107 results about "Region growing" patented technology

The invention provides an image tampering detection method based on a consistency graph and region growing clustering, and the method comprises the steps: guiding a region growing clustering process through introducing local consistency measurement, carrying out the explicit modeling of the spatial connectivity between pixels, and dynamically determining a tampering region in combination with the feature distance between clusters. The method effectively solves the problem of isolated noise regions caused by lack of spatial connectivity constraints and the problem of unstable region segmentation caused by fixed cluster number setting in the prior art, and has the advantages of improving the positioning precision of image tampering regions, enhancing the spatial connectivity constraint capability, reducing isolated noise regions and improving the precision of detection results.

The invention relates to the technical field of image analysis, in particular to a vegetation fine classification and recognition method and system based on an unmanned aerial vehicle, and the method comprises the following steps: obtaining a multispectral image through the unmanned aerial vehicle, extracting red edge reflectivity, NDVI and gray-level co-occurrence contrast, generating a feature vector in a standardized manner, calculating neighborhood offset to obtain a dynamic weight, and combining the dynamic weight into a weighted vector; high discrete features are screened as effective channels, multi-scale clustering is carried out, center and region growth extension recognition is optimized, and a vegetation classification atlas is generated. According to the method, a neighborhood pixel feature offset dynamic weight mechanism is introduced, multi-spectral feature dimension contribution degree is adjusted in a self-matching mode, effective channels are screened based on full-image dispersion, redundant interference is eliminated, image pyramid multi-scale clustering and consistency constraint are fused, the complex vegetation boundary recognition capability is improved, dynamic weight and multi-scale optimization are coordinated, and the method is high in robustness and high in robustness. Sample dependence is reduced, and accurate distinguishing of spectrum similar vegetation is achieved.

The invention relates to the technical field of image processing, in particular to a particle size detection method and system based on image processing. The method comprises the steps of obtaining a particle image and preprocessing to obtain a to-be-detected image; constructing a density index and a boundary adhesion degree for any pixel point in the to-be-detected image; taking a density index normalization result and a boundary adhesion normalization result corresponding to the pixel point as a segmentation feature vector of the pixel point; taking each pixel point in the image as a seed point, and segmenting the image by using a region growing method based on the distance of each segmentation feature vector to obtain a segmented image; and extracting a particle region in the segmented image, and analyzing the particle quality. The method has the effect of improving the particle quality detection accuracy.

The invention discloses a sugarcane live-action three-dimensional phenotype extraction method based on a 3D Gaussian splashing technology. The method comprises the following steps of data acquisition, wherein a 360-degree omnibearing shooting mode is executed around a single sugarcane plant for shooting or video recording to obtain complete image or video data of the sugarcane plant; preprocessing the data; 3D model construction comprises the following sub-steps: SfM sparse reconstruction: instance segmentation; the method comprises the following steps: carrying out 3D Gaussian Splicing guided by a mask; point cloud data processing includes the following sub-steps: point cloud preprocessing including voxel downsampling, statistical filtering and radius filtering; segmenting and identifying stalks / leaves; the segmentation comprises RANSAC plane segmentation, DBSCAN clustering and region growth; the phenotype analysis comprises grid reconstruction and area / volume / perimeter / plant height calculation. According to the invention, the problems of difficult data acquisition, large workload and incomplete information acquisition in the prior art are solved.

The invention relates to the technical field of video generation, in particular to a video frame generation system and display equipment based on texture reconstruction and trajectory optimization, and the system comprises a motion breakpoint detection module, a trajectory interpolation frame generation module, a texture anomaly positioning module, a texture interpolation reconstruction module and a video frame synthesis module. According to the method, an edge feature map is generated through edge detection and non-maximum suppression, the boundary recognition precision is improved, noise is suppressed, candidate regions are divided through region growth, the local feature extraction efficiency is enhanced, global redundancy is reduced, three-frame optical flow analysis captures trajectory breakpoints, and motion vector correction path deviation is predicted in combination with a Lucas-Kanade method. Color consistency verification optimizes interpolation to reduce cross-frame abrupt change; 64 * 64 grids and a gray level co-occurrence matrix locate texture abnormity; five-frame similarity strengthens time sequence stability; Gabor filtering extracts a main direction and combines an adjacent frame reconstruction texture direction to ensure consistency; and trajectory smoothness and visual consistency are optimized through multi-frame cooperation and layered detection.

