Automated detection method for automobile brake system blank parts

Abstract

The present application relates to a kind of automobile brake system blank parts automatic detection method, method includes: selecting the kind of part to be measured and setting determination threshold;Image is collected from detection line;To the image to be measured is filtered and binarized;Extension collection image;Edge extraction is obtained to the image after processing to the profile to be measured;To the profile to be measured is normalized;To the profile to be measured after processing is focused according to feature site respectively;To the profile after focusing, select corresponding algorithm to analyze;To the part to be measured, give final detection result;Detection data is automatically saved and provides data analysis and traceability basis;Finally realize the automatic detection of automobile brake system blank parts.The present application has the characteristics of objective detection process, detection data is automatically saved, and detection result is traceable, improves the accuracy and efficiency of detection.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of image recognition, and particularly relates to an automatic detection method for automobile brake system blank parts. BACKGROUND

[0002] The automobile brake system blank parts are directly related to the safety performance of the automobile. In the production process, the parts may be deformed in several positions due to collision, surface treatment, handling and other reasons, which may cause assembly abnormalities in the whole vehicle assembly process, even cause abnormal noise of the automobile, and cause safety hazards.

[0003] There are two traditional detection methods for the blank parts, namely the manual tool detection method and the machine vision method. The detection process of the manual tool detection method depends on human operation, and the detection process cannot be dataized, the parts cannot be traced, and the detection efficiency is low. This link becomes a bottleneck in the production enterprise in terms of detection effect, detection efficiency and daily labor cost. The machine vision method mainly analyzes the image of the overall contour of the part to achieve deformation detection effect. This method has high requirements for the placement position of the part and the detection environment, and the detection accuracy is unstable.

[0004] In view of the shortcomings of the traditional detection method, the application adopts automatic transmission and improved machine vision to provide the following solutions:

[0005] (1) Under the condition of relatively uniform placement position of the part and appropriate transmission beat, the detection efficiency is greatly improved compared with the manual tool detection method.

[0006] (2) Combined with the actual production environment, the light is supplemented and the camera frame rate is adjusted, the detection accuracy is close to the detection level of experienced workers by focusing analysis of each feature part of the blank part.

[0007] At the same time, the application has the characteristics of objective detection process, automatic saving of detection data and traceability of detection results. SUMMARY

[0008] The application provides a kind of high efficiency reliable automatic detection method for the low detection efficiency, false positive rate and other shortcomings of the traditional detection method of the automobile brake system blank parts.

[0009] The high efficiency reliable automatic detection method for automobile brake system blank parts provided by the application comprises the following steps:

[0010] S1: selecting the type of parts to be detected in the system, and setting the determination threshold;

[0011] S2: collecting images from the detection line as the images to be detected;

[0012] S3: filtering and binarization processing is performed on the image to be measured;

[0013] S4: if the profile to be detected is connected to the edge of the collected image, the collected image size is expanded with white as the background color to ensure the correctness of the profile to be measured;

[0014] S5: edge extraction is performed on the processed image to obtain the profile to be measured;

[0015] S6: the profile to be measured is normalized and then rotated to a similar angle through affine transformation;

[0016] S7: the processed profile to be measured is focused on the feature parts respectively;

[0017] S8: different feature analysis algorithms are selected for different feature parts of the profile to be measured;

[0018] Among them, the feature part 1 is the mounting base of the part, which is an irregular trapezoidal structure on the upper part of the part; the feature part 2 is the lower beam of the part, which is an arc structure on the lower part of the part; the feature part 3 is the positioning rod on both sides of the part, which is a straight line structure connecting the lower left side and the lower right side of the part; there are three kinds of feature analysis algorithms for the feature parts of the automobile brake system blank part, and the feature analysis algorithms are as follows:

[0019] S81: algorithm 1 is used for feature part 1 analysis, angle point detection is performed on the profile, noise removal algorithm is used to select the angle points to be calculated, and the difference between the feature areas is determined by the angle points;

[0020] S82: algorithm 2 is used for feature part 2 analysis, convex hull detection is performed on the profile, and the feature deformation degree is determined by the feature quantity of the convex hull region;

[0021] S83: algorithm 3 is used for feature part 3 analysis, Hough transform straight line detection is performed on the profile, weighted detection points are obtained through Hough transform straight line, weighted slope comparison is performed, and the feature symmetry degree is determined;

[0022] S9: according to the type of the part to be measured, the determination results of S81, S82 and S83 are combined to give the final detection result of the part to be measured;

[0023] S10: the detection data is automatically saved to provide data analysis and traceability basis.

