Miniature bearing detection system and detection method based on image detection

Abstract

The invention discloses a miniature bearing detection system and detection method based on image detection, and relates to the technical field of image processing, the miniature bearing detection system comprises a control center, and the control center is connected with an image acquisition module, an image processing module, an intelligent analysis module and a comprehensive detection module; the image acquisition module acquires production video data of the miniature bearing; a frame frequency is set in the image processing module to carry out picture extraction on the production video data, and a standard size is set to carry out pattern standardization to obtain a standard node pattern; setting a boundary extraction frame in the intelligent analysis module to perform region extraction and image gray level conversion on the standard node pattern to obtain a gray level boundary diagram; setting detection nodes in the comprehensive detection module, constructing a test boundary image set, carrying out sampling inspection and grouping, obtaining a reference boundary image set and a comparison boundary image set, and setting a feature comparison image to carry out boundary detection so as to obtain an abnormal comparison image; defects and abnormal conditions in the production process are monitored in real time, the rate of unqualified products is reduced, and the production cost is saved.

Description

Technical Field

[0001] The present invention relates to the technical field of image processing, and in particular to a micro bearing detection system and a detection method based on image detection. Background Art

[0002] Image detection refers to the use of image processing and analysis technology to identify and understand the content in an image, and to perform relevant processing and analysis. Through image detection, it is possible to extract information such as identification, location, counting, and attributes of objects in an image, thereby achieving understanding and decision-making of the scene.

[0003] The micro bearing detection system is a system specially designed for detecting small bearings. It usually includes components such as sensors, image processing equipment, control units and actuators. The purpose of this system is to ensure that micro bearings meet strict quality standards to ensure their normal operation and performance in engineering applications. Although micro bearing detection can improve the reliability and efficiency of bearing quality, it also has some disadvantages and challenges, including: the detection process requires higher accuracy and resolution, increased equipment and technical requirements, and thus increased production costs; at the same time, the detection of micro bearings requires operators with professional training and technical support, otherwise false alarms or missed detections may occur; therefore, optimizing the detection of micro bearings through image detection has important theoretical and practical significance.

[0004] How to use image processing technology to extract pictures from the collected production video data and set standard sizes to standardize the pattern to obtain a standard node pattern; set a boundary extraction box to perform area extraction and image grayscale conversion on the standard node pattern to obtain a grayscale boundary map; construct a test boundary atlas for boundary detection to obtain an abnormal comparison map is a problem we need to solve. To this end, a micro-bearing detection system and detection method based on image detection are provided. Summary of the invention

[0005] The object of the present invention is to provide a micro bearing detection system and a detection method based on image detection.

[0006] The purpose of the present invention can be achieved through the following technical solutions: A micro bearing detection system based on image detection includes a control center, wherein the control center is connected with an image acquisition module, an image processing module, an intelligent analysis module and a comprehensive detection module; The image acquisition module is used to set an acquisition port, and to acquire production video data, acquisition time and video duration of the miniature bearing through the acquisition port; The process of obtaining a production frame pattern includes: Set sampling nodes according to the acquired collection ports; Perform node positioning on the collected production video data according to the obtained sampling nodes to obtain the node video time; The frame frequency is set, and pictures are extracted from the obtained production video data according to the obtained frame frequency to obtain a production frame pattern.

[0007] The process of obtaining a standard node pattern includes: Time-segment the obtained video duration according to the obtained frame frequency to obtain frame time points, and associate the obtained production frame pattern with the corresponding frame time points; Match the obtained frame time point with the node video time to obtain the node matching time, and mark the production frame pattern corresponding to the node matching time as the matching node pattern; Marking pixels on the obtained matching node pattern to obtain matching pixel points; Setting a canonical size, the canonical size including a pixel limit and an input recognition channel; The matching node pattern is standardized according to the obtained standard size to obtain a standard node pattern.

[0008] The process of obtaining the feature boundary map includes: A boundary extraction frame is set according to the obtained matching pixel points, the obtained boundary extraction frame is uploaded to the standard node pattern, and the standard node pattern is extracted and cut according to the obtained boundary extraction frame to obtain a regional node graph; The obtained regional node graphs are sorted according to the obtained standard node patterns to obtain a regional extraction graph sequence; The obtained region extraction image sequence is subjected to region extraction according to the obtained boundary extraction frame to obtain a feature boundary image, and matching pixel points in the feature boundary image are marked as feature pixel points.