The invention discloses a binocular line laser reconstruction method and system, and relates to the technical field of computer vision, and the method comprises the following steps: S001, calibrating an internal reference and an external reference of a binocular camera, and obtaining a left image and a right image; arranging a line laser to respectively project parallel light knife surfaces on the left and right images; s002, respectively carrying out sub-pixel-level center extraction on the left image and the right image to obtain a binary mask, and calculating a candidate intersection point set; s003, calculating a basic matrix, determining a corresponding point on the right image based on the intersection point of the left image, and outputting a matching pair; s004, triangularizing the matching pair to obtain a 3D intersection point, and then marking the attributes of the light knife surface; s005, region growing is carried out after initialization; s006, calibrating a smooth knife surface equation; and S007, laser stripe full-pixel matching is carried out, and a final 3D point cloud is output. According to the invention, two or more groups of parallel line lasers in different directions are simultaneously utilized in the binocular system, so that unambiguous and high-density three-dimensional reconstruction is realized, the scanning efficiency is improved, and the ambiguity problem existing in the prior art is solved.

The invention discloses a large-parallax image splicing method and device, equipment and a storage medium, and belongs to the technical field of image processing. The method comprises the following steps: performing feature point matching on a to-be-spliced feature point set and a reference feature point set to obtain a feature matching point pair set; according to the parallax value of each feature matching point pair in the feature matching point pair set, layering the parallax of an overlapping region intersected with the reference image in the to-be-spliced image to obtain a plurality of parallax layers and a matching feature point pair set corresponding to each parallax layer; according to the matching feature points to be spliced in the matching feature point pair set of each parallax layer, determining each geometric feature region by using a region growing method; and according to the global homography matrix and the local homography matrix of each geometric feature region, performing fusion processing on the to-be-spliced image and the reference image to obtain a target spliced image. By using the technical scheme provided by the invention, the accuracy of large parallax scene image splicing can be improved.

The invention relates to a floating algae detection method, system and device. The method comprises the following steps: preprocessing a multi-vision original image sequence to obtain a multi-vision preprocessed image sequence; performing image splicing on the multi-vision preprocessed image sequence according to the boundary features to obtain a panoramic image; carrying out step-by-step edge identification and labeling on the panoramic image by adopting a multi-operator and region growing method to obtain a labeled image; and performing type identification and counting on the floating algae in the labeled image based on a stacked integrated learning method to obtain floating algae biological information. According to the method, image splicing is performed firstly, then edge recognition is performed, and multi-operator step-by-step edge recognition from low precision to high precision is performed on the panoramic image, so that the detection efficiency can be improved; in addition, the types of the floating algae in the annotated image are identified and counted based on the stacked integrated learning method, and the identification accuracy is higher than that of an existing single model.

Provided in the present disclosure are a segmentation method for fluorescent spots in fluorescent images, and a computer device and a storage medium. The segmentation method comprises: performing base calling on at least some pixel positions in a plurality of fluorescent images, so as to obtain base sequences of the at least some pixel positions, wherein the plurality of fluorescent images are obtained by means of a plurality of consecutive sequencing cycles; aligning the base sequences of the at least some pixel positions with base sequences in a reference genome, so as to determine the error rates of the base sequences of the at least some pixel positions; using the pixel positions, the error rates of the base sequences of which are less than a preset error rate threshold value, as candidate pixel positions for region growing; and determining, from among the candidate pixel positions for region growing, seed points for base sequence region growing, performing region growing on the basis of the similarity between the base sequences of the seed points and the base sequences of the candidate pixel positions within the neighborhood of the seed points, and using a plurality of regions obtained after growth as a plurality of fluorescent spots in the fluorescent images.

The present application relates to the technical field of medical image processing, and provides a knee femur and tibia segmentation method based on region growing, comprising the following steps: S1, acquiring knee medical images; S2, selecting a growth seed point; S3, coarsely segmenting the knee medical images to obtain knee stem data; S4, finely segmenting the knee medical images to obtain knee connection data; S5, fusing the knee stem data and the knee connection data to obtain knee femur and tibia image data with complete contours; S6, coarsely filling the knee femur and tibia image data to obtain knee femur and tibia image data after coarse filling; and S7, iteratively filling holes in the knee femur and tibia image data by using a three-dimensional sliding window, so as to obtain image data that is consistent with the surrounding data in shape, position and completeness and smoothness. The present application solves the problems of slow segmentation speed, unstable segmentation effect, incomplete segmentation and adhesion in preoperative planning of knee replacement surgery.

