342 results about "Gradient direction" patented technology

The invention relates to a machine vision-based meshing precision measurement calibration method for a gear transmission system, and belongs to the technical field of calibration workpieces. The method comprises the following steps: extracting a gradient direction of a contact edge point in an image, analyzing a direction change trend, removing a mutation point, and constructing an edge direction paragraph sequence; positioning a tooth profile contour intersection point according to the sequence, analyzing longitudinal coordinate fluctuation, and identifying an axial dislocation structure; comparing the actual tooth crest contour with a standard path, extracting the direction and length of a difference section, and generating an adjustment direction mark group; converting the direction mark into a fine adjustment angle sequence, collecting an image and identifying a direction stable section; and finally, dividing image sub-blocks, analyzing the curvature direction of a spliced boundary, identifying curvature mutation points and establishing a calibration scheme. According to the method, the integrity of gear meshing area structure recognition and the continuity of an image sequence are effectively improved, high-sensitivity axial dislocation detection and accurate direction guiding fine adjustment are achieved, the curvature sudden change recognition and multi-dimensional error response capability is enhanced, the engineering practicability and the automation level are high, and the method is suitable for large-scale popularization and application. The method is suitable for metering calibration and quality control of a high-precision gear transmission system.

The invention discloses a multi-modal learning method and device under different perception sources, equipment and a medium, and relates to the technical field of machine learning. Semantic energy scores of different modals are adjusted by utilizing learnable temperature parameters, and interaction features are weighted by utilizing semantic energy weight scores, so that the learning efficiency is improved. In the process, semantic quality of each mode is dynamically evaluated through an energy score, a noise mode is suppressed, key information is highlighted, and a temperature regulation and dynamic gating mechanism is introduced to eliminate quality difference and semantic asymmetry between the modes and realize adaptive feature fusion; then, a gradient adjustment factor is obtained according to the confidence coefficient ratio, a direction consistency adjustment factor is generated according to cosine similarity, and in the process, gradient scores represented by the gradient adjustment factor obtained based on the confidence coefficient ratio are sensed; and the gradient direction represented by the direction consistency adjustment factor generated by the cosine similarity is aligned with two dimensions to adjust the modal gradient, so that the overall performance of multi-modal learning is finally improved.

The invention belongs to the technical field of image processing, and particularly relates to an aluminum bar quality detection method and system based on machine vision, and the method comprises the steps: obtaining a gray image of the surface of an aluminum bar; constructing a structure tensor for describing the distribution of the pixel points in the local gradient direction, and obtaining gradient coherence indexes of the pixel points according to feature values of the structure tensor; according to the gradient coherence index, the gray value of the pixel point and the gray value mean value and the gray value standard deviation in the neighborhood, obtaining coherence weighted saliency of the pixel point; according to the coherence weighted saliency and the coherence weighted saliency of all the pixel points in the pixel point neighborhood, acquiring a defect probability index of the pixel points; and performing aluminum bar surface quality discrimination according to the defect probability index. According to the method, through combination of gradient coherence and local brightness statistical characteristics, interference of specular reflection light spots on the surface of the aluminum bar and wiredrawing texture noise is effectively inhibited, and accuracy of detection of weak defects such as scratches is improved.

The invention relates to the technical field of image recognition, in particular to a shearing behavior dynamic correction method and system based on wear state recognition. Acquiring a digital image sequence of the cutting edge area; constructing an image feature separation network, and performing parallel feature extraction on the preprocessed digital image sequence; identifying a pixel-level high-frequency texture discontinuous region and an edge gradient direction field by utilizing continuity characteristics of a cutting edge surface periodic texture mode to obtain a target defect probability graph; identifying a projection shadow area and a low-frequency illumination halation of the edge of the bulge by using a backlighting imaging model to obtain an interference artifact probability graph; establishing spatial mutual exclusion constraints of the target defect probability graph and the interference artifact probability graph in a pixel space, and generating a defect binary mask; performing multi-dimensional texture feature calculation on an area corresponding to the defect binary mask, and constructing a surface state feature vector; according to the invention, based on the surface state feature vector, the defect mode category is discriminated, and the corresponding shearing correction parameter is generated.

