognition of a multi-scene park weed image data set. Multi-stage model training and optimization are executed; and step 6, weed real-time identification and result output are realized. According to the method, a lightweight attention mechanism and an improved multi-scale feature fusion structure are introduced, the extraction and distinguishing capability of the network on small-scale weed features under a complex background is remarkably enhanced, the false detection and omission ratio is effectively reduced, and the dynamic label distribution strategy and a loss function fusing global context information are adopted, so that the robustness of the network is improved. And the detection robustness and the positioning precision of the model in dense, shielded and form-variable weed scenes are improved.

The invention discloses a defect feature recognition method based on a clothing visual image and a quality inspection system, and aims to solve the problems that traditional manual quality inspection is low in efficiency and poor in precision and an existing automatic system is insufficient in robustness and generalization ability. The method comprises the following steps: after image preprocessing, segmenting by using an SAM network to obtain a clothes main body and an initial flaw; constructing multi-scale features by the FPN, inputting the multi-scale features into the improved YOLOv11 network, and outputting an initial candidate box; correcting the candidate frame according to IoU / Dice to obtain a refined frame; extracting texture and morphological characteristics and calculating abnormal scores; carrying out weighted fusion on multiple indexes to obtain a comprehensive confidence coefficient to identify flaws; and carrying out secondary training to update the model after manually annotating the low-confidence sample. The system comprises eight corresponding modules, realizes full-process automatic detection, can significantly improve the precision, robustness and automation level of garment defect identification, and has important practical application value.

The application belongs to the technical field of application software development test, and discloses an application software development test system with real-time vulnerability detection, which comprises a code real-time collection module, a multi-dimensional vulnerability detection module, a vulnerability accurate positioning module, a vulnerability risk quantitative evaluation module, a dynamic repair guidance module, a data storage module, a visual interaction module and an iterative optimization module, and each module cooperates to form a whole-process closed-loop vulnerability detection and management and control system. The application software development test system with real-time vulnerability detection is adopted, real-time capture, accurate positioning, risk evaluation and dynamic repair suggestion output of the vulnerability are realized, and the software development test efficiency and application program security are improved.

This utility model provides an adjustable AOI inspection machine, belonging to the field of AOI inspection technology. It includes a support and a workpiece to be inspected, as well as a drive assembly. The drive assembly includes a slide rail fixedly connected to the top end; a flexible assembly including a connecting slider slidably connected to the slide rail. Pressure flaps are fixedly connected to both ends of the connecting slider's outer side, and a vertical plate is fixedly connected to the outer side of each pressure flap. A horizontal plate is fixedly connected to the end of the vertical plate away from the pressure flaps. A rack is fixedly connected to the top end of the support, and a gear is movably connected to the top end of the support. This design achieves high-precision horizontal transport, flexible support, and anti-deformation protection for the workpiece to be inspected, while maintaining stability and precise positioning during high-speed movement. It effectively solves the problem that fixed lens positions are difficult to adapt to products of different specifications. Precise transport and stable support ensure clear and accurate images are obtained during inspection, reducing missed or false detections and improving inspection accuracy.

The invention discloses a pipeline abnormal state recognition and positioning method and an inspection robot thereof, relates to the technical field of machine vision recognition, and can solve the problems of low recognition precision and high false detection and omission ratio of structural posture abnormity such as falling, inclination and hanging drooping of an underground coal mine pipeline in the prior art. The pipeline abnormal state identification and positioning method comprises the following steps: S1, constructing a global map of a coal mine tunnel, and establishing a coordinate corresponding relation between an image acquisition position and the global map; s2, acquiring a pipeline image in the roadway and position information corresponding to the pipeline image, and performing noise reduction processing on the pipeline image to obtain a to-be-identified image; s3, judging whether the pipeline is in an abnormal posture state or not, and outputting a pipeline abnormity recognition result; and S4, according to the position information corresponding to the pipeline abnormity identification result, completing the positioning of the abnormal pipeline.

The invention discloses a low-altitude flight collision detection method based on conical ray detection. Acquiring environmental space data and position and attitude information of the unmanned aerial vehicle in a three-dimensional space; the flight direction of the unmanned aerial vehicle is obtained according to attitude information processing of the unmanned aerial vehicle in the three-dimensional space, the position of the unmanned aerial vehicle serves as a conical vertex, the flight direction serves as a central axis, and a conical detection area is constructed in combination with the field angle of the unmanned aerial vehicle and a preset detection distance; discretizing the conical detection area to generate a plurality of sub-rays covering the conical detection area; detecting a spatial relationship between each sub-ray and a spatial obstacle according to the environmental spatial data to obtain spatial relationship information of each sub-ray; and adjusting the attitude information of the unmanned aerial vehicle according to the spatial relationship information of each sub-ray. A conical ray detection model is introduced, and compared with traditional single-straight-line ray detection, missing detection and false detection are reduced.

