A device crack detection system based on millimeter wave imaging radar

Abstract

The present invention relates to the field of workpiece detection, and discloses a device crack detection system based on millimeter wave imaging radar, a data module, which is used to collect point cloud data of a good workpiece and generate a point cloud model of the whole good workpiece based on the point cloud data; a deviation module, which is used to divide the point cloud model of the good workpiece into sub-areas, generate first point cloud data of each sub-area, and use the corresponding part of the point cloud data in the whole point cloud model of the good workpiece as the second point cloud data, calculate the difference based on the first point cloud data and the second point cloud data, and generate the deviation factor of each sub-area; a detection module, which is used to generate feature point cloud data of each sub-area of ​​the target workpiece.

Description

Technical Field

[0001] The present invention relates to the field of workpiece detection, and more particularly, to a device crack detection system based on millimeter wave imaging radar. Background Art

[0002] Millimeter-wave imaging radar can effectively identify cracks in workpieces through the reflection characteristics of high-frequency electromagnetic waves. An existing crack detection system based on millimeter-wave imaging radar usually uses point cloud data as a basis, maps the point cloud data generated by millimeter-wave scanning into a three-dimensional model, and uses this to display the geometric shape and surface features of the workpiece. During the inspection process, engineers usually manually determine whether there are cracks or defects in the workpiece by observing the shape or differences of the three-dimensional model. In summary, the problems of high detection cost and low detection efficiency in the existing technology need to be solved urgently. Summary of the invention

[0003] The present invention provides a device crack detection system based on millimeter wave imaging radar, which solves the technical problems raised in the background technology.

[0004] The present invention provides a device crack detection system based on millimeter wave imaging radar, comprising:

[0005] The data module is used to collect point cloud data of good workpieces and generate a point cloud model of the whole good workpiece based on the point cloud data through millimeter wave imaging radar;

[0006] The deviation module is used to divide the point cloud model of the good workpiece into sub-areas, generate the first point cloud data of each sub-area, and use the corresponding part of the point cloud data in the overall point cloud model of the good workpiece as the second point cloud data, calculate the difference based on the first point cloud data and the second point cloud data, and generate the deviation factor of each sub-area; wherein the deviation factor includes:

[0007] Step 1: Physically segment the good workpiece into several sub-blocks, each of which corresponds to a sub-region. The physical segmentation includes:

[0008] Determine the symmetry plane of good workpieces;

[0009] Based on the symmetry plane, the good workpiece is divided in the same way on both sides according to the principle of symmetry;

[0010] Step 2, generating first point cloud data for each sub-block by using a millimeter wave imaging radar;

[0011] Step 3, extracting partial point cloud data corresponding to each sub-block from the overall point cloud data of the good workpiece to generate second point cloud data;

[0012] Step 4, calculating the difference between the first point cloud data and the second point cloud data of each sub-region, and using the difference as the deviation factor of the corresponding sub-region;

[0013] The detection module is used to generate feature point cloud data of each sub-region of the target workpiece, and correct the feature point cloud data based on the deviation factor to generate corrected point cloud data; compare the corrected point cloud data with the second point cloud data to determine the abnormal state of the sub-region; wherein the good workpiece and the target workpiece are both the same workpiece and are symmetrical in structure;

[0014] The verification module is used to comprehensively analyze the abnormal status of all sub-areas, screen the abnormal sub-areas of the target workpiece based on the results of the comprehensive analysis, and determine the crack area of ​​the target workpiece.

[0015] Furthermore, the feature point cloud data is corrected based on the deviation factor to generate corrected point cloud data, including:

[0016] Step 5: Generate feature point cloud data for each sub-region of the target workpiece, and calculate feature difference between the feature point cloud data of each sub-region and the corresponding deviation factor, and use the feature difference as the corrected point cloud data of the corresponding sub-region.