The application discloses a remote sensing image water body extraction method based on physical statistical distribution guidance and large model space constraint, comprising the following steps: acquiring a remote sensing image and performing pretreatment, calculating a physical statistical feature map and performing probability density distribution analysis, automatically determining a scene category to which the remote sensing image belongs, adaptively matching a hierarchical threshold calculation strategy, and acquiring a seed threshold and a growth threshold; performing region growing processing based on the seed threshold and the growth threshold to generate a physical prior coarse mask; performing connected domain analysis and adaptive outer expansion processing on the physical prior coarse mask to form a prompt box; inputting the pretreated remote sensing image and the prompt box into a visual basic model to output a fine segmentation prediction map; constructing a physical constraint search area based on the physical prior coarse mask, performing cascade fusion on the fine segmentation prediction map and the physical constraint search area, and obtaining a high-precision water body binary mask. Based on this, the application realizes automatic water body extraction without training samples, with strong noise resistance and fine edges.

A kind of photovoltaic panel foreign matter identification method and cleaning device suitable for micro-grid system, including the coordinate positioning of photovoltaic panel;Industrial camera collects the photovoltaic panel picture of positioning area;Photograph is identified using YOLOv3 algorithm and is identified again positioning;The focal length of industrial camera is adjusted using new positioning information to enlarge photovoltaic panel to the central part of industrial camera for picture capture;After the complete photovoltaic panel information is identified, the outline of processed photovoltaic panel picture is extracted, the area of photovoltaic panel is extracted using region growing method, and whether photovoltaic panel has foreign matter is judged;The positioning and identification problem of photovoltaic panel are realized by using the image recognition analysis technology of deep learning and convolutional neural network in the application, and the world coordinate system conversion method is used as the basis to enlarge and capture the corresponding position of photovoltaic panel by industrial camera, which has the advantage of accurate foreign matter identification, secondly, the cleaning device provided by the application has the advantages of simple structure, energy saving and high cleaning efficiency.

The present application relates to the technical field, especially to a kind of based on three-dimensional information's forming weld appearance defect extraction method.The method specifically includes: the three-dimensional point cloud data of weld is collected;Three-dimensional point cloud data is preprocessed, including using statistical filtering method to remove outlier, and using the point closest to the voxel grid centroid in voxel grid is selected to carry out point cloud data downsampling;After three-dimensional point cloud data is preprocessed, through LMedS adaptive threshold segmentation method, the extraction of weld is carried out, in the weld point cloud that extraction is completed, the framework of region growing method is established, seed point is selected, whether the point in the neighborhood of seed point belongs to the same category according to normal angle threshold and curvature threshold is judged, the extraction of weld surface defect point cloud is realized.The method provided by the present application can more completely obtain three-dimensional defect geometric information, improve the overall operation efficiency of system, and be more conducive to the complete extraction of weld defect.

The invention discloses a real-time detection method and system for surface defects of an aluminum-plastic composite film based on machine vision, and particularly relates to the technical field of machine vision and image analysis. A high detection rate and a low false alarm rate cannot be considered due to the fact that a target signal is weak and is aliasing with noise; a sequence is formed by acquiring continuous multi-frame surface images; calculating the statistical magnitude of each pixel position in the time dimension; analyzing a collaborative change mode of the pixel points and spatial neighborhoods thereof based on the statistics to screen candidate points; performing region growth on the candidate points to form a connected abnormal region, and extracting the region feature quantity of the region; and finally, evaluating and marking by combining the time statistics and the spatial region characteristic quantity, thereby reliably identifying and positioning the micro-defects on the surface of the aluminum-plastic composite film from the noise background.

This invention relates to the field of tunnel detection and identification technology, specifically a method for detecting defects in water diversion tunnels based on a 3D model. The method includes: collecting laser point cloud data, panoramic image sequences, and ground-penetrating radar (GPR) profile data of the target water diversion tunnel section to form a multi-source detection data set. A 3D mesh model of the tunnel structure is reconstructed using the laser point cloud data. The panoramic image sequence is mapped onto the model to generate a textured 3D model. Abnormal signal segments from the GPR profile data are identified to obtain a set of radar anomaly depth markers, which are then projected onto the textured 3D model to construct a 3D detection model with defect markers. Candidate defect regions are extracted through region growing and segmentation. The opening width and extension length of each candidate defect region are quantified. Based on geometric features, the defect type is classified, and a defect detection report is output. This method achieves the fusion and correlation of multi-source detection data in 3D space, completing automatic defect region segmentation and geometric feature quantification.

