A tire line detection method based on machine vision

Abstract

The present invention provides a tire line detection method based on machine vision, which solves the technical problem that the existing tire color line drawing lacks automated detection, and the steps mainly include: S1, image shooting; S2, image preprocessing; S3, image analysis, matching and obtaining the color identification and processing position identification of each color line; S4, obtaining the process identification information of the color line; S5, identification information comparison; S6, sending instructions, when the comparison result in S5 is abnormal, sending preset instructions to the alarm system and MES system. The present invention realizes automatic detection of line quality by online shooting of tread color lines and combines machine vision applications, improves production efficiency, reduces the defective rate, and thus reduces cost expenditure, and can be widely used in the field of tire processing technology.

Description

Technical Field

[0001] The invention relates to the technical field of tire processing, and in particular to a tire line detection method based on machine vision. Background Art

[0002] When tires are produced, multiple specifications and multiple patterns are often produced at the same time. In addition, there are also tires with the same specifications and patterns but different rubber formulas that need to be distinguished.

[0003] In order to facilitate identification and avoid misuse of tread rubber in the above situation, two forms of marking are generally adopted: one way is to attach a label, usually a barcode, and make a judgment by scanning the label; the other way is to add different colored lines to the tire tread rubber when the tread rubber is compounded to distinguish specifications, patterns and other information