47 results about "Gradient magnitude" patented technology

The invention belongs to the field of cable eccentricity detection, and particularly relates to a cable eccentricity detection method and system based on X rays. The method comprises the following steps: calculating a gradient magnitude diagram and a gradient direction diagram through an X-ray image of a cable, screening a point with the local maximum gradient magnitude and low neighborhood divergence as a contour starting point, performing contour tracking to generate a contour point set, and selecting a next contour point based on a tangential prediction direction; performing ellipse fitting on the contour point set to obtain a geometric center, long and short axis parameters and a root-mean-square error of a fitting ellipse, dividing the fitting ellipse into an inner candidate ellipse and an outer candidate ellipse according to a long axis, and matching the inner candidate ellipse meeting the condition for each outer candidate ellipse; and screening out an outer candidate ellipse and an inner candidate ellipse which meet conditions from the candidate pairs, and calculating the eccentricity of the cable to be measured based on geometric center coordinates of the outer candidate ellipse and the inner candidate ellipse. According to the invention, the accuracy and reliability of cable eccentricity measurement results can be improved.

The invention discloses a mold defect visual inspection method, particularly relates to the technical field of industrial machine visual inspection, and is used for solving the technical problem of image spatial variation blurring caused by mechanical vibration under a mobile shooting condition. The method comprises the following steps: acquiring a to-be-detected image on the surface of a mold, analyzing the gradient magnitude of each region, determining the fuzzy characteristics of different regions in the to-be-detected image according to the difference of the gradient magnitudes, evaluating the expected confidence of each region for executing the deblurring operation based on the fuzzy characteristics, and executing the deblurring operation on the regions to obtain a preliminary restored image; an artifact index is calculated in the uniform background area of the preliminary restored image, the distribution concentration degree of image components in each local feature area in the frequency domain is analyzed in the preliminary restored image, and the distribution concentration degree is compared with a preset defect judgment threshold value adjusted according to the artifact index; judging whether the corresponding local feature region is a defect region or not; accurate recognition of mold surface defects under complex imaging conditions is realized.

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 relates to the technical field of image segmentation, in particular to an anorectal focus automatic segmentation method based on deep learning, which comprises the following steps: acquiring an anorectal image pixel map, extracting contrast and direction offset to mark candidate focus points, screening overlapped marks to generate a focus activation mark map, and establishing a response map to generate a boundary response distribution map. And training the network to output a classification graph, and extracting a truncation path to complete image segmentation. According to the invention, through extracting the contrast value and the gradient amplitude of the local gray level co-occurrence matrix, accurate capturing of the spatial difference of the lesion area under a complex background is realized, through constructing a response map and direction consistency comparison mechanism and combining multi-dimensional features such as a direction gradient histogram and a structure tensor, the area discrimination capability and the edge classification precision are improved, and the accuracy of edge classification is improved. The texture stability is judged by means of anisotropic standard deviation, a fuzzy edge mask is set, truncation paths are screened in combination with a main direction vector included angle deviation trend, and continuity and stability of a boundary convergence position are ensured.

The invention discloses a feature point detection method based on edge saliency and scale sensitivity. The method comprises the following steps of performing Gaussian smooth denoising on an input image; calculating the edge saliency of the pixel points through a Laplace operator; calculating a texture change degree based on the local gradient magnitude; generating a scale sensitivity weighted value through multi-scale analysis; candidate points are screened in combination with edge saliency, texture weighting and scale weighted values, and mismatching points are removed through non-maximum suppression; and finally, key points are enhanced and marked. According to the method, the edge structure, the texture information and the multi-scale features are fused, the robustness and accuracy of feature point detection are improved, and the method is particularly suitable for image matching and target recognition tasks in complex scenes.

The invention discloses a network optimization InSAR large gradient deformation phase unwrapping method based on edge detection, and the method specifically comprises the steps: 1, carrying out the differential interference processing of an original image, and obtaining a differential interference pattern; 2, Gaussian filtering is carried out on the differential interferogram, and the gradient magnitude and the gradient direction of each pixel in the image after Gaussian filtering are calculated; a gradient magnitude image is obtained; step 3, performing non-maximum suppression on the gradient magnitude image, and completing preliminary extraction of edges; 4, setting double thresholds to further screen edge points, and finally completing edge extraction; 5, generating a coherence graph based on the differential interferogram, extracting coherence points, and generating an unwrapping network based on the extracted coherence points and the edges extracted in the step 4; and step 6, phase unwrapping is carried out based on the unwrapping network. The method has certain universality for monitoring application scenes with large-gradient characteristic deformation, such as landslide, mining and the like, and can provide important technical support for geological disaster prevention and control.