The present application relates to the technical field of visual detection, and discloses a gear tooth profile defect detection method and system based on image recognition, the method comprising: carrying out pixel-level gray difference coding on a gear end face original image to obtain a gray coding sequence; based on the gray coding sequence, carrying out neighborhood comparison on the Hamming distance between pixels in the gear end face original image to obtain a local anomaly distribution; carrying out intensity sorting screening on the local anomaly distribution to obtain a candidate abnormal seed point; taking the candidate abnormal seed point as a starting point, carrying out similarity weighted region growing on the candidate abnormal seed point to obtain a suspected defect region; carrying out consistency checking on the suspected defect region, carrying out outlier elimination on the pixel coding deviation value of the checked region to obtain a refined defect region, and carrying out edge smoothing processing on the refined defect region to obtain a gear face damage region of the gear end face original image; the present application can improve the efficiency of a gear tooth profile defect detection based on image recognition.

The present application relates to the field of human state monitoring, and particularly relates to a human state monitoring method and system based on thermal imaging technology, the method comprising: acquiring an infrared thermal image and a visible light image of a human body; registering the infrared thermal image to the coordinate system of the visible light image and extracting a temperature field; performing posture evaluation on the visible light image to extract skeleton key points, and dividing the human body into multiple left-right symmetrical regions of interest based on the skeleton key points; calculating correction thermal symmetry difference values between the left-right symmetrical regions of interest to obtain a thermal symmetry difference map; performing region growing based on the point with the maximum pixel value in the thermal symmetry difference map to obtain an abnormal thermal spot region; calculating the product of an inflammation activity index and a boundary confidence of the abnormal thermal spot region to obtain a state risk value; and in response to the state risk value being greater than a set threshold, generating an alarm signal. The present application solves the problem of high false alarm rate of human state monitoring in a dynamic and complex environment.

This invention discloses a non-contact tire deformation recognition method based on a fine-tuned large-scale visual model. The method includes calibrating a monocular high-speed camera, fine-tuning the parameters of the mask decoder of the large-scale visual model using a tire image dataset, generating a point coordinate cue sequence based on pixel calculation using OpenCV, inputting the cue sequence into the fine-tuned large-scale model for segmentation, further processing the pixel matrix using OpenCV to generate a point cue sequence, and post-processing using iterative geometric fitting and region growing algorithms to obtain the mechanical deformation parameters of the target sample tire. This invention can achieve accurate and rapid tire deformation recognition, reaching pixel-level precise segmentation. It solves the problems of large errors, limited measurement environments, nonlinear image distortion caused by cameras, weak generalization ability, and high training costs in existing computer vision-based vehicle tire recognition methods.

The application provides a robot autonomous planning method based on a scene graph, and the method comprises the following steps: S1: reconstructing a three-dimensional point cloud of a scene based on multi-view images, and performing over-segmentation on the three-dimensional point cloud to obtain a plurality of point cloud segments; meanwhile, multi-view boundary maps are extracted from the multi-view images; S2: based on the multi-view boundary maps, the multi-view boundary affinity between any two adjacent point cloud segments is calculated; S3: based on the multi-view boundary affinity, region growing clustering is performed on all the point cloud segments to obtain object instances in the scene; S4: semantic information is given to each object instance, and a Voronoi diagram is constructed based on the spatial positions of all the object instances. The application utilizes the description of the spatial information and open semantic features of the object instances in the scene topology to assist a large language model in combining the task and behavior planning of the robot itself, and solves the problem that the task planner of the large language model is generally disconnected with the actual environment.

The invention provides a point cloud segmentation method and device based on improved region growth and concavity-convexity clustering, which can be applied to the field of image analysis, and comprises the following steps: carrying out region growth segmentation on a target point cloud according to a neighborhood curvature and a neighborhood normal of each sampling point in the target point cloud to obtain a discrete point set and a plurality of initial plane point sets; according to a spatial geometrical relationship between the sampling points and the plurality of initial plane point sets and a spatial geometrical relationship between the plurality of first neighborhood points corresponding to the sampling points and the plurality of initial plane point sets, distributing the plurality of sampling points of the discrete point set to the plurality of initial plane point sets to obtain a plurality of target plane point sets; and performing concavity and convexity clustering on the plurality of target plane point sets to obtain point clouds of the plurality of target objects.