[0017] Further, comparing the corrected point cloud data with the second point cloud data to determine the abnormal state of the sub-region includes:

[0018] Step 6, determining whether the number of point clouds in the corrected point cloud data of the i-th sub-region is greater than the number of point clouds in the second point cloud data; if greater, fitting the second point cloud data to match the number of point clouds in the corrected point cloud, specifically including:

[0019] Generate multiple fitting schemes according to the point cloud position information in the corrected point cloud data and the second point cloud data;

[0020] The degree of fit is calculated for each fitting scheme, and the calculation formula for the degree of fit of the fitting scheme is as follows:

[0021] ;

[0022] ;

[0023] ;

[0024] ;

[0025] in, represents the fitness of the fitting scheme, , and denotes weight parameters, and , and All are not 0, , and The sum is 1, , and They represent the first similarity, the second similarity and the third similarity of the ith point in the corrected point cloud data in terms of spatial coordinates, color information and reflection intensity, respectively. Indicates the number of point clouds of the second point cloud data, The index indicating the number of points in the second point cloud data. Represents the three-dimensional spatial coordinates of the i-th point in the corrected point cloud data, represents the three-dimensional spatial coordinates of the jth point in the second point cloud data, The bandwidth weight representing the similarity of spatial coordinates, Represents the color information of the i-th point in the corrected point cloud data, Represents the color information of the jth point in the second point cloud data, The bandwidth weight representing the similarity of color information, Represents the reflection intensity of the i-th point in the corrected point cloud data, Represents the reflection intensity of the jth point in the second point cloud data, The bandwidth weight representing the similarity of reflection intensity, represents the square of the Euclidean distance, represents the Gaussian kernel function, express Activation function;

[0026] Step 7: Select the solution with the highest fitting degree and calculate the fitting deviation between the corrected point cloud data and the second point cloud data. The calculation formula of the fitting deviation is as follows:

[0027] ;in, represents the fitting deviation;

[0028] Step 8: If the fitting deviation is greater than a preset fitting deviation threshold, the i-th sub-region is judged to be in an abnormal state; otherwise, the i-th sub-region is judged to be in a normal state.

[0029] Furthermore, the abnormal status of all sub-areas is comprehensively analyzed, including:

[0030] Step 9, obtaining shape and size information of cracks of the historical workpiece, and clustering the cracks based on the shape and size information to generate a number of characteristic cracks;

[0031] Step 10, performing connectivity analysis on adjacent abnormal state sub-regions to determine connected regions;

[0032] Step 11, performing a first verification based on the size of the characteristic crack and the size of the connected area; performing a second verification based on the shape of the characteristic crack and the shape of the connected area.

[0033] Furthermore, the abnormal status of all sub-areas is comprehensively analyzed, including:

[0034] Step 12, determining the sub-regions that are symmetrical about the symmetry plane, and judging that the abnormal states of the sub-regions that are symmetrical about the symmetry plane are the same, then the sub-regions that are symmetrical about the symmetry plane are thirdly verified.

[0035] Furthermore, based on the results of the comprehensive analysis, the abnormal sub-regions of the target workpiece are screened to determine the crack region of the target workpiece, including:

[0036] Step 13, if the connected region passes the first verification, the subregion in the connected region passes the first screening; if the connected region passes the second verification, the subregion in the connected region passes the second screening;

[0037] Step 14, performing symmetric matching on the sub-regions that pass the first screening and the second screening about the symmetry plane; if the matching is successful, performing a third screening on the successfully matched sub-regions according to the fitting deviation, specifically including:

[0038] The fitting deviations of the sub-regions on both sides of the symmetry plane that have passed the first verification and the second verification and have been successfully matched are compared. If the fitting deviation of the a-th sub-region is greater than the fitting deviation of the b-th sub-region, the a-th sub-region is the crack region of the target workpiece.

[0039] The beneficial effect of the present invention is that through millimeter wave radar scanning combined with point cloud data processing and symmetry matching, the noise caused by symmetric surface reflection can be accurately identified and eliminated, thereby effectively improving the accuracy of crack detection. This method can not only deeply detect tiny cracks on complex surfaces (such as vases) without traditional optical or visual interference, but also effectively avoid misjudgment and missed judgment through a multi-level screening and verification mechanism combined with characteristic crack morphology analysis, providing an efficient, accurate and robust surface crack detection solution. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 It is a module diagram of a device crack detection system based on millimeter wave imaging radar of the present invention. DETAILED DESCRIPTION

[0041] The subject matter described herein will now be discussed with reference to example embodiments. It should be understood that the discussion of these embodiments is only to enable those skilled in the art to better understand and implement the subject matter described herein, and the functions and arrangements of the elements discussed may be changed without departing from the scope of protection of the contents of this specification. Each example may omit, replace or add various processes or components as needed. In addition, the features described relative to some examples may also be combined in other examples.