The embodiment of the invention provides an edge detection method and device based on gradient weighted fusion and an adaptive threshold, and is applied to the field of computer vision and digital image processing. The method comprises the steps of firstly preprocessing an input image, then extracting two groups of gradient magnitude diagrams and directional diagrams through an adaptive morphological operator and a traditional difference operator, and constructing a weighted fusion function according to the gradient direction consistency of each pixel point to obtain a fused gradient magnitude diagram; and adaptively determining a high threshold and a low threshold based on an information entropy maximization principle, executing an improved Canny process by combining the fused gradient magnitude diagram and the second gradient directional diagram to obtain an initial edge diagram, and outputting a final edge detection result after dynamic structure element optimization. In this way, the defects that in a traditional edge detection method, gradient information extraction is not precise, threshold selection lacks adaptability, and an edge result is fractured can be overcome, more robust and more accurate image edge detection is achieved, and the reliability of an edge detection algorithm in a complex image scene is improved.

The invention belongs to the technical field of image processing, and particularly relates to a screw fastening quality evaluation method and system based on image features, and the method comprises the steps: collecting a screw fastening image, and obtaining a complex response diagram through a Log-Gabor filter; obtaining a phase congruency diagram according to each complex response diagram; acquiring gradient directions of pixel points in the phase consistency graph, and constructing an accumulator graph according to the gradient directions and offset points in a preset radius range; according to salient points in the accumulator graph and local signal-to-noise ratios of the salient points, screw positioning points are determined through two-dimensional quadratic polynomial function fitting; the torque value of the screw is obtained according to the screw positioning point driving torque measuring device, and whether the screw is fastened or not is judged. According to the method, the screw is positioned by using phase consistency in the frequency domain, so that the interference of specular reflection and extreme shadow on positioning is overcome, the accuracy of screw positioning is improved, and screw fastening quality evaluation is assisted.

The invention relates to the technical field of wiring robot system fault detection, in particular to a transformer wiring robot system fault intelligent detection method. The method comprises the following steps: acquiring an image in a wiring process and graying the image to obtain a grayscale image; obtaining candidate pixel points in the grayscale image and a pixel point sequence of the candidate pixel points; acquiring a local gradient amplitude change feature value and a local gradient direction change feature value of each neighbor pixel point, and further acquiring initial edge feature credibility of the candidate pixel points; obtaining the final edge feature credibility based on the initial edge feature credibility of each marked pixel point of one candidate pixel point; performing clustering analysis based on the final edge feature credibility of each candidate pixel point in a gray level image to obtain a low threshold value and a high threshold value of the gray level image; and carrying out edge detection on the grayscale image by using the low threshold and the high threshold, and identifying the abnormity of the wiring robot. According to the invention, the wiring abnormity of the wiring robot can be effectively processed.

The invention relates to the technical field of medical instrument detection, discloses a medical whole basket instrument multi-parameter detection system and method, and aims to solve the problem that in an existing CCD image detection technology, a high-brightness reflection area interferes with a detection result due to tiny bloodstain or superposition on the surface of an instrument. The system comprises an imaging module, a light source control module, a data processing module and a carrying platform. According to the method, reflection is reduced through combination of multispectral imaging and an adjustable light source, a reflection area is accurately identified and removed by using an algorithm of combining gradient amplitude and gradient direction entropy, and finally geometric parameters and cleanliness indexes of the instrument are synchronously calculated.

The invention relates to the technical field of image processing, and discloses a mechanical part contour extraction method and system based on edge detection. The method comprises the following steps: carrying out overlapping region division on a part image, implementing adaptive illumination compensation according to gray level statistics, carrying out gradient joint detection on a preprocessed image through adaptive double thresholds, screening reliable edge points according to neighborhood gradient direction deviation, continuously carrying out contour tracking connection according to a gradient direction, and carrying out contour tracking connection according to the gradient direction; and realizing inner and outer contour layered identification according to curvature statistical characteristics and topological nesting depth. The integrity of contour extraction of the mechanical part and the accuracy of layered recognition are improved.