[0009] The obtained feature boundary map is subjected to image grayscale conversion, and the process of obtaining the grayscale boundary map includes: Color-marking the characteristic pixel points of the obtained characteristic boundary image according to the set input recognition channel to obtain a first characteristic point, a second characteristic point, and a third characteristic point; Setting a replacement pixel point according to the obtained characteristic pixel point, and obtaining a grayscale pixel point according to the obtained first characteristic point, second characteristic point and third characteristic point based on the replacement pixel point; The obtained grayscale pixels are replaced with replacement pixels until all feature pixels are replaced with corresponding grayscale pixels, a grayscale boundary map is obtained, and the obtained grayscale boundary map is associated with the corresponding sampling nodes.

[0010] The process of obtaining the reference boundary atlas and the comparison boundary atlas includes: Set the detection node according to the set sampling node; A sampling cycle is set according to the acquired acquisition time, a grayscale boundary map associated with the detection node is acquired based on the sampling cycle, a test boundary atlas is constructed according to the acquired grayscale boundary map, and the acquired grayscale boundary map is uploaded to the test boundary atlas; The obtained test boundary atlas is sampled and grouped to obtain a reference boundary atlas and a comparison boundary atlas.

[0011] The process of obtaining anomaly comparison graphs includes: Marking the grayscale boundary map in the reference boundary atlas as the reference boundary map, and marking the grayscale boundary map in the comparison boundary atlas as the comparison boundary map; Setting a feature comparison graph in the comparison boundary atlas, performing homology conversion between the obtained feature comparison graph and the reference boundary atlas to obtain a homology comparison graph; Perform boundary detection on the homology alignment graph according to the obtained reference boundary atlas to obtain an alignment prominence value; The obtained outstanding alignment values ​​are counted to obtain an abnormal alignment diagram.

[0012] Based on the above-mentioned micro-bearing detection system based on image detection, the present invention also provides a micro-bearing detection method based on image detection, comprising the following steps: Step 1: Set the acquisition port to collect the production video data, acquisition time and video duration of the miniature bearing; Step 2: Set the frame frequency to extract images from the production video data to obtain a production frame pattern, perform time matching on the production frame pattern according to the obtained video duration to obtain a matching node pattern, set a standard size, perform pattern standardization on the matching node pattern according to the set standard size, and obtain a standard node pattern; Step 3: Set a boundary extraction box to extract and cut the standard node pattern to obtain a regional node map, perform regional extraction on the obtained boundary extraction box and the regional node map to obtain a feature boundary map, and perform image grayscale conversion on the feature boundary map to obtain a grayscale boundary map; Step 4: Set the detection node and obtain the grayscale boundary map of the detection node, build a test boundary atlas based on the obtained grayscale boundary map, sample and group the obtained test boundary atlas, obtain a reference boundary atlas and a comparison boundary atlas, set the feature comparison map and the reference boundary atlas for boundary detection and statistics, and obtain an abnormal comparison map.

[0013] Compared with the prior art, the beneficial effects of the present invention are as follows: performing image extraction on the collected production video data to obtain the production frame pattern, and performing pattern standardization on the production frame pattern to obtain the standard node pattern, setting a boundary extraction frame to perform extraction cutting and region extraction on the standard node pattern to obtain a feature boundary map, and converting the video data of the production of miniature bearings into the image data of the sampling nodes, and performing feature extraction on the image data, which helps to remove the redundant information of the feature boundary map, retain meaningful features, and improve the accuracy and efficiency of image analysis; Perform image grayscale conversion on the feature boundary map to obtain a grayscale boundary map; convert the color image into a grayscale image to reduce processing complexity, retain the brightness information of the image, and facilitate unified processing of the feature boundary map, saving computing resources, and increasing compatibility and versatility. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0015] Figure 1 It is a schematic diagram of the present invention. DETAILED DESCRIPTION