The invention discloses an enhanced straight line detection method and system based on gradient pooling. The method comprises the following steps: defining a detection area in an image and dividing the detection area into a plurality of sampling calipers; carrying out average pooling processing on the data to generate a noise-reduced and dimension-reduced low-dimensional feature map; on the low-dimensional feature map, coarse positioning edge points are screened out based on double constraints of gradient amplitude and direction consistency; then, returning to the original pixel gray data, and carrying out sub-pixel-level interpolation positioning along the normal direction of the gradient direction of the coarse positioning edge point to obtain a high-precision fine positioning edge point; and finally, after outliers are eliminated by adopting a random sampling consistency algorithm, fitting is carried out on inner points through a least square method, and a final straight line parameter is obtained.According to the method, through the synergistic effect of pooling noise reduction, gradient constraint screening, sub-pixel fine positioning and robust fitting, the contradiction in precision, efficiency and noise resistance in the prior art is effectively solved, and the accuracy, efficiency and noise resistance of the system are improved. And the comprehensive performance of straight line detection is obviously improved.

The invention discloses a gradient-guided panchromatic sharpening method for unlocking spatial textures, which comprises two image processing branches: a guide feature extraction branch for capturing gradient features from gradient amplitudes and generating auxiliary gradient features grads and refined guide gradient features Gsr, Gsr, Gsr, Gsr, Gsr, Gsr, Gsr, Gsr, Gsr, Gsr, Gsr, Gsr, Gsr, Gsr, Gsr, Gsr, Gsr and Gsr; gradient features extracted from gradient amplitudes can provide complementary information to promote optimization and improve the overall performance; one branch is a gradient-guided fusion branch, the other branch is a gradient-guided fusion branch, PAN and MSI are combined with extra information of gradient features, and in the fusion branch, a multi-image cross attention module is designed to gradually integrate image features with different spectral bands and resolutions, so that the PAN and MSI fusion process is effectively guided, and the panchromatic sharpening performance is remarkably improved.

This application discloses a method, apparatus, device, and storage medium for edge detection of images, relating to the field of image edge detection technology. The method includes: acquiring an input image and performing structure-preserving filtering on the input image to obtain a smoothed image; calculating a first-order gradient magnitude map based on the smoothed image; generating an adaptive parameter mapping based on the statistical characteristics of the first-order gradient magnitude map; obtaining an edge magnitude map based on the adaptive parameter mapping; determining a global threshold based on the edge magnitude map, and setting a high threshold and a low threshold based on the global threshold; and performing double-threshold hysteresis connection processing on the edge magnitude map based on the high threshold and the low threshold to output a binary edge map. This method can improve the detection accuracy of images.

The invention belongs to the technical field of image data processing, and particularly relates to a brake pad surface damage detection method and system based on machine vision, and the method comprises the steps: obtaining an image of the surface of a brake pad, and calculating the gradient vector of each pixel point; constructing a gradient interaction field based on the intensity of the gradient magnitude and the gradient direction angle, and obtaining the structural significance of each pixel point; path searching is carried out in the local main direction, and linear propagation potential energy of each pixel point is obtained by accumulating structural significance and direction consistency weights of the pixel points on the path; and adaptive threshold segmentation and morphological processing are carried out based on the linear propagation potential energy diagram, and accurate detection of the surface damage area of the brake pad is realized. According to the invention, the technical problems of strong noise interference on the surface of the brake pad and difficulty in detection of discontinuous micro crack characteristics are solved, and the robustness and accuracy of damage detection are improved.

The invention provides a method, device and equipment for determining checkerboard angular points, and the method comprises the steps: carrying out the convolution of an initial image, and determining a gradient size image and a gradient direction image; since the angular points are located at the intersection of the background and the foreground, the pixel points at the intersection are initially screened according to the gradient size image, and then the screened pixel points are classified according to the gradient direction image, so that different types of checkerboard edge contours (in different directions) can be obtained; secondly, determining an intersection point of the edge contours of the intersected checkerboard, roughly determining an initial angular point at the moment, determining straight lines passing through the initial angular point in the horizontal direction and the vertical direction in order to further improve the detection precision of the angular point, solving the two straight lines to determine an intersection point of the two straight lines, and taking the intersection point as a target intersection point; in this way, the final angular point can be accurately determined by fitting the first straight line and the second straight line and then solving, and the subsequent calibration precision of the module is improved.