The invention discloses a ship bollard identification method and system based on adaptive multi-scale multi-grid division, and belongs to the technical field of computer vision and image processing. The system divides an image foreground region and a background region through a visual saliency calculation module, carries out dense sampling in the foreground region and sparse sampling in the background region by adopting a non-uniform grid generation module, screens candidate grids in combination with gradient direction consistency and texture features, fuses candidate frames of different scales through a multi-scale image pyramid, and finally obtains a multi-scale image. And accurate positioning of the bollard is realized through the fine grid accurate positioning module. The method solves the problem that the prior art is insufficient in adaptability to ship size, shooting distance and resolution change, improves the precision and generalization ability of bollard positioning in a complex scene, and is suitable for bollard detection scenes of various ship images.

The invention relates to the technical field of product design, in particular to a household appliance appearance defect optimization product design method based on image discrimination, which comprises the following steps: acquiring a household appliance appearance image, calculating a gradient direction, constructing a distribution matrix, analyzing consistency to generate a defect edge image, extracting high-frequency feature scores to identify defect areas, and clustering point cloud to output defect clusters. According to the method, by calculating the gradient direction numerical value of each pixel point and constructing the gradient direction distribution matrix, the processing capacity of the directivity difference of the household appliance surface texture is effectively improved, the defect point cloud is accurately divided through the density-based noise application spatial clustering algorithm, the defect point cloud is accurately divided, and the defect point cloud is accurately divided. According to the method, the space positioning accuracy of the defect area is improved, the precision of optimization design is ensured, curved surface reconstruction is performed in combination with the least square method, the appearance design of the household appliance is accurately corrected, and the production efficiency and the design quality are optimized.

The invention relates to the technical field of image processing, in particular to a lens edge detection method and system based on image generation, and the method comprises the following steps: collecting a lens image through an industrial camera, carrying out the weighted graying, extracting an edge coordinate through a Laplace operator, constructing an edge pixel coordinate set, and carrying out the image processing; screening an edge validity marking result by combining a gray difference threshold, calculating a phase difference between a gradient direction and light source incidence based on a Sobel operator, extracting pixels in a consistent direction, carrying out spatial clustering, generating a lens edge pixel cluster, and fitting a continuous curve to construct a complete lens edge contour. Noise and reflection interference are restrained through pixel neighborhood gray difference, effective pixels are screened in combination with a gradient direction and light source incident phase relation, edge direction consistency is enhanced, pixels are clustered according to spatial continuity and gradient intensity, contours are continuously fitted based on pixel cluster distribution, and lens edge integrity and geometric consistency under complex illumination are improved.

The invention provides an unmanned vehicle local high-precision positioning mapping system and method based on reinforcement learning polarization normal deambiguity and depth information fusion, and the method comprises the steps: synchronously obtaining a polarization image and a depth image of a target scene in the driving process of an unmanned vehicle in a low-texture and low-illumination scene; constructing the polarization normal candidate, the depth normal, the neighborhood surface curvature and the depth gradient information into a state vector, inputting the state vector into a reinforcement learning agent based on Dueling DQN, and obtaining the depth gradient according to a reward function fusing the included angle error of the polarization normal and the depth normal, the neighborhood normal smoothness loss and the consistency error of the polarization normal and the depth gradient direction. Outputting an optimal polarization normal selection action to obtain an unambiguous surface normal; polarization estimation depths are generated, and confidence coefficients are calculated respectively; according to the depth confidence coefficient and the polarization confidence coefficient, pixel-level weighted fusion is carried out, a fused depth map is obtained, synchronous positioning and mapping are carried out on the fused depth map and the synchronous RGB image, and a point cloud map of the target scene and the moving track of the unmanned vehicle are output.