[0042] Specifically, assuming that the workpiece is a vase, obtain the point cloud data of a good vase:

[0043] First, the good curved vase is scanned using traditional radar scanning technology to obtain its overall point cloud data. Due to the penetrating power of radar, the obtained point cloud data includes not only the point cloud data reflected from the vase surface, but also the point cloud noise reflected from the symmetric surface of the vase. At this point, the point cloud data is processed by the existing algorithm to generate a point cloud model of the good vase.

[0044] Remove reflection noise from symmetrical surfaces:

[0045] In order to eliminate the symmetry surface reflection noise introduced by radar scanning, we physically segment the vase according to a predetermined rule to obtain multiple segmented sub-blocks. Each sub-block is scanned separately, and the resulting point cloud data is called the first point cloud data. Since the scan data of each sub-block does not contain the symmetry surface reflection noise, the first point cloud data only contains the real point cloud information of the corresponding sub-block.

[0046] At the same time, for the good vase that has not been segmented, the point cloud data of the vase is extracted using the overall scanning method, and the point cloud data of each sub-block is extracted from the overall point cloud to obtain the second point cloud data. By calculating the difference between the first point cloud data and the second point cloud data, the point cloud noise generated by the reflection of the symmetry surface can be obtained, and these differences are the deviation factors. The deviation factor represents the reflection noise of each sub-block and can be used as a reference for correcting the point cloud data of the target vase.

[0047] Correct the point cloud data of the target vase:

[0048] For the target vase to be detected, it is divided into several sub-areas and scanned according to the same segmentation rules as the qualified vase. The point cloud data of the target vase in each sub-area is obtained by scanning, and these point cloud data are used as the feature point cloud data of the target vase.

[0049] The obtained feature point cloud data is compared with the deviation factor obtained in the above step, and the deviation factor is subtracted from the feature point cloud data, thereby eliminating the noise caused by the reflection of the symmetry surface and obtaining the corrected point cloud data. At this point, the corrected point cloud data represents the true surface morphology of the target vase.

[0050] Fitting and anomaly detection of point cloud data:

[0051] Since the sub-block point cloud data of the good vase is used as the standard data, the corrected point cloud data of the target vase is compared with the first point cloud data of the sub-block of the good vase. Since the size and quantity of point cloud data in different sub-areas may be different, it is necessary to fit the smaller point cloud data with the larger point cloud data first. For example, assuming that the first point cloud data of a sub-block of the good vase is a 3×3 matrix, and the corrected point cloud data of the target vase is a 4×4 matrix, the 3×3 matrix needs to be fitted to the 4×4 matrix.

[0052] At this point, four fitting schemes can be generated, and the degree of fit of each scheme will be different. By calculating the degree of fit of these schemes, the scheme with the highest degree of fit is selected for further deviation calculation. If the scheme with the highest degree of fit is 99.8%, the fitting deviation is 0.02%. By comparing with the preset process design allowable range, if the fitting deviation is within the allowable range (for example, ±0.05%), the sub-area state is considered normal; if the fitting deviation exceeds the process allowable range (for example, ±0.01%), the sub-area state is considered abnormal.

[0053] According to the above steps, the abnormal state of each sub-region of the target vase can be obtained, and then the state of the entire vase can be determined.

[0054] Joint analysis and connectivity verification:

[0055] Considering that the cracks on vases are usually in a net-like or open shape, and the crack size is within a certain range, the abnormal state of each sub-region can be further analyzed jointly. If multiple adjacent sub-regions (for example, three sub-regions) are judged to be abnormal, and these abnormal sub-regions are adjacent in space, these sub-regions can be regarded as connected regions. By analyzing the shape and size of the connected region, it is determined whether it can accommodate the characteristic crack.

[0056] On this basis, the shape and size of the connected area are verified to determine whether the area conforms to the morphology of the characteristic crack. If the characteristic crack is in a mesh shape, the connected area should also be distributed in a mesh shape; if the crack is in an open shape, the shape of the connected area should conform to the open shape. By verifying the shape, size and matching degree of the connected area with the characteristic crack, the first verification and second verification steps are completed, thereby further confirming whether the vase has cracks.

[0057] Symmetry matching and abnormal area removal:

[0058] It is worth noting that since the deviation factor is based on the noise data generated by the good vase, in the actual detection process, when the abnormal sub-region of the target vase is symmetrical with the normal sub-region, there may still be a certain degree of noise influence. In particular, when the abnormal sub-region is symmetrical with the normal sub-region, the reflection noise of the abnormal sub-region is different from that of the normal sub-region, which will cause the symmetrical normal sub-region to be misjudged as abnormal during calculation.