The invention discloses a welding path priority method and welding equipment for an automobile electronic ignition coil, and the welding method comprises the steps: obtaining point cloud data through preset track global scanning, generating a digital three-dimensional model consistent with a welding driving motion coordinate system through Gaussian filtering denoising, iterative nearest point registration, B-spline surface fitting and other strategies, and obtaining a welding driving motion coordinate system; the problem of coordinate deviation of traditional path planning is solved. A contour recognition strategy is combined with region growth, feature extraction and template comparison, and a least square method is used for fitting a center line to serve as a reference track; a node setting strategy is adopted, forced and uniform nodes are scientifically arranged, and accurate positioning of a welding head is guaranteed. A double correction mechanism of molten pool monitoring and node feedback is innovatively adopted, a thermal imager collects images and extracts molten pool parameters after preprocessing, a real-time correction instruction is generated in combination with three-level judgment, a lower node adjustment instruction is generated through image processing after node welding, closed-loop control is achieved, welding defects are reduced, and the percent of pass is increased. Corollary equipment adopts an integrated disc design and is matched with a rotary bearing mechanism to realize batch positioning and conveying, a multi-degree-of-freedom mechanical arm performs automatic welding according to a planned path, parameters are adjusted according to welding seam characteristics, and both precision and production efficiency are considered.

The application discloses a hyperspectral remote sensing image change detection method based on adaptive iterative filtering, and specifically comprises the following steps: step one, selecting three bands with the largest information quantity from two groups of hyperspectral remote sensing images at different times; step two, writing the selected two groups of bands into the red, green and blue bands of a natural image in the order of the index value from small to large; step three, performing adaptive region growing on the two pictures pixel by pixel, if the difference between the pixel value of the neighborhood and the center pixel is less than the standard deviation of the eight-neighborhood pixels, the pixel is merged into the adaptive region, otherwise, the pixel is not merged; step four, performing region growing on the two filtered pictures in step three again to obtain a change amplitude graph; and step five, calculating the threshold of the obtained change amplitude graph to obtain a final change detection result graph. By adopting the method, the homogenous regions on the picture are more obvious, and the precision of the hyperspectral remote sensing image change detection is improved.

The invention belongs to the technical field of image processing, and particularly relates to a steel wheel weld joint welding quality detection method based on artificial intelligence. The method comprises the following steps: acquiring a steel wheel welding position image; carrying out weighted fusion Gaussian median hybrid filtering denoising on the image, positioning a weld boundary by combining Canny edge detection with Hough linear transformation, and extracting a weld ROI image by using a region growing method; constructing an ROI image gray matrix, straightening along the center line of the welding seam to form a gray sequence, and smoothing through moving average filtering; calculating a gray difference sequence, marking and merging abnormal segments through a sliding window, and extracting gray mean deviation degree and gray distribution entropy features; and normalizing the features, inputting the normalized features into a preset neural network model, and outputting a weld quality score. The detection accuracy and stability of the welding position of the steel wheel are improved, and the efficient quality control requirement of large-scale automatic production is met.

The application discloses an image feature processing method for fire source detection in a dynamic scene, and belongs to the field of image feature processing; the method comprises the following steps: collecting an original infrared image of a dynamic scene, and obtaining a normalized infrared image through preprocessing; a dynamic region segmentation algorithm is used to generate a binary image; a radiation gradient weighted feature extraction algorithm is used to generate a weighted feature image; a dynamic boundary feature enhancement algorithm is used to generate a dynamic enhanced feature image; an edge-guided region growing algorithm is used to take edge pixels as seed points for region growing, and finally a complete fire source region image is output; and an adaptive threshold segmentation method is used to realize accurate segmentation of a high-temperature region and a background region based on local temperature statistical characteristics, the threshold is dynamically adjusted, the temperature distribution change of different scenes is adapted, the robustness of segmentation is significantly improved, the high-temperature region and the background region can be accurately distinguished in a complex dynamic scene, and the influence of environmental interference on fire source detection is reduced.

The invention relates to the technical field of fabric defect identification, in particular to a fabric defect identification method and system based on image processing, and the method comprises the steps: collecting a normal image of a jacquard knitted fabric, and processing the normal image to obtain a normal grayscale image; sampling the normal grayscale image to obtain a plurality of window images, and performing hierarchical clustering on the window images to obtain a plurality of clusters; training a convolution auto-encoder for each cluster, counting reconstruction errors of each pixel after all window images in each cluster are reconstructed by the convolution auto-encoder, and establishing a reconstruction error distribution reference library; the method comprises the following steps: acquiring a to-be-detected image of a to-be-detected jacquard knitted fabric, processing the to-be-detected image to obtain a to-be-detected window image, inputting the to-be-detected window image into a convolutional auto-encoder of a matched cluster to obtain a reconstruction error graph, and recognizing and extracting a defect area through a self-adaptive area growing method in combination with a reconstruction error distribution reference library of the cluster. According to the invention, the precision and robustness of defect detection are obviously improved.