The invention provides a semiconductor device part surface micro defect image preprocessing data processing method and system, and relates to the technical field of data processing, and the method comprises the steps: carrying out the image quality normalization processing of an obtained high-magnification amplification surface original image, and generating a quality enhancement image; performing defect feature selective enhancement processing on the quality enhanced image to generate a feature optimized image; performing three-dimensional shape reconstruction on the feature optimization image to generate surface three-dimensional shape topological data containing height information; on the basis of the surface three-dimensional topography topological data, a topography reference origin is determined at the highest point of a local topography protrusion of the preliminary defect candidate area, and a first feature radiation vector and a second feature radiation vector are generated in a fitting mode from the topography reference origin in the main direction and the normal change gradient direction of the surface texture respectively; according to the invention, precise characterization of micro defects in a high-magnification magnification scene can be realized, and the accuracy and real-time performance of defect detection are guaranteed.

The invention relates to the technical field of endoscopic image analysis, in particular to a cystoscope tumor focus real-time positioning and boundary intelligent recognition system and method, and the system comprises a feature extraction module, a breakpoint detection module, a contour closing module, a boundary evaluation module and a smooth display module. According to the method, edge features are accurately captured by calculating pixel gradient vector data, homology matching is performed on end points of a fracture contour according to a gradient direction trend, connection pixels are automatically interpolated and filled between the breakpoints along a tangential direction, and non-closed edge gaps caused by illumination or shielding are repaired to form a complete closed contour. A boundary uncertainty evaluation model is constructed by combining gray uniformity and gradient intensity in a local neighborhood, a smoothing processing intensity radius is adaptively adjusted according to the boundary uncertainty evaluation model, depth smoothing is applied to a high-risk area, high-confidence area details are reserved, zigzag artifact interference is eliminated, and a high-risk area is obtained. And a continuous and high-precision self-adaptive focus boundary conforming to the real anatomical form of the tissue is generated.

The invention discloses a polyester staple fiber spinning on-line detection system and method based on machine vision. The system comprises a pneumatic optical constraint assembly, an imaging assembly and a processing unit. The pneumatic optical constraint assembly is integrated with a Venturi flow channel and a back plate with an inclined stripe background; the imaging assembly obtains image data of the tows constrained by the venturi flow channel. The processing unit calculates a vertical gradient component and a gradient direction, and generates a first feature map marking a defect area by calculating a difference value between the gradient direction and a stripe background extension direction; and weighting the vertical gradient component by using the first feature map, executing projection integral operation in a direction vertical to the motion direction of the tow to offset the gradient component of the stripe background, and calculating and outputting tow width data according to a projection curve generated by integral. According to the method, transparent defect detection and tow width measurement are simultaneously realized in a single-frame image through cooperation of fluid dynamic constraint and an integral cancellation algorithm.

The invention relates to the technical field of image data processing, in particular to a chip surface defect area judgment method in combination with depth map abnormal point aggregation degree, which comprises the following steps: S1, acquiring a chip surface depth map, extracting abnormal points based on local fitting residual errors and constructing a structure evolution graph, the abnormal point description module is used for describing a multi-scale evolution trend of abnormal points from point shapes to line shapes or sheet shapes; s2, a linear evolution region is extracted, a global direction tensor field is constructed, direction and space association is carried out, and a collaborative defect channel graph is generated; and S3, matching the channel graph with the gradient direction consistency field, screening regions which are consistent in direction and continuous in connection, and outputting a defect region labeling graph. According to the method, through three-stage linkage of multi-scale evolution modeling, direction tensor collaborative analysis and gradient consistency verification, a real chip surface defect area with directivity and structural continuity is accurately identified, and the detection sensitivity and the judgment reliability are remarkably improved.