[0016] The technical solution of the present invention will be clearly and completely described below in conjunction with the embodiments. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0017] like Figure 1 As shown, the micro bearing detection system based on image detection includes a control center, and the control center is connected to an image acquisition module, an image processing module, an intelligent analysis module and a comprehensive detection module; The image acquisition module is used to collect production video data of miniature bearings, and the specific process includes: Acquire a production line of a micro bearing, set a collection port according to the obtained production line, collect production video data of the micro bearing through the collection port, mark the time of collecting the production video data as the collection time, and mark the video duration of the collected production video data; It should be further explained that, in the specific implementation process, the production line refers to a series of equipment and process flow specifically used for the production of miniature bearings; the acquisition port is a high-definition camera that performs all-round monitoring and acquisition of the production process on the production line.

[0018] The image processing module is used to process the collected production video data to obtain a standard node pattern. The specific process includes: Set sampling nodes according to the obtained production lines and collection ports; Perform node positioning on the collected production video data according to the obtained sampling nodes to obtain the node video time; Furthermore, the node positioning is to match the position of the sampling node in the production line with the position in the production video data, obtain the sampling node in the production video data, and mark the time corresponding to the matched sampling node as the node video time, and the node video time represents the time of the corresponding production state of the produced miniature bearing on the sampling node; Setting the frame frequency, performing image extraction on the production video data according to the obtained frame frequency, and obtaining a production frame pattern; Time-segment the obtained video duration according to the obtained frame frequency to obtain frame time points; Associating the obtained production frame pattern with the corresponding frame time point; Match the obtained frame time point with the node video time to obtain the node matching time; Mark the production frame pattern corresponding to the node matching time as the matching node pattern; Marking pixels on the obtained matching node pattern to obtain matching pixel points; Further, the matching pixel points represent pixels in the matching node pattern; Setting the standard size, and standardizing the matching node pattern according to the obtained standard size to obtain the standard node pattern; The canonical size includes a pixel limit and an input identification channel, the pixel limit includes an input width and an input length; Furthermore, the input width and the input length jointly determine the number of matching pixels of the input image, wherein the input width represents the number of matching pixels of the image in the horizontal direction, and the input length represents the number of matching pixels of the image in the vertical direction; the input recognition channel represents the number of identifiable colors, and the color dimension of the input image is standardized by the input recognition channel. For example, if there are three input recognition channels, three different colors can be recognized. In this embodiment, there are three input recognition channels, namely red, green and blue; The intelligent analysis module is used to analyze the obtained standard node pattern to obtain a grayscale boundary map. The specific process includes: Setting a boundary extraction frame according to the obtained matching pixel points, wherein the boundary extraction frame includes an extraction frame size and an extraction frame quantity; Furthermore, the boundary extraction frame is composed of elements for calculation with matching pixel points. The size of the extraction frame includes extraction width and extraction length. The number of extraction frames is proportional to the clarity of the output image, that is, the more extraction frames there are, the more times the standard node pattern is extracted, and the higher the clarity of the output image. The obtained boundary extraction frame is uploaded to the standard node pattern, and the standard node pattern is extracted and cut according to the obtained boundary extraction frame to obtain a regional node graph; Furthermore, the extraction and cutting is to mark the boundary extraction box in the corresponding area of ​​the standard node pattern as the extraction node area, and cut out the extraction node area to obtain the regional node map, translate the boundary extraction box to reach the next extraction node area, and cut it again until all areas of the standard node pattern are covered to obtain the regional node map; The obtained regional node graphs are sorted according to the obtained standard node patterns to obtain a regional extraction graph sequence; Performing region extraction on the obtained region extraction image sequence according to the obtained boundary extraction frame to obtain a feature boundary image, and marking matching pixel points in the feature boundary image as feature pixel points; It should be further explained that, in the specific implementation process, the region extraction is to convolve the boundary extraction box with the matching pixel points in the region node graph in the order of the region extraction graph sequence to obtain a feature region graph, and combine the obtained feature region graphs according to the region extraction graph sequence to obtain a feature boundary graph; Performing image grayscale conversion on the obtained feature boundary map to obtain a grayscale boundary map; Furthermore, the image grayscale conversion is to convert the color feature boundary map into an image with only gray; The process of image grayscale conversion includes: Color-marking the characteristic pixel points of the obtained characteristic boundary image according to the set input recognition channel to obtain a first characteristic point, a second characteristic point, and a third characteristic point; Further, according to this embodiment, there are three input recognition channels, namely red, green and blue, then the first feature point is red, the second feature point is green and the third feature point is blue; Select any one of the feature pixel points in the feature boundary map as a replacement pixel point, and mark the obtained replacement pixel point as i, where i=1, 2, 3, ..., u1, and u1 is a positive integer; Mark the first feature point, the second feature point and the third feature point of the replaced pixel point, and mark the first feature point as , mark the second feature point as , mark the third feature point as ; Obtain grayscale pixels based on the first feature point, the second feature point, and the third feature point, and mark the obtained grayscale pixels as ,in, =0.299* +0.587* +0.114*