[0059] Therefore, when making anomaly judgments, it is necessary to match the sub-regions symmetrically. If the abnormal sub-region matches the normal sub-region successfully, the abnormal sub-region with a smaller fitting deviation can be eliminated according to the size of the fitting deviation, while the abnormal sub-region with a larger fitting deviation can be retained. In this way, the misjudgment caused by noise can be effectively eliminated, thereby ensuring the accuracy of the abnormal judgment.

[0060] Therefore, the sub-regions are symmetrically matched. If the match is successful, the abnormal sub-region on the side with the smaller fitting deviation is eliminated, and the abnormal sub-region on the side with the larger fitting deviation is retained, and the retained sub-region is verified for the third time.

[0061] The sub-region that passes the first verification, the second verification and the third verification is taken as the crack region of the target vase, and the target vase completes the inspection.

[0062] like Figure 1 , a device crack detection system based on millimeter wave imaging radar, comprising:

[0063] The data module is used to collect point cloud data of a good workpiece and generate a point cloud model of the whole good workpiece based on the point cloud data through a millimeter wave imaging radar; wherein the good workpiece is structurally symmetrical;

[0064] The deviation module is used to divide the point cloud model of the good workpiece into sub-areas, generate the first point cloud data of each sub-area, and use the corresponding part of the point cloud data in the overall point cloud model of the good workpiece as the second point cloud data, calculate the difference based on the first point cloud data and the second point cloud data, and generate the deviation factor of each sub-area; wherein the deviation factor includes:

[0065] Step 1: Physically segment the good workpiece into several sub-blocks, each of which corresponds to a sub-region. The physical segmentation includes:

[0066] Determine the symmetry plane of good workpieces;

[0067] Based on the symmetry plane, the good workpiece is divided in the same way on both sides according to the principle of symmetry;

[0068] Step 2, generating first point cloud data for each sub-block by using a millimeter wave imaging radar;

[0069] Step 3, extracting partial point cloud data corresponding to each sub-block from the overall point cloud data of the good workpiece to generate second point cloud data;

[0070] Step 4, calculating the difference between the first point cloud data and the second point cloud data of each sub-region, and using the difference as the deviation factor of the corresponding sub-region;

[0071] The detection module is used to generate feature point cloud data of each sub-region of the target workpiece, and correct the feature point cloud data based on the deviation factor to generate corrected point cloud data; compare the corrected point cloud data with the second point cloud data to determine the abnormal state of the sub-region; wherein the good workpiece and the target workpiece are both the same workpiece and are symmetrical in structure;

[0072] The verification module is used to comprehensively analyze the abnormal status of all sub-areas, screen the abnormal sub-areas of the target workpiece based on the results of the comprehensive analysis, and determine the crack area of ​​the target workpiece.

[0073] In one embodiment of the present invention, correcting the feature point cloud data based on the deviation factor to generate corrected point cloud data includes:

[0074] Step 5: Generate feature point cloud data for each sub-region of the target workpiece, and calculate feature difference between the feature point cloud data of each sub-region and the corresponding deviation factor, and use the feature difference as the corrected point cloud data of the corresponding sub-region.

[0075] In one embodiment of the present invention, comparing the corrected point cloud data with the second point cloud data to determine the abnormal state of the sub-region includes:

[0076] Step 6, determining whether the number of point clouds in the corrected point cloud data of the i-th sub-region is greater than the number of point clouds in the second point cloud data; if greater, fitting the second point cloud data to match the number of point clouds in the corrected point cloud, specifically including:

[0077] Generate multiple fitting schemes according to the point cloud position information in the corrected point cloud data and the second point cloud data;

[0078] The degree of fit is calculated for each fitting scheme, and the calculation formula for the degree of fit of the fitting scheme is as follows:

[0079] ;

[0080] ;

[0081] ;

[0082] ;

[0083] in, represents the fitness of the fitting scheme, , and denotes weight parameters, and , and All are not 0, , and The sum is 1, , and They represent the first similarity, the second similarity and the third similarity of the ith point in the corrected point cloud data in terms of spatial coordinates, color information and reflection intensity, respectively. Indicates the number of point clouds of the second point cloud data, The index indicating the number of points in the second point cloud data. Represents the three-dimensional spatial coordinates of the i-th point in the corrected point cloud data, represents the three-dimensional spatial coordinates of the jth point in the second point cloud data, The bandwidth weight representing the similarity of spatial coordinates, Represents the color information of the i-th point in the corrected point cloud data, Represents the color information of the jth point in the second point cloud data, The bandwidth weight representing the similarity of color information, Represents the reflection intensity of the i-th point in the corrected point cloud data, Represents the reflection intensity of the jth point in the second point cloud data, The bandwidth weight representing the similarity of reflection intensity, represents the square of the Euclidean distance, represents the Gaussian kernel function, express Activation function;