The invention discloses an automatic driving test scene generation method and system based on a hybrid expert model, and belongs to the technical field of automatic driving test verification and artificial intelligence cross, and the method comprises the following steps: obtaining a scene description file through the hybrid expert model; generating a first scene according to the scene description file, and collecting first scene data; and obtaining a sensitive area of the first scene according to the entropy of the first scene data and the gradient direction thereof. A scene description file is generated based on the hybrid expert model, and the scene generation efficiency is improved; by identifying the sensitive area of the scene, the scene is iterated based on the sensitive area, and a challenging scene is generated; closed-loop optimization is carried out on the hybrid expert model through test feedback and a reinforcement learning reward function, directional iteration is carried out based on the gradient direction of scene decision uncertainty, an automatic and directional iterative'generation-test-evaluation-iteration-optimization 'intelligent closed loop is formed, and the efficiency, coverage and pertinence of scene generation are remarkably improved.

The invention provides a CT metal artifact correction algorithm based on combination of double-domain diffusion and wavelet attention, and relates to the technical field of image correction. According to the method, the accuracy and effectiveness of CT metal artifact correction are improved by fusing the wavelet attention mechanism and the double-domain diffusion model, and the accuracy and effectiveness of CT metal artifact correction are improved by means of targeted extraction of the wavelet attention module on the high-frequency component of the image and quantitative analysis of the gradient direction consistency index and the curvature entropy. Accurate distinguishing of real edges and metal artifact fragments is achieved, and the problem of structure loss caused by confusion of edges and artifacts in a traditional method is effectively solved. Besides, on the basis of the design of edge geometric attribute dynamic distribution diffusion parameters, a small convolution kernel and a slow step length are adopted for a high-curvature edge to reserve a fine structure, and a large convolution kernel and a fast step length are adopted for an artifact area to strengthen the suppression effect, so that the artifact removal efficiency is improved, and the damage of excessive smoothness to key edge features is avoided; and the balance between the local fine structure and the global smooth demand is realized.

The invention discloses a crack sub-pixel precision edge detection method based on an improved Canny operator. The crack sub-pixel precision edge detection method comprises the steps of obtaining a crack image corresponding to a structural body surface crack; the crack image is preprocessed; carrying out gradient calculation on the preprocessed crack image through an improved Sobel operator to obtain a gradient image containing a gradient magnitude and a gradient direction; performing non-maximum suppression on the gradient image, adaptively obtaining an optimal threshold by adopting an improved maximum between-class variance method, judging a crack edge according to the optimal threshold, and further obtaining a coarse edge pixel-level image; processing edge information in the coarse edge pixel-level image by using a Zernike orthogonal moment to obtain coordinates of sub-pixel edge points; and connecting all the sub-pixel edge points, and outputting a crack detection result image. On the basis, the crack edge detection precision is improved to a sub-pixel level.

The invention relates to the technical field of image data processing, in particular to a flange forge piece surface defect detection method and system based on image processing, and the method comprises the steps: obtaining an original image of a flange forge piece, and carrying out the differential Gaussian band-pass filtering to obtain a preprocessed image; performing multi-scale structure tensor analysis on the preprocessed image, and determining a texture abnormal value according to a gradient magnitude and an included angle between a gradient direction and a local texture direction; extracting a background image through morphological reconstruction, and calculating a geometric suppression weight based on brightness difference; and fusing the texture abnormal value and the geometric suppression weight to obtain a defect response value, and identifying the surface defect through threshold segmentation. According to the method, through combination of multi-scale texture analysis and a geometric suppression mechanism, weak defects can be accurately extracted under a strong texture background, artifacts generated by a flange geometric structure are effectively suppressed, and the accuracy and robustness of defect detection are improved.

The invention belongs to the technical field of image processing, and particularly relates to a pole adaptive positioning method and system based on image processing, and the method comprises the steps: obtaining edge pixel points in a pole gray level image and the gradient direction of the edge pixel points; estimating an illumination distortion main direction according to the gradient direction, and dividing edge pixel points into a positive edge set and a negative edge set; according to the forward edge set, calculating weighted votes of edge pixel points to gradient direction intersection points and accumulating the weighted votes to a forward accumulator, clustering accumulation results to obtain an initial forward candidate center, and finally performing iterative updating through corrected voting weights based on geometric consistency to obtain forward center estimation; obtaining negative center estimation according to the negative edge set; and determining a circle center positioning point of the pole hole by combining the positive center estimation and the negative center estimation. The method effectively overcomes the positioning error caused by uneven illumination and metal reflection, and improves the positioning precision and robustness.