[0019] Further, according to the first feature point being red, the second feature point being green, and the third feature point being blue, the first feature point marks Represented as the intensity value of red, the second feature point marked Represented as the green intensity value, the third feature point marked It is expressed as the intensity value of blue; The obtained grayscale pixels are replaced with the replacement pixels, one of the remaining feature pixels is selected as the replacement pixel, the process of obtaining grayscale pixels is repeated, and the grayscale pixels are replaced with the replacement pixels until all feature pixels are replaced with the corresponding grayscale pixels, thereby obtaining a grayscale boundary map; The obtained grayscale boundary map is associated with the corresponding sampling nodes.

[0020] The comprehensive detection module is used to perform boundary detection on the obtained grayscale boundary image to obtain an abnormal comparison image. The specific process includes: Set the detection node according to the set sampling node; Furthermore, the detection node is to select any one of the sampling nodes set in the production line as the detection node to detect the produced micro bearings. Accordingly, the grayscale boundary image of the micro bearing obtained at the detection node is the image produced by the current sampling node. According to the same production line, the images of the micro bearings produced at the same sampling node are the same. A sampling cycle is set according to the acquired acquisition time, a grayscale boundary map associated with the detection node is acquired based on the sampling cycle, and a test boundary atlas is constructed according to the acquired grayscale boundary map; Upload the obtained grayscale boundary map to the test boundary atlas; The obtained test boundary atlas is sampled and grouped to obtain a reference boundary atlas and a comparison boundary atlas; Marking the grayscale boundary map in the reference boundary atlas as the reference boundary map, and marking the grayscale boundary map in the comparison boundary atlas as the comparison boundary map; It should be further explained that, in the specific implementation process, the sampling grouping is to first sort the grayscale boundary maps in the test boundary atlas according to the acquisition time, and upload the grayscale boundary maps to the reference boundary atlas and the comparison boundary atlas in chronological order. For example, the first grayscale boundary map is uploaded to the reference boundary atlas, the second grayscale boundary map is uploaded to the comparison boundary atlas, the third grayscale boundary map is uploaded to the reference boundary atlas, the fourth grayscale boundary map is uploaded to the comparison boundary atlas, and so on, until all grayscale boundary map grouping is completed; Selecting a comparison boundary map in the comparison boundary map set as a feature comparison map, performing homology conversion between the obtained feature comparison map and the reference boundary map set to obtain a homology comparison map; Furthermore, the homology conversion is based on the feature comparison map, and any reference boundary map is selected from the reference boundary map set, and the selected reference boundary map is converted into the feature comparison map, and the center of the feature comparison map is used as the conversion point, and the homology comparison map obtained is a feature comparison map and a reference boundary map that share the same center; Perform boundary detection on the homology alignment graph according to the obtained reference boundary atlas to obtain an alignment prominence value; The obtained alignment prominence value is marked as BD, where BD= , represents the segmentation area of ​​the reference boundary map in the reference boundary map set, represents the segmentation area of ​​the homology alignment graph, and " "express The absolute value of The obtained outstanding comparison values ​​are counted to obtain an abnormal comparison diagram; It should be further explained that, in the specific implementation process, the boundary detection is to concentrically segment the homology comparison map and the reference boundary map, use the homology comparison map as the concentric reference and fix it, and segment the reference boundary map according to the same center, that is, segment the boundary of the reference boundary map with the boundary of the homology comparison map, and obtain the segmentation area of ​​the reference boundary map, that is, ; Use the reference boundary map as the concentric reference and fix it, and divide the homology comparison map according to the same center, that is, divide the boundary of the homology comparison map with the boundary of the reference boundary map, and obtain the segmentation area of ​​the homology comparison map, that is, ; and the homology alignment graph and all reference boundary graphs in the reference boundary atlas are concentrically segmented; In particular, if the comparison highlight value is equal to 0, it means that the miniature bearings produced at the same inspection node are qualified during production, that is, the grayscale boundary map during the random inspection is qualified during production; if the comparison highlight value is not equal to 0, it means that the homologous comparison map is different from the reference boundary, and the grayscale boundary map during the random inspection is unqualified during production, that is, it is an abnormal comparison map.