[0084] Step 7: Select the solution with the highest fitting degree and calculate the fitting deviation between the corrected point cloud data and the second point cloud data. The calculation formula of the fitting deviation is as follows:

[0085] ;in, represents the fitting deviation;

[0086] Step 8: If the fitting deviation is greater than a preset fitting deviation threshold, the i-th sub-region is judged to be in an abnormal state; otherwise, the i-th sub-region is judged to be in a normal state.

[0087] In one embodiment of the present invention, the abnormal status of all sub-areas is comprehensively analyzed, including:

[0088] Step 9, obtaining shape and size information of cracks of the historical workpiece, and clustering the cracks based on the shape and size information to generate a number of characteristic cracks;

[0089] Step 10, performing connectivity analysis on adjacent abnormal state sub-regions to determine connected regions;

[0090] Step 11, performing a first verification based on the size of the characteristic crack and the size of the connected area; performing a second verification based on the shape of the characteristic crack and the shape of the connected area.

[0091] In one embodiment of the present invention, the comprehensive analysis of abnormal conditions of all sub-areas further includes:

[0092] Step 12, determining the sub-regions that are symmetrical about the symmetry plane, and judging that the abnormal states of the sub-regions that are symmetrical about the symmetry plane are the same, then the sub-regions that are symmetrical about the symmetry plane are thirdly verified.

[0093] In one embodiment of the present invention, screening the abnormal sub-region of the target workpiece based on the result of the comprehensive analysis to determine the crack region of the target workpiece includes:

[0094] Step 13, if the connected region passes the first verification, the subregion in the connected region passes the first screening; if the connected region passes the second verification, the subregion in the connected region passes the second screening;

[0095] Step 14, symmetrically matching the sub-regions that pass the first screening and the second screening about the symmetry plane; if the matching is successful, performing a third screening on the successfully matched sub-regions according to the fitting deviation, specifically including:

[0096] The fitting deviations of the sub-regions on both sides of the symmetry plane that have passed the first verification and the second verification and have been successfully matched are compared. If the fitting deviation of the a-th sub-region is greater than the fitting deviation of the b-th sub-region, the a-th sub-region is the crack region of the target workpiece.

[0097] The above describes an embodiment of the present embodiment, but the present embodiment is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the guidance of the present embodiment, ordinary technicians in this field can also make many forms, all of which are within the protection of the present embodiment.

Claims

1. A device crack detection system based on millimeter wave imaging radar, characterized in that: include: The data module is used to collect point cloud data of a good workpiece and generate a point cloud model of the whole good workpiece based on the point cloud data by using a millimeter wave imaging radar; wherein the good workpiece is structurally symmetrical; The deviation module is used to divide the point cloud model of the good workpiece into sub-areas, generate the first point cloud data of each sub-area, and use the corresponding part of the point cloud data in the overall point cloud model of the good workpiece as the second point cloud data, calculate the difference based on the first point cloud data and the second point cloud data, and generate the deviation factor of each sub-area; wherein the deviation factor includes: Step 1: Physically segment the good workpiece into several sub-blocks, each of which corresponds to a sub-region. The physical segmentation includes: Determine the symmetry plane of good workpieces; Based on the symmetry plane, the good workpiece is divided in the same way on both sides according to the principle of symmetry; Step 2, generating first point cloud data for each sub-block by using a millimeter wave imaging radar; Step 3, extracting partial point cloud data corresponding to each sub-block from the overall point cloud data of the good workpiece to generate second point cloud data; Step 4, calculating the difference between the first point cloud data and the second point cloud data of each sub-region, and using the difference as the deviation factor of the corresponding sub-region; The detection module is used to generate feature point cloud data of each sub-region of the target workpiece, and correct the feature point cloud data based on the deviation factor to generate corrected point cloud data; compare the corrected point cloud data with the second point cloud data to determine the abnormal state of the sub-region; wherein the good workpiece and the target workpiece are both the same workpiece and are symmetrical in structure; The verification module is used to comprehensively analyze the abnormal status of all sub-areas, screen the abnormal sub-areas of the target workpiece based on the results of the comprehensive analysis, and determine the crack area of ​​the target workpiece.