The invention relates to the technical field of image classification, in particular to a geographic surveying and mapping image classification processing system and method based on big data, and the system comprises a regional feature module, a fluctuation analysis module, a spectrum anomaly module, an edge adjustment module and a classification correction module. According to the method, the gray brightness and the multi-band reflectivity are extracted through region division, variance contrast and spectral stability analysis are combined, dimensionality reduction and redundancy reduction are performed through principal component analysis, key features are reserved, constraint intensity is counted and tested through cooperation of interval division and fluctuation frequency extraction, potential anomalies are recognized in advance, and noise interference is reduced. Reflectivity difference and spectrum extreme value detection are combined to capture abnormal changes through a stability threshold value, gradient direction deviation and an edge closing proportion monitor boundary continuity and are complementary with spectrum detection, the polymerization degree and the sparse ratio are calculated, the priority is dynamically corrected, deviation caused by a single feature is avoided, and the classification result is more accurate and stable.

The invention relates to the technical field of image processing, in particular to a method and system for detecting surface defects of an insulating layer of a polyethylene insulated cable, and the method comprises the steps: obtaining a surface image of the insulating layer; the defect response index of each pixel in the surface image is calculated, and the defect response index is in positive correlation with the product of the gradient amplitude of each pixel in the corresponding local window and is in positive correlation with the absolute value of the difference value of the gray average values of all the pixels in the local window. According to the method, a defect response index is constructed by introducing a gradient direction angle variance, the defect response index and the defect response index are distinguished by utilizing the physical characteristics that the gradient direction of a highlight region is consistent and a defect region is disordered, and a defect significance enhancement factor is constructed according to the physical characteristics to dynamically adjust the Gaussian kernel scale. According to the method, large-kernel smooth interference is automatically adopted in a highlight area, small-kernel reserved details are adopted in a defect area, the limitation of a traditional fixed parameter algorithm is effectively overcome, and the defect detection precision under the complex reflective background is remarkably improved.

The invention belongs to the technical field of building boundary segmentation, and provides a remote sensing image building boundary segmentation method based on a U-net network model, and the method comprises the steps: extracting a gradient magnitude image through a Sobel operator, generating a binary boundary mask, dynamically adjusting the weight of a convolution kernel in combination with a gradient direction, and enhancing the feature extraction capability of a boundary region; boundary artifacts output by cavity convolution are suppressed through a high-frequency residual boundary mask, an artifact area is restored in combination with gradient direction constraint, and original detail features are reserved; calculating complexity scores based on boundary density and tortuosity, adaptively selecting voidage, and balancing global context and local detail capture capability; extracting a high-frequency component through a Laplace operator, marking an artifact point by combining a dynamic threshold value, and repairing an abnormal region by utilizing a neighborhood gradient direction; the method solves the performance bottleneck of a traditional method in a boundary fuzzy, artifact interference and receptive field fixed scene, and is suitable for remote sensing image analysis tasks such as urban planning and disaster assessment.

The invention relates to the technical field of image data processing, in particular to an impurity identification method and system for fruit and vegetable powder, and the method comprises the steps: collecting a fruit and vegetable powder image on a conveying belt, preprocessing the image into a gray-scale map, and calculating a dark channel index based on the gray-scale distribution characteristics in a neighborhood window to preliminarily screen a dark region; counting dispersion of gradient amplitude and information entropy of gradient direction in a pixel point neighborhood window, distinguishing smooth shadow and rough impurities by using the dispersion and the information entropy, and correcting a dark channel index to obtain impurity confidence; constructing a local histogram by using the impurity confidence as a weight, and carrying out adaptive weighted histogram equalization processing on the grey-scale map to enhance the impurity contrast and suppress the shadow; and completing impurity identification through Otsu threshold segmentation and connected domain analysis. According to the method, the problem that shadow is easily misjudged as impurities when accumulated powder is processed by a traditional algorithm is effectively solved, and the detection accuracy is improved.