The invention relates to the technical field of tunnel image data processing, in particular to a tunnel smoke image recognition method and system based on multi-scale feature fusion. The method comprises the following steps: obtaining the spatial weight of each pixel point, wherein the spatial weight is in negative correlation with the distance from the pixel point to a vanishing point; obtaining an enhanced gradient of each pixel point, wherein the enhanced gradient is in positive correlation with the closeness degree between the second characteristic value and the first characteristic value of the pixel point, the first-order gradient module value and the spatial weight; taking the enhanced gradient greater than a gradient threshold as a seed point to perform connected domain analysis to obtain a candidate cluster in the tunnel image; according to the method, the extinction coefficient of the candidate cluster where the seed point is located is obtained, the pixel point corresponding to the candidate cluster is obtained from the historical image sequence of the current tunnel image, time domain accumulation is carried out on the enhancement gradient and the extinction coefficient of the pixel point, the smoke judgment score of the current tunnel image is obtained, tunnel smoke is recognized, and the accuracy of the recognition result is effectively improved.

This invention provides an industrial image anomaly detection method based on prior knowledge and image reconstruction, comprising the following steps: Step 10) Inputting the image to be tested into an anomaly judgment model based on multivariate Gaussian distribution to obtain a mask of the image to be tested, and using the mask to occlude the image to be tested, obtaining an occluded image; Step 20) Inputting the occluded image into an image reconstruction model for reconstruction, obtaining a reconstructed image; Step 30) Calculating the multi-scale gradient magnitude similarity map between the image to be tested and the reconstructed image to obtain the anomaly detection result. This invention provides an industrial image anomaly detection method based on prior knowledge and image reconstruction, which makes the image reconstruction process unaffected by abnormal pixels, improving detection efficiency and accuracy.

An inventive method for testing the feasibility of acquiring diffusion-weighted measurement data of a test object using a magnetic resonance system with a measurement protocol with non-trapezoidal gradient pulse shapes for diffusion coding comprises the steps a) Loading a non-trapezoidal gradient pulse shape, b) Loading pre-prepared characteristics for the loaded non-trapezoidal gradient pulse shape, c) Receiving a condition that must be met during the execution of the acquisition of the diffusion-weighted measurement data, d) Determine at least one gradient amplitude relevant for the gradient pulse shape based on the charged characteristics and the condition, e) Checking the feasibility of acquiring diffusion-weighted measurement data of the object under investigation using the magnetic resonance system with a measurement protocol with the non-trapezoidal gradient pulse shape based on at least one specific relevant gradient amplitude.

The invention discloses a laser center line extraction method based on improved non-maximum suppression, and the method comprises the steps: (1) preprocessing a laser image, and carrying out the graying of the image; (2) carrying out noise filtering on the image by using a Gamma algorithm, deleting too dark or excessive noise in the image, and reserving a proper pixel point set; and (3) gradient acquisition by a Sobel operator: determining the position and direction of the edge by calculating the gradient value of each pixel point in the image through a Sobel edge detection algorithm. The gradient size (Gx) and the gradient direction (Gy) in the laser image are calculated by using a Sobel operator. And (4) suppressing the pixel points according to the size and direction of the gradient: finding out a local maximum point on the same gradient by using a non-maximum suppression algorithm, and extracting a center line. And (5) the straight line obtained by the method of interpolation only in the Y direction is smoother and more continuous. According to the method, the problems of overlarge exposure and excessive noise are effectively solved, the center line of the laser image is accurately extracted, and the method is of great significance to extraction of the center line.