The application relates to a cross-layer structure feature fusion detection method for complex road surface disease identification, and relates to the technical fields of road engineering intelligent detection and computer vision target identification. The application solves the problems that the existing road surface disease target detection method still has insufficient local detail expression, insufficient disease boundary structure response and weak deep and shallow layer semantic information cooperation capability in complex road surface disease identification. The application forms a feature expression mode of shallow layer detail feature, middle layer structure feature and deep layer semantic feature cooperative modeling; through spatial scale adjustment and channel dimension adjustment, an alignment fusion mechanism of different layer features is constructed; further, a structure weight guiding strategy is introduced, and enhanced expression of disease related areas and key structure features is established. The application is mainly used for road surface disease identification.

This application provides a method, apparatus, and vehicle network cloud platform for vehicle system performance testing. The method includes: acquiring raw performance data reported by the vehicle system; performing anomaly cleaning and trip segmentation on the raw performance data to obtain performance time-series data corresponding to each trip; performing trend analysis based on the performance time-series data to obtain trend information corresponding to the performance time-series data; and determining the performance testing result based on the trend information corresponding to the performance time-series data. This method aims to improve the timeliness of performance anomaly detection.

The invention discloses a point cloud data efficient three-dimensional target detection method for an intelligent construction scene, and belongs to the field of engineering management, and the detection method comprises the following specific steps: I, periodically collecting and preprocessing construction site point cloud data, building a continuous time frame sequence, and constructing a construction twinborn model corresponding to a construction site; iI, aligning the semantic topology of the point cloud data and the construction twinborn model, identifying the abnormal area of the construction site, and then analyzing the evolution trajectory of each object on the construction site in real time; according to the method, the occurrence probability of false detection and unreasonable engineering semantics results is remarkably reduced, the detection stability of slowly moving or periodically changing targets is effectively enhanced, the problems of jumping and missing detection which are easily generated by single-frame detection are avoided, the recognition accuracy of shielded components in a complex construction environment is improved, and the construction quality is improved. And the overall detection efficiency is improved while the precision is ensured.

This invention discloses a method, apparatus, and product for detecting small targets in drone aerial photography, relating to the field of computer vision, to improve the detection performance and parameter efficiency of small targets in complex background scenes. In the visual feature extraction process, the aerial image is downsampled hierarchically using a spatial-to-channel rearrangement approach to obtain multi-layered downsampled feature maps. A multi-branch convolutional structure is used to enhance the receptive field of each downsampled feature map, resulting in multi-scale visual features. In the text semantic enhancement process, at least one cue vector most relevant to the aerial image is selected from a cue vector library. The cue vector is semantically enhanced to obtain enhanced semantic features, which are then fused cross-modally with the multi-scale visual features. A detection head is then used to identify the target based on the cross-modal fused features. This invention improves the accuracy, robustness, and generalization ability of small target detection in complex environments.

The application discloses a billiards event detection method, system, device and computer storage medium, relates to the technical field of computer vision and video analysis, and obtains a target video stream of a billiards table; determines a pocket area in the target video stream; for a frame image in the target video stream, determines pixel features of the pocket area in the frame image; according to the pixel features, determines local texture complexity of pixels in the pocket area; according to the local texture complexity, determines a ViBe detection threshold of the pixels in the pocket area; based on the ViBe detection threshold, identifies a pixel category of the pocket area through a ViBe algorithm, and the pixel category is used for representing whether the pixel belongs to a background or a billiard ball; and performs billiards event detection according to the pixel category. The application can keep stable pixel segmentation in scenes such as strong inverse light, weak light and color temperature mutation, reduces false detection caused by background texture fluctuation, and thus can perform billiards event detection according to accurate pixel categories, and improves detection accuracy.

This invention relates to the technical field of gas detection, and discloses an SF6 gas detector. 6 A dual-band differential imaging detection method and system for gas infrared images. The dual-band differential imaging detection system includes: a binocular imaging module for simultaneously acquiring a reference band infrared image and a probe band infrared image of the area under test; the reference channel of the binocular imaging module is equipped with a reference infrared filter, the film structure of which includes a multi-cavity interference structure; the probe channel of the binocular imaging module is equipped with a probe infrared filter, the film structure of which includes a single-cavity Fabry-Perot structure; and an image processing module for registering and differentially processing the reference band infrared image and the probe band infrared image to identify SF6. 6 The gas leakage area. The reference infrared filter and the detector infrared filter of the dual-band differential imaging detection method and system of this invention are complementary, jointly optimizing the signal-to-noise ratio of the detection system.