The present invention relates to the technical field of material identification, and more specifically to a method and system for image recognition of material types and capacities, comprising: obtaining the similarity between any two adjacent superpixel blocks based on the grayscale information, gradient information, and saturation information of pixel points in any two adjacent superpixel blocks, and using this information for region growing to obtain a number of regions in each storage bin material image; dividing all regions into light-explosive regions, light-weak regions, and normal regions; determining the gamma value of each light-explosive region or light-weak region during gamma transformation based on the grayscale difference between each light-explosive region or light-weak region and all normal regions, and using this value for image enhancement to obtain an enhanced storage bin material image; and training each quarter based on the enhanced storage bin material image to identify the material type and capacity. The present invention improves the accuracy of material type and capacity recognition.

This application relates to the field of image processing technology, specifically to a high-voltage power supply DNA sequencing visual inspection method and system. The method includes: acquiring a four-channel DNA electrophoresis image using Sanger sequencing technology; acquiring contours in the DNA electrophoresis grayscale image; constructing a DNA local density confidence coefficient and local DNA density parameters for each pixel based on the grayscale values ​​of each pixel within a channel and its neighboring pixels; acquiring the density run matrix for each pixel; constructing a DNA density distribution heterogeneity coefficient and channel tailing effect evaluation parameters for each pixel, and based on these, constructing a sequencing segmentation probability coefficient for each pixel; acquiring a decision threshold for each pixel; using the center of each contour as a seed point, and acquiring each DNA fragment based on an adaptive threshold using a region growing method, thereby obtaining the DNA sequence. This application can improve the accuracy of detecting the base sequence of a DNA sequence.

The embodiment of the invention provides a bone microstructure segmentation method and device, a storage medium and a program product. In the embodiment of the invention, a bone microstructure image is obtained, a weighted graph network is constructed according to the spatial change rate of voxel gray values, and the discrete curvature of each voxel block is calculated based on the weighted graph network so as to extract positive curvature components in a specified data range to form a positive curvature feature graph; and screening a target voxel block in combination with a bone tissue HU range, determining a maximum connected bone tissue region in the positive curvature feature map, and gradually complementing a boundary and an internal structure through region growth and pore filling so as to obtain a segmentation result of a bone microstructure. According to the method, curvature-driven geometric feature extraction and region growth evolution based on gray constraint are taken as the core, the segmentation problem in ultrahigh-resolution bone microstructure image processing is effectively solved, and automatic and high-precision segmentation of the bone microstructure image is realized on the premise of not depending on a large amount of manual annotation data.

The present invention discloses an image processing method, apparatus, device and computer-readable storage medium, wherein the method comprises: obtaining an initial segmentation mask image of a preset shooting scene in an image to be processed; performing region growth processing on the initial segmentation mask image to obtain a grown segmentation mask image; determining multiple target connected domains in a difference binary image corresponding to both the initial segmentation mask image and the grown segmentation mask image; when the maximum area of ​​the multiple target connected domains meets a preset condition, generating a target segmentation mask image according to the initial segmentation mask image; processing the image to be processed according to the target segmentation mask image, wherein the target segmentation mask image clearly displays the details at the boundary between different regions in the image, and is easy to process more finely. The image processing method provided by the present invention improves the accuracy of image processing.

The invention relates to the technical field of image data processing, in particular to a logistics violation behavior recognition method and device based on a monitoring video, and the method comprises the steps: obtaining an image in a logistics transportation process, extracting a plurality of shadow boundaries, carrying out the region growth through employing the midpoint of the shadow boundaries as a seed point, and generating an initial shadow; if the area ratio of the initial shadow to the shadow boundary is smaller than a set threshold value, removing the initial shadow to obtain a second shadow, and performing behavior recognition on the second shadow by adopting a behavior analysis algorithm; the shadow boundary extraction specifically comprises the steps of corroding a connected domain in an image foreground to obtain a first target, obtaining a second target through subtraction operation, and obtaining points on a unit circle by taking the geometric center of the second target as a circle center to form a feature vector; and mapping the feature vectors into a three-dimensional space for clustering, and calculating a mean value of target values in a cluster to obtain a shadow boundary. According to the invention, the problem of low logistics monitoring accuracy is solved.