The invention discloses a semi-supervised medical image segmentation method and system, and relates to the technical field of image segmentation, and the method comprises the steps: calculating the spatial average weighted sum of the three-dimensional gradient magnitude and a discrete three-dimensional Laplacian operator through a foreground probability graph outputted by a student network, and obtaining a global boundary roughness scalar; the voxel-level entropy and the soft boundary indicator are combined to construct the consistency loss of roughness perception, and adaptive course learning is realized; carrying out residual modulation on geometric contrast loss by driving a scaling factor, and enhancing the structural discrimination force of a boundary region; boundary gradient alignment, curvature smoothing and a pseudo-supervision threshold are guided, and the student network and the teacher network are optimized through weighted joint of the three so as to perform index moving average updating. According to the method, the global boundary roughness scalar is introduced, entropy consistency, geometric boundary comparison and gradient-curvature-pseudo supervision joint regularization are cooperatively regulated, and the boundary precision and robustness of medical image segmentation are remarkably improved under the semi-supervised condition.

The invention discloses a light area contour extraction method and device, equipment and a storage medium. The method comprises the following steps: carrying out light area extraction on a high-exposure grayscale picture based on a gradient magnitude to obtain a light response diagram; performing light area boundary expansion on the light response diagram to obtain a boundary expansion diagram; a bright light extraction threshold value is constructed according to the boundary expansion graph, a bright light mask graph is constructed according to the bright light extraction threshold value, and the bright light mask graph comprises effective boundary points; and carrying out connected region filtering and closed operation on the bright light mask pattern to obtain a bright light region contour. A light response diagram and a self-adaptive extraction mechanism are constructed through gradient amplitudes, a real set structure contour of a highlight area can be recovered under the condition of not depending on edge closure and gradient continuity, and the method has higher structure adaptability and interference robustness and is suitable for large-scale popularization and application. Therefore, structural highlight area contour information is stably extracted from a complex industrial image with multiple exposure interferences, uneven illumination and bright area overexposure.

This invention relates to the field of image recognition technology, specifically to an image recognition method and system for precision weld formation quality, comprising the following steps: S1: acquiring an original grayscale image containing the weld region; S2: calculating an adaptive segmentation threshold and extracting the weld target region accordingly; S3: performing Sobel convolution on the weld target region to generate a gradient magnitude image; S4: fusing the convolution results at various scales to generate an enhanced gradient image; S5: extracting closed contours from the enhanced gradient image and calculating the Euler number, number of holes, and rate of change of boundary curvature for each contour; S6: outputting the judgment result of weld formation quality defects based on combined features. This invention, through the collaborative design of synchronous triggering of welding current, high-order gradient enhancement, and topological feature judgment, achieves high-precision automation of weld image acquisition and quality defect recognition.

The invention relates to the technical field of temperature monitoring, in particular to a temperature monitoring method and system for a reaction kettle based on machine learning. The method comprises the following steps: acquiring infrared thermal images of the outer wall of the reaction kettle at multiple moments; calculating the weighted gradient magnitude of the image at the current moment, and inhibiting the gradient of the background region based on the entropy value of the image at the previous moment to obtain a corrected gradient map; determining a threshold value for screening the key pixel points based on historical image changes; extracting global gradient features from the corrected gradient map, and screening key pixel points to extract local focus features; splicing the global and local features into a fusion feature vector; and inputting the fusion feature vectors at the current moment and the last moment into a machine learning model to obtain the current temperature state of the reaction kettle. According to the scheme of the invention, noise interference can be suppressed, multi-dimensional features are fused, judgment is carried out by using a temperature change trend, and the monitoring reliability is improved.

This invention provides a method for generating a seabed DEM based on large-scale nautical chart depth points and high-resolution remote sensing surfaces, belonging to the field of marine surveying technology. The method includes: acquiring discrete depth points from remote sensing images and nautical charts; extracting a relative depth trend surface based on a dual-band logarithmic ratio model; calculating the image gradient field and pixel-level structure tensor, solving for the principal eigenvectors, and constructing an anisotropic diffusion tensor by combining local gradient magnitudes; using the nautical chart depth points to perform regression calibration on the trend surface to generate sparse residuals, and establishing a residual calculation grid; using the anisotropic diffusion tensor as a guiding constraint, performing non-uniform diffusion calculation on the sparse residuals to generate a global continuous residual correction surface; superimposing the correction surface onto the trend surface, and obtaining the final seabed DEM through anomaly removal and edge smoothing. This invention solves the problems of low accuracy and loss of detail in complex shallow sea topography inversion.