This invention discloses a 3D target detection method based on sparse dynamic attention and star-shaped interaction. The method obtains basic voxel features from the original LiDAR point cloud through voxelization and sparse convolution, then introduces a sparse dynamic parallel attention module. This module achieves efficient enhancement of global context and channel dimensions through dynamic attention branches and parallel channel interaction branches. A sparse star-shaped interaction module is then used to construct a star-shaped neighborhood interaction structure with a central voxel, completing local geometric modeling and nonlinear feature interaction only on non-empty voxels. Finally, keypoint sampling, RoI pooling, and a detection head output the 3D detection box, category, and confidence score. This invention, through the synergistic complementarity of SDPA and SSB, significantly improves the detection accuracy of long-distance, small-scale, and occluded targets while maintaining linear growth in computational complexity and meeting real-time requirements. It achieves balanced performance optimization across multiple categories, including vehicles, pedestrians, and cyclists, and is suitable for 3D perception scenarios with high precision and real-time requirements, such as autonomous driving.

The application discloses a diver handheld navigation sonar cross-modal navigation method and system, and relates to the technical field of underwater navigation. It aims to solve the problems of low navigation recognition, high power consumption and cross-modal fusion deviation in low visibility environment. The application includes synchronously collecting underwater acoustic, visual and positioning data; correcting the physical field of acoustic and visual data based on a fluid dynamics model and an electromagnetic wave propagation model; extracting geometric and texture features from the corrected data, while detecting acoustic anomalies and visual anomalies, and comprehensively outputting obstacle information; matching the extracted features with a pre-stored feature library for similarity, and outputting fusion features; generating a basic path based on the fusion features through a dynamic weight cost function, correcting the generated final path using positioning data, and planning a detour route; and superimposing and displaying the path, the detour route and the obstacle information, and giving a warning when the distance of the obstacle is less than a safety threshold. The technical scheme significantly improves the navigation recognition and reduces the power consumption in low visibility.

The invention discloses a cigarette carton missing detection method for a cigarette carton box. The method comprises the following steps: S1, collecting image data of the cigarette carton through a plurality of cameras when the carton is opened; s2, performing illumination adjustment on the acquired image data, and applying image enhancement processing to generate an enhanced image; s3, inputting the enhanced image into a target detection model, detecting whether the cigarette bar is missing, outputting a cigarette bar bounding box and confidence coefficient by the target detection model through extracting multi-scale features of the image and applying attention weight calculation, and judging a bar missing state according to a confidence coefficient threshold value; and S4, when it is detected that the cigarette bar is missing, triggering a data recording operation, recording related data of a missing event, including a timestamp, a box body number and an image fragment, and packaging, transmitting and storing the data. Through the steps of collecting images by multiple cameras, adjusting illumination, integrating a target detection model of an attention mechanism and automatically recording and transmitting data, the accuracy of detecting the carton lacking of the cigarette box of the cigarette piece is remarkably improved.

This invention discloses a method for real-time 3D tracking and localization of microscale targets in a bright-field microscopic environment, relating to the fields of image recognition and reconstruction technology. The method includes inputting continuous microscopic images into a target detection model, outputting the target's 2D position, regressing the defocus height and direction from a single frame of local images using a fine-grained ROI-based defocus regression model, directly using the model output to drive a 2D motion platform and a focusing device, fusing the 2D position information of each frame with the corresponding defocus axis information, and incorporating the position information of the previous frame as a priori using a temporal correlation mechanism to achieve continuous correlation of the microscale target in the time dimension, thereby stably reconstructing the target's 3D position and outputting a smooth real-time motion trajectory. This invention achieves real-time 3D localization and stable tracking of microscale targets under single-view bright-field microscopic conditions, ensuring both positioning accuracy and system real-time performance and accuracy.

The invention discloses a Mama enhanced industrial product surface defect detection method based on YOLOv11, and belongs to the technical field of defect detection.The method comprises the steps that preprocessed industrial product surface image data are input into an industrial product surface defect detection model; according to industrial product surface image data, feature vectors are extracted through a plurality of C3k2 modules in sequence, and the feature vectors are input into an SPPF module to obtain the feature vectors; according to the SPPF module, obtaining a feature vector, splicing the feature vector with a feature vector in any C3k2 module so as to obtain a feature vector through the Mama module, splicing the feature vector obtained by the Mama module with any other feature vector so as to obtain a feature vector through the Mama module, and repeating the steps; and randomly selecting the feature vectors obtained by a plurality of Mamba modules, and inputting the feature vectors into a plurality of detection heads in a one-to-one correspondence manner for detecting surface defects. According to the method, both the detection precision and the reasoning efficiency are considered, and particularly, the detection rate and the positioning accuracy of surface defects of complex, tiny and long-strip-shaped industrial products are remarkably improved.