2. The device crack detection system based on millimeter wave imaging radar according to claim 1, characterized in that: Correct the feature point cloud data based on the deviation factor to generate corrected point cloud data, including: Step 5: Generate feature point cloud data for each sub-region of the target workpiece, and calculate feature difference between the feature point cloud data of each sub-region and the corresponding deviation factor, and use the feature difference as the corrected point cloud data of the corresponding sub-region.

3. The device crack detection system based on millimeter wave imaging radar according to claim 2, characterized in that: Comparing the corrected point cloud data with the second point cloud data to determine the abnormal state of the sub-area, including: Step 6, determining whether the number of point clouds in the corrected point cloud data of the i-th sub-region is greater than the number of point clouds in the second point cloud data; if greater, fitting the second point cloud data to match the number of point clouds in the corrected point cloud, specifically including: Generate multiple fitting schemes according to the point cloud position information in the corrected point cloud data and the second point cloud data; The degree of fit is calculated for each fitting scheme, and the calculation formula for the degree of fit of the fitting scheme is as follows: ； ； ； ； in, represents the fitness of the fitting scheme, , and denotes weight parameters, and , and All are not 0, , and The sum is 1, , and They represent the first similarity, the second similarity and the third similarity of the ith point in the corrected point cloud data in terms of spatial coordinates, color information and reflection intensity, respectively. Indicates the number of point clouds of the second point cloud data, The index indicating the number of points in the second point cloud data. Represents the three-dimensional spatial coordinates of the i-th point in the corrected point cloud data, represents the three-dimensional spatial coordinates of the jth point in the second point cloud data, The bandwidth weight representing the similarity of spatial coordinates, Represents the color information of the i-th point in the corrected point cloud data, Represents the color information of the jth point in the second point cloud data, The bandwidth weight representing the similarity of color information, Represents the reflection intensity of the i-th point in the corrected point cloud data, Represents the reflection intensity of the jth point in the second point cloud data, The bandwidth weight representing the similarity of reflection intensity, represents the square of the Euclidean distance, represents the Gaussian kernel function, express Activation function; Step 7: Select the solution with the highest fitting degree and calculate the fitting deviation between the corrected point cloud data and the second point cloud data. The calculation formula of the fitting deviation is as follows: ;in, represents the fitting deviation; Step 8: If the fitting deviation is greater than a preset fitting deviation threshold, the i-th sub-region is judged to be in an abnormal state; otherwise, the i-th sub-region is judged to be in a normal state.

4. The device crack detection system based on millimeter wave imaging radar according to claim 3 is characterized in that: Comprehensive analysis of abnormal status of all sub-areas, including: Step 9, obtaining shape and size information of cracks of the historical workpiece, and clustering the cracks based on the shape and size information to generate a number of characteristic cracks; Step 10, performing connectivity analysis on adjacent abnormal state sub-regions to determine connected regions; Step 11, performing a first verification based on the size of the characteristic crack and the size of the connected area; performing a second verification based on the shape of the characteristic crack and the shape of the connected area.

5. The device crack detection system based on millimeter wave imaging radar according to claim 4 is characterized in that: Comprehensive analysis of abnormal conditions in all sub-areas, including: Step 12, determining the sub-regions that are symmetrical about the symmetry plane, and judging that the abnormal states of the sub-regions that are symmetrical about the symmetry plane are the same, then the sub-regions that are symmetrical about the symmetry plane are thirdly verified.

6. The device crack detection system based on millimeter wave imaging radar according to claim 5, characterized in that: Based on the results of comprehensive analysis, the abnormal sub-areas of the target workpiece are screened to determine the crack area of ​​the target workpiece, including: Step 13, if the connected region passes the first verification, the subregion in the connected region passes the first screening; if the connected region passes the second verification, the subregion in the connected region passes the second screening; Step 14, performing symmetric matching on the sub-regions that pass the first screening and the second screening about the symmetry plane; if the matching is successful, performing a third screening on the successfully matched sub-regions according to the fitting deviation, specifically including: The fitting deviations of the sub-regions on both sides of the symmetry plane that have passed the first verification and the second verification and have been successfully matched are compared. If the fitting deviation of the a-th sub-region is greater than the fitting deviation of the b-th sub-region, the a-th sub-region is the crack region of the target workpiece.

Images