[0021] Based on the above-mentioned micro-bearing detection system based on image detection, the present invention also provides a micro-bearing detection method based on image detection, comprising the following steps: Step 1: Set the acquisition port to collect the production video data, acquisition time and video duration of the miniature bearing; Step 2: Set the frame frequency to extract images from the production video data to obtain a production frame pattern, perform time matching on the production frame pattern according to the obtained video duration to obtain a matching node pattern, set a standard size, perform pattern standardization on the matching node pattern according to the set standard size, and obtain a standard node pattern; Step 3: Set a boundary extraction box to extract and cut the standard node pattern to obtain a regional node map, perform regional extraction on the obtained boundary extraction box and the regional node map to obtain a feature boundary map, and perform image grayscale conversion on the feature boundary map to obtain a grayscale boundary map; Step 4: Set the detection node and obtain the grayscale boundary map of the detection node, build a test boundary atlas based on the obtained grayscale boundary map, sample and group the obtained test boundary atlas, obtain a reference boundary atlas and a comparison boundary atlas, set the feature comparison map and the reference boundary atlas for boundary detection and statistics, and obtain an abnormal comparison map.

[0022] The preferred embodiments of the present invention disclosed above are only used to help explain the present invention. The preferred embodiments do not describe all the details in detail, nor do they limit the invention to only specific implementation methods. Obviously, many modifications and changes can be made according to the content of this specification. This specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the present invention, so that those skilled in the art can understand and use the present invention well. The present invention is limited only by the claims and their full scope and equivalents.

Claims

1. A micro bearing detection system based on image detection, including a control center, characterized in that: The control center is connected to an image acquisition module, an image processing module, an intelligent analysis module and a comprehensive detection module; The image acquisition module is used to set an acquisition port, and to acquire production video data, acquisition time and video duration of the miniature bearing through the acquisition port; The image processing module is used to set the frame frequency to extract pictures from the production video data, obtain the production frame pattern, perform time matching on the production frame pattern according to the obtained video duration, obtain the matching node pattern, set the standard size, perform pattern standardization on the matching node pattern according to the set standard size, and obtain the standard node pattern; The intelligent analysis module is used to set a boundary extraction box to extract and cut the standard node pattern to obtain a regional node graph, perform regional extraction on the obtained boundary extraction box and the regional node graph to obtain a feature boundary graph, and perform image grayscale conversion on the feature boundary graph to obtain a grayscale boundary graph; The comprehensive detection module is used to set the detection node and obtain the grayscale boundary map of the detection node, build a test boundary atlas according to the obtained grayscale boundary map, perform random sampling and grouping on the obtained test boundary atlas, obtain a reference boundary atlas and a comparison boundary atlas, set a feature comparison map and a reference boundary atlas for boundary detection, and obtain an abnormal comparison map.

2. The micro bearing detection system based on image detection according to claim 1 is characterized in that: The process of obtaining a production frame pattern includes: Set sampling nodes according to the acquired collection ports; Perform node positioning on the collected production video data according to the obtained sampling nodes to obtain the node video time; The frame frequency is set, and pictures are extracted from the obtained production video data according to the obtained frame frequency to obtain a production frame pattern.