The invention provides a low-dose CT (Computed Tomography) image denoising method fusing gradient extraction and mixed loss. Comprising the following steps: S1, acquiring a low-dose CT image as input, performing numerical truncation and normalization preprocessing on input data, and constructing a de-noising model based on a U-Net backbone network; s2, constructing a gradient extraction module at the output end of the U-Net backbone network in parallel, extracting high-frequency edge features of the image in horizontal and vertical directions by using a Sobel operator, and mapping the image from an intensity domain to a gradient magnitude domain; s3, constructing a mixed loss function including pixel consistency loss, structural similarity loss and gradient perception loss; and S4, calculating the difference between the predicted image and the gold standard image by using the mixed loss function, and optimizing network parameters through back propagation to obtain a denoised CT image. According to the method, competition conflicts between mean square errors and structure indexes can be effectively relieved, and edge artifacts of a high-density skeleton region are eliminated while noise is suppressed.

A visible-light SAR image registration algorithm based on OS-SIFT computes the consistent gradient of the SAR image, the gradient magnitude image of the visible image, and the consistent gradient for each visible-light SAR image pair. Two Harris scale spaces are then constructed, and local maxima are searched in each Harris scale space to detect repeatable keypoints. Gradient position and orientation histogram descriptors are extracted from multiple image blocks to improve image saliency. Keypoint pairs are obtained using the NNDR method and further filtered using FSC. More accurate keypoint pairs are then filtered using translation, scale, and orientation constraints. FSC is then used again to remove outliers and identify the correct matching pairs. Finally, the transformation parameters between the SAR and visible images are calculated to register the visible-light SAR image pairs.

This invention relates to the field of tunnel image data processing technology, and particularly to a method and system for tunnel smoke image recognition based on multi-scale feature fusion. The method includes the following steps: obtaining the spatial weight of each pixel, where the spatial weight is negatively correlated with the distance from the pixel to the vanishing point; obtaining the enhancement gradient of each pixel, where the enhancement gradient is positively correlated with the proximity between the second and first feature values ​​of the pixel, the first-order gradient magnitude, and the spatial weight; using enhancement gradients greater than a gradient threshold as seed points for connected component analysis to obtain candidate clusters in the tunnel image; obtaining the extinction coefficient of the candidate cluster where the seed point is located; obtaining the pixel corresponding to the candidate cluster in the historical image sequence of the current tunnel image; and performing temporal accumulation on the enhancement gradient and extinction coefficient of the pixel to obtain the smoke determination score of the current tunnel image, thereby identifying tunnel smoke and effectively improving the accuracy of the recognition results.

This invention relates to the field of industrial vision technology, specifically to a deep learning-based method for detecting bubble defects in rubber bonding. The method includes the following steps: acquiring a grayscale image of the rubber bonding surface; calculating the gradient magnitude of the image pixels; obtaining a gradient magnitude distribution map; performing iterative calculations on the gradient magnitude distribution map; smoothing background noise on the rubber surface while simultaneously sharpening the edges; and generating an anisotropic diffusion-enhanced image. In this invention, a set of gradient vectors and position vectors is constructed based on candidate regions. The overall directionality of the gradient is quantified by calculating the centripetal convergence index of the vector field. The unique optical morphological characteristics of bubbles are used to distinguish true defects from surface stains or planar texture differences at the physical property level, eliminating false detections caused by texture interference. This achieves high-precision bubble defect localization and morphological confirmation without requiring extensive pixel-level annotation.

A method and system for image recognition of stripe spacing in liquid crystal textures, relating to the field of liquid crystal texture experimental data processing technology, includes: segmenting and grayscale processing a liquid crystal texture image to obtain multiple sub-images; calculating the gradient of each sub-image to obtain image gradient information; transforming the gradient direction matrix by treating gradient directions differing by 180° as the same direction to obtain a transformed direction matrix; statistically analyzing the gradient directions corresponding to gradients greater than a first multiple of the maximum gradient value in the gradient magnitude matrix in the transformed direction matrix to obtain an directional gradient histogram; and, when determining that stripes in a sub-image have a main direction, dividing multiple horizontal sampling lines and extracting their grayscale data signals to calculate the corresponding stripe spacing. This application only statistically analyzes the gradient directions corresponding to gradients greater than a first multiple of the maximum gradient value in the gradient magnitude matrix to obtain the directional gradient histogram, which can filter noise, eliminate brightness unevenness, and improve the accuracy of image recognition results.