The invention relates to the technical field of image data processing and deep learning, in particular to a blind sidewalk image segmentation method based on a self-attention mechanism, and the method comprises the steps: obtaining a blind sidewalk image, and calculating a weak edge tension index of each pixel point; the method comprises the following steps: dividing a blind sidewalk image into grids and local windows through a Swindow-Transform network, calculating a cross-grid splitting flux between adjacent grids in combination with a weak edge tension index and an initial gradient direction of pixel points on a shared boundary line of the adjacent grids, calculating a boundary perception attention compensation component based on the cross-grid splitting flux, and calculating the blind sidewalk image according to the boundary perception attention compensation component. The method comprises the following steps: acquiring a blind sidewalk image, integrating the blind sidewalk image into a boundary perception attention compensation offset matrix of a local window, taking the boundary perception attention compensation offset matrix of the local window as an intervention item, injecting the intervention item into a self-attention calculation module of a Swindow-Transform network, and carrying out semantic segmentation on the blind sidewalk image to obtain a segmentation result. The method improves the accuracy of blind sidewalk image segmentation.

The invention relates to the technical field of defect detection, in particular to an electronic component welding spot detection method and system, and the method comprises the following steps: based on a circuit board imaging image, extracting a welding spot gray scale gradient direction and a pixel change trend, marking a gray scale path, recognizing a pixel space communication relation, and analyzing the welding spot gray scale consistency and edge stability. According to the method, the gray gradient direction and neighborhood pixel intensity change of the welding spot region are extracted, so that the local structural features of the welding spot are effectively recognized, the gray consistency and edge stability of the welding spot are optimized, and the welding spot defect classification and partition recognition label set is obtained. The morphological operation processes isolated pixels and cavity pixels, background noise and fragments are effectively removed, accurate closing of welding spot contours is ensured, the defect classification precision is improved, the defect type recognition capability is enhanced, false detection is reduced, and the precision and efficiency of defect classification and region recognition are optimized.

The invention relates to the technical field of cable detection, in particular to a method and system for detecting the thickness of an insulating layer of a cross-linked insulated cable. The method comprises the following steps: acquiring a cable cross section image, performing region segmentation to identify three topological connected domains, and extracting inner and outer boundaries; establishing a physical field model and applying a first type of boundary conditions; solving a Laplacian equation in the insulating layer region to generate a convergent scalar potential field; performing differential operation on the scalar potential field to obtain a gradient vector field, and searching gradient modulus length extreme points on the virtual equipotential line to mark thinnest and thickest feature points; and taking the thinnest feature point and the thickest feature point as starting points, respectively executing streamline tracking along positive and negative gradient directions, and calculating the sum of path lengths so as to respectively obtain the minimum thickness of the insulating layer and the maximum thickness of the insulating layer. Noise is eliminated by solving the Laplacian equation, rapid positioning is performed by using the gradient modulus length extreme value, and the detection precision is remarkably improved.

The invention discloses an image edge matching degree calculation and inspection point position deviation correction control method and system, and innovatively adopts a semantic segmentation and multi-level feature fusion mechanism to solve the problems of low precision and poor robustness of traditional IMU coarse adjustment, single template matching and SLAM algorithms. The method comprises the following steps of: acquiring an image through a pan-tilt camera, performing de-noising processing, and calculating an initial translation offset by using an RANSAC algorithm; generating a mask matrix in combination with the pre-training model, and extracting the edge gradient direction and strength; region pairing is established based on semantic consistency, and the matching degree is calculated through Fourier descriptor conversion; the control quantity is reversely deduced through the pixel-holder coordinate mapping relation, image acquisition is executed recursively until the matching degree reaches a threshold value, and semantic-level high-precision automatic correction and intelligent control of the target area are achieved.