[0024] In step S4, edge scanning is performed on the collected image after binarization, and if the values of the image edges are inconsistent, it is determined that the profile to be detected is connected to the edge of the collected image.

[0025] The step S7 is to cut the normalized part contour by feature region, and then extract the region contour of the cut contour image to obtain the feature region contour for further analysis.

[0026] In the step S81, for the feature region 1, four corner points A, B, C and D are obtained by corner point detection and denoising. The following calculation is performed on the contour curve between A and B:

[0027]

[0028]

[0029] wherein, represents the area on the side of C point of the structure with the CD median line as the boundary; represents the area on the side of D point of the structure with the CD median line as the boundary; by comparing with the set threshold value, the difference value between the areas of the feature region 1 is determined.

[0030] In the step S82, for the feature region 2, the contour bottom A and B points are obtained by convex hull detection, the slope k of the line connecting A and B is calculated, and the convex defect analysis is performed on the region above the line connecting A and B. The formula is as follows:

[0031]

[0032] wherein, Si is the distance from point i on the contour to the line AB; by comparing with the set threshold value, the deformation degree of the feature region 2 is determined.

[0033] In the step S83, for the feature region 3, based on the adjusted Hough transform, the following calculation is performed on every two points of the M points on one side of the contour:

[0034]

[0035]

[0036] The distance between each point and other points is counted , and the number of elements less than the threshold value is counted . wherein, is the distance between point i and point j;

[0037] All points greater than the threshold value are found out , and the number of points is counted Size, to Sort, the first Value is recorded as , the corresponding Recorded as , and so on, the selected The following calculations are made:

[0038]

[0039] The same calculation is made on the other side of the profile, and ;

[0040] Wherein, The integrated slope of the slope of the plurality of transformed straight lines; by Comparison with the set threshold value, the symmetry degree determination result of the feature part 3 is given.

[0041] In step S9, for each algorithm in step S8, for different parts to be tested, as long as the feature part meets the algorithm requirements, the threshold value can be preset, and different algorithm combinations are used for detection.

[0042] The beneficial effects of the present application are as follows:

[0043] (1) The present application adopts automatic transmission and machine vision for part detection, which greatly improves the detection efficiency compared with manual detection, under the condition that the placement position of the part is relatively uniform and the transmission beat is appropriate.

[0044] (2) The present application focuses on each feature part of the part to be tested and designs corresponding feature analysis algorithms, so that the detection result in complex environment is more stable and accurate.

[0045] At the same time, the present application has the characteristics of objective detection process, automatic saving of detection data and traceability of detection results. BRIEF DESCRIPTION OF DRAWINGS

[0046] Figure 1 It is a flowchart of the automatic detection method of the automobile brake system blank part in the embodiment of the present application;

[0047] Figure 2 It is a legend of a part to be tested in the embodiment of the present application;

[0048] Figure 3 It is a profile of the feature part 1 after focusing of the part to be tested in the embodiment of the present application;

[0049] Figure 4 It is a profile of the feature part 2 after focusing of the part to be tested in the embodiment of the present application;

[0050] Figure 5The profile of the focused feature part 3 of the part to be detected in the embodiment of the present application;

[0051] Figure 6 The schematic diagram of the corner point detection in the embodiment of the present application;

[0052] Figure 7 The schematic diagram of the convex hull detection in the embodiment of the present application. DETAILED DESCRIPTION

[0053] The purpose and technical solution of the present application are further described in detail below in combination with the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and do not limit the present application.