3. The micro bearing detection system based on image detection according to claim 2 is characterized in that: The process of obtaining a standard node pattern includes: Time-segment the obtained video duration according to the obtained frame frequency to obtain frame time points, and associate the obtained production frame pattern with the corresponding frame time points; Match the obtained frame time point with the node video time to obtain the node matching time, and mark the production frame pattern corresponding to the node matching time as the matching node pattern; Marking pixels on the obtained matching node pattern to obtain matching pixel points; Setting a canonical size, the canonical size including a pixel limit and an input recognition channel; The matching node pattern is standardized according to the obtained standard size to obtain a standard node pattern.

4. The micro bearing detection system based on image detection according to claim 3 is characterized in that: The process of obtaining the feature boundary map includes: A boundary extraction frame is set according to the obtained matching pixel points, the obtained boundary extraction frame is uploaded to the standard node pattern, and the standard node pattern is extracted and cut according to the obtained boundary extraction frame to obtain a regional node graph; The obtained regional node graphs are sorted according to the obtained standard node patterns to obtain a regional extraction graph sequence; The obtained region extraction image sequence is subjected to region extraction according to the obtained boundary extraction frame to obtain a feature boundary image, and matching pixel points in the feature boundary image are marked as feature pixel points.

5. The micro bearing detection system based on image detection according to claim 4 is characterized in that: The obtained feature boundary map is subjected to image grayscale conversion, and the process of obtaining the grayscale boundary map includes: Color-marking the characteristic pixel points of the obtained characteristic boundary image according to the set input recognition channel to obtain a first characteristic point, a second characteristic point, and a third characteristic point; Setting a replacement pixel point according to the obtained characteristic pixel point, and obtaining a grayscale pixel point according to the obtained first characteristic point, second characteristic point and third characteristic point based on the replacement pixel point; The obtained grayscale pixels are replaced with replacement pixels until all feature pixels are replaced with corresponding grayscale pixels, a grayscale boundary map is obtained, and the obtained grayscale boundary map is associated with the corresponding sampling nodes.

6. The micro bearing detection system based on image detection according to claim 5 is characterized in that: The process of obtaining the reference boundary atlas and the comparison boundary atlas includes: Set the detection node according to the set sampling node; A sampling cycle is set according to the acquired acquisition time, a grayscale boundary map associated with the detection node is acquired based on the sampling cycle, a test boundary atlas is constructed according to the acquired grayscale boundary map, and the acquired grayscale boundary map is uploaded to the test boundary atlas; The obtained test boundary atlas is sampled and grouped to obtain a reference boundary atlas and a comparison boundary atlas.

7. The micro bearing detection system based on image detection according to claim 6 is characterized in that: The process of obtaining anomaly comparison graphs includes: Marking the grayscale boundary map in the reference boundary atlas as the reference boundary map, and marking the grayscale boundary map in the comparison boundary atlas as the comparison boundary map; Setting a feature comparison graph in the comparison boundary atlas, performing homology conversion between the obtained feature comparison graph and the reference boundary atlas to obtain a homology comparison graph; Perform boundary detection on the homology alignment graph according to the obtained reference boundary atlas to obtain an alignment prominence value; The obtained alignment prominence value is marked as BD, where BD= , Indicates the segmentation area of ​​the homology alignment graph; The obtained outstanding alignment values ​​are counted to obtain an abnormal alignment diagram.

8. The micro-bearing detection method of the micro-bearing detection system based on image detection according to any one of claims 1 to 7, characterized in that: The following steps are involved: Step 1: Set the acquisition port to collect the production video data, acquisition time and video duration of the miniature bearing; Step 2: Set the frame frequency to extract images from the production video data to obtain a production frame pattern, perform time matching on the production frame pattern according to the obtained video duration to obtain a matching node pattern, set a standard size, perform pattern standardization on the matching node pattern according to the set standard size, and obtain a standard node pattern; Step 3: Set a boundary extraction box to extract and cut the standard node pattern to obtain a regional node map, perform regional extraction on the obtained boundary extraction box and the regional node map to obtain a feature boundary map, and perform image grayscale conversion on the feature boundary map to obtain a grayscale boundary map; Step 4: Set the detection node and obtain the grayscale boundary map of the detection node, build a test boundary atlas based on the obtained grayscale boundary map, sample and group the obtained test boundary atlas, obtain a reference boundary atlas and a comparison boundary atlas, set the feature comparison map and the reference boundary atlas for boundary detection and statistics, and obtain an abnormal comparison map.

Images