This invention belongs to the field of image processing technology, specifically relating to a method and system for detecting the wear state of rotary drilling bits based on image recognition. The method includes: acquiring scalar brightness field and global potential flow gradient magnitude field data of the drill bit surface image; calculating the local potential well smoothing index and potential flow divergence coefficient based on the scalar brightness and potential flow gradient magnitude within a local window of each pixel; performing a top-hat transformation on the scalar brightness field to obtain the entity protrusion weight of each pixel, and weighting it with the local potential well smoothing index and potential flow divergence coefficient to obtain the degradation retention weight; mapping all pixels to graph network nodes, calculating the walk transition probability of a node moving to any of its neighboring nodes to construct a global graph transition probability matrix; performing PageRank iteration based on the transition probability matrix until convergence to obtain the retention probability of each pixel, thereby determining the wear state of the drill bit. This invention effectively suppresses mud interference and improves detection accuracy.

This invention discloses a computer vision-based method and system for analyzing surface defects in electrical porcelain blanks. The method includes: comparing each image in a historical image sequence with a standard template image; determining the interference region image corresponding to each image based on the covariance eigenvalue and gradient magnitude of the gray-level gradient matrix; acquiring a real-time image; filtering out the target historical image most similar to the real-time image through multi-scale feature matching; mapping the interference region of the target historical image to the real-time image through feature point registration; dynamically adjusting the expansion coefficient according to the local gray-level variance of the mapped region; predicting the number of pixels to be removed and generating a circular region mask; obtaining the image to be analyzed after removing interference; extracting candidate defect regions in the image to be analyzed through local contrast mapping and graph cut segmentation to obtain the defect type and parameters. This invention effectively eliminates process interference and significantly improves the accuracy and robustness of surface defect detection in electrical porcelain blanks.

This invention discloses a method for fusing InSAR and GNSS heterogeneous surface deformation based on an adaptive sparse spatial basis, belonging to the field of geological disaster monitoring and data processing technology. The method first acquires and aligns InSAR and GNSS deformation sequences; calculates the InSAR deformation feature field and its spatial gradient magnitude; sparsifies control points according to gradient thresholds to construct an adaptive sparse spatial basis set that is sparse in low-gradient regions and dense in high-gradient regions; models the deformation observations, combines the spatial basis set with the state transition equation, and uses filtering smoothing and the EM algorithm for iterative solution to reconstruct a high spatiotemporal resolution deformation sequence. This invention can adaptively match deformation spatial features, reducing the computational dimensionality while preserving local deformation details, and achieving high-precision fusion of heterogeneous data.

The invention relates to a flaw recognition system for a special-shaped sealing bolt forming process based on image processing, and belongs to the technical field of image recognition. The system considers that a complex structure on the surface of the special-shaped sealing bolt enables the gray gradient magnitude difference of different angular points to be large, abandons a traditional angular point judgment method for comparing the gradient magnitude with a threshold value, and determines the angular point degree according to the multi-dimensional features of a target pixel point and a neighborhood pixel point about the distance, the gradient vector difference and the gradient magnitude of the target pixel point and the neighborhood pixel point. All the pixel points are clustered in combination with the gray values and the angular point degrees of the pixel points, and the error degree caused by local too dark influence of the collected image when the angular point degrees of the pixel points in the class clusters are determined is determined according to the gray values, the sizes and the angular point degree mean values of all the classes of clusters; and on the basis of the error degree, combining the outlier degree of the target pixel point in the class cluster to determine an angular point degree correction coefficient, and completing angular point judgment. According to the method, the angular point judgment accuracy of the special-shaped sealing bolt can be remarkably improved, so that the defect recognition accuracy is improved.

This invention relates to an autofocus method based on Gaussian filtering and optimized Sobel gradient value calculation. It falls under the field of UAV pod optical imaging technology, specifically focusing technology for UAV pod optical cameras. It addresses the technical problem that after the camera acquires the light signal of an image, it performs some processing to improve image quality, resulting in overexposure and high noise, thus affecting the accuracy of autofocus. The invention involves cropping the middle 1 / 9 region of a grayscale image, applying Gaussian filtering, and then calculating the gradient magnitude of each pixel in each frame of the image using horizontal and vertical gradient kernels. Finally, the image corresponding to the maximum gradient value is selected to verify whether it is the clearest image, thus serving as the clearest image for focusing. The camera's focus is then adjusted to the position where this image was generated, completing the focusing process.