[0054] As shown in the figure, the embodiment provides an automatic detection method for automobile brake system blank parts, which is described in detail as follows: Figure 1

[0055] S1: The system is powered on, the type of the part to be detected is selected in the detection setting, and the determination threshold is set. After the detection line is started, the part to be detected is conveyed according to the rhythm;

[0056] S2: The complete part-to-be-detected image is collected from the detection line by a CCD industrial camera or an image collection device, as a to-be-detected image;

[0057] As shown in the figure, the collection angle can collect images from two straight directions of top-down and left-to-right, completely collect the profile that needs to be analyzed for features, and at the same time, additional light can be added according to the environment, and the camera frame rate can be adjusted, etc. Figure 2

[0058] S3: The to-be-detected image is filtered and binarized to highlight the profile features of the part to be detected as much as possible;

[0059] S4: If the profile to be detected and analyzed is connected with the edge of the collected image, the collected image size is expanded with white as the background color to ensure the correctness of the extraction of the to-be-detected profile;

[0060] If the values of the image edges are inconsistent, it is determined that the profile to be detected and analyzed is connected with the edge of the collected image. After the edge of the collected image is expanded, the complete to-be-detected profile can be extracted in step S5;

[0061] S5: The edge of the processed image is extracted to obtain the to-be-detected profile;

[0062] The Canny operator is used to extract the edge of the image. In all the extracted profiles, the largest profile is selected as the outer profile of the part to be detected for the next step of analysis;

[0063] ​​S6: The measured profile is normalized, and then rotated to a similar angle through affine transformation, so as to further focus on the feature parts;

[0064] S7: The measured profile is focused on the feature parts according to the type of the measured part selected in S1, and the process is as follows:

[0065] (1) According to the type of the measured part, the corresponding feature part data is obtained, including the feature part area, the feature part name, the analysis algorithm, the determination threshold, etc.

[0066] (2) The normalized measured profile is regionally cropped according to the feature part area;

[0067] (3) The cropped profile image is subjected to regional contour extraction again to obtain the feature part contour for further analysis;

[0068] As shown in Figure 3 , Figure 4 , Figure 5 , the feature parts of the measured part in this embodiment have three parts, corresponding to feature part 1, feature part 2, and feature part 3, respectively.

[0069] S8: For different feature parts of the measured profile, the corresponding feature analysis algorithm is selected;

[0070] Among them, feature part 1 is the mounting base of the part, which is an irregular trapezoidal structure on the upper part of the part; feature part 2 is the lower beam of the part, which is an arc structure on the lower part of the part; feature part 3 is the positioning rod on both sides of the part, which is a straight line structure connecting the lower left side and the lower right side of the part; for the feature parts of the automobile brake system blank part, there are three kinds of feature analysis algorithms respectively, and the feature analysis algorithms are as follows:

[0071] S81: Algorithm 1 is used for feature part 1 to analyze the contour and detect the corner points, select the corner points that need to be calculated through the denoising algorithm, and combine the difference between the corner point determination feature area;

[0072] Among them, for feature part 1, four corner points A, B, C, and D are obtained through corner point detection and denoising, as shown in Figure 6 The following calculations are performed on the contour curve between A and B:

[0073]

[0074]

[0075] Among them, represents the area on the side of the structure close to point C with CD as the dividing line; represents the area on the side of the structure close to point D with CD as the dividing line; and The comparison with the set threshold value gives a difference value determination result between the feature part 1 areas;

[0076] S82: The feature part 2 uses algorithm 2 analysis, performs convex hull detection on the contour, and determines the feature deformation degree through the feature quantity of the convex hull region;

[0077] For the feature part 2, two points A and B at the bottom of the contour are obtained through the convex hull detection, as shown in the formula below is used to calculate the area above the line connecting the two points A and B, taking A as the origin, and the formula is as follows: Figure 7

[0078]

[0079] Where Si is the distance from point i on the contour to the line AB; through The comparison with the set threshold value gives a feature part 2 deformation degree determination result;

[0080] S83: The feature part 3 uses algorithm 3 analysis, performs Hough transform straight line detection on the contour, obtains weighted detection points through the Hough transform straight line, and performs weighted slope comparison to determine the feature symmetry degree;

[0081] For the feature part 3, based on the adjusted Hough transform, M points on one side of the contour are calculated as follows:

[0082]

[0083]

[0084] The distance between each point and other points is collected , and the number of elements less than the threshold value is counted , where is the distance between point i and point j;

[0085] All points greater than the threshold value are found , sorted by size, and the first value is recorded as , and the corresponding is recorded as , and so on. The selected is calculated as follows:

[0086]

[0087] ​​​The same calculation is performed on the other side of the profile, and the result is ;

[0088] wherein, is the comprehensive slope of the slopes of the plurality of transformed straight lines, and is obtained by The symmetry degree determination result of the feature part 3 is given by comparison with the set threshold value;

[0089] For the three feature parts of the part to be measured in the embodiment, feature analysis is performed on S81, S82 and S83, respectively, and the analysis process is as follows:

[0090] (1) The three feature parts focusing profiles, i.e., feature part 1, feature part 2 and feature part 3, are obtained by S7;

[0091] (2) The difference value between areas of the feature part 1 is calculated by S81, and the calculation result is compared with the determination threshold value set by S1 to obtain the determination result a;

[0092] (3) The deformation degree of the feature part 2 is calculated by S82, and the calculation result is compared with the determination threshold value set by S1 to obtain the determination result b;

[0093] (4) The symmetry degree of the feature part 3 is calculated by S83, and the calculation result is compared with the determination threshold value set by S1 to obtain the determination result c;

[0094] S9: The determination results a, b and c are comprehensively determined, and the final detection result of the part to be measured is given;

[0095] S10: The detection data is automatically saved, and the system provides data analysis and traceability basis.

[0096] The final detection result of the part to be measured is displayed in the form of pictures and texts in the system. In the case of unqualified detection, the system can list the detailed information of each unqualified feature part to help the detection personnel analyze and decide, so as to ensure that the detection result is more reasonable.

[0097] The above is only a specific embodiment of the present application, and any feature or feature analysis method disclosed in the description can be combined in any way, except for mutually exclusive features or steps.

Claims

1. A method for automated inspection of automotive brake system green parts, characterized by, It comprises the following steps: S1: selecting the type of the part to be tested in the system and setting the determination threshold; S2: collecting an image from the detection line as the image to be tested; S3: filtering and binarizing the image to be tested; S4: if the contour to be detected is connected with the edge of the collected image, expanding the size of the collected image with white as the background color to ensure the correctness of the extraction of the contour to be tested; S5: performing edge extraction on the processed image to obtain the contour to be tested; S6: performing normalization on the contour to be tested and rotating the contour to a similar angle through affine transformation; S7: focusing on the processed contour to be tested according to the feature parts; S8: selecting a corresponding feature analysis algorithm for each feature part of the contour to be tested; Among them, the feature part 1 is the mounting base of the part, which is an irregular trapezoidal structure on the upper part of the part; the feature part 2 is the lower beam of the part, which is an arc-shaped structure on the lower part of the part; the feature part 3 is the positioning rod on both sides of the part, which is a straight line structure connecting the lower left side and the lower right side of the part; there are three kinds of feature analysis algorithms for the feature parts of the automobile brake system blank parts, and the feature analysis algorithms are as follows: S81: using algorithm 1 to analyze the feature part 1, performing corner point detection on the contour, selecting the corner points to be calculated through the denoising algorithm, and combining the difference between the feature areas to determine the corner points; S82: using algorithm 2 to analyze the feature part 2, performing convex hull detection on the contour, and determining the feature deformation degree through the feature quantity of the convex hull region; S83: using algorithm 3 to analyze the feature part 3, performing Hough transform straight line detection on the contour, obtaining weighted detection points through Hough transform straight line, and comparing the weighted slopes to determine the feature symmetry degree; S9: according to the type of the part to be tested, comprehensively determining the final detection result of the part to be tested based on the determination results of S81, S82 and S83; S10: automatically saving the detection data to provide data analysis and traceability basis; The step S4 further comprises: performing edge scanning on the binarized collected image, and if the values of the image edges are inconsistent, it is determined that the contour to be detected is connected with the edge of the collected image; The step S7 further comprises: focusing on the contour according to the feature parts means that the normalized part contour is cut according to the feature part region, and the cut contour image is further subjected to region contour extraction to obtain the feature part contour for further analysis; The step S9 further comprises: for each algorithm in step S8, as long as the feature part meets the algorithm requirement, a threshold value can be preset, and different algorithm combinations can be used for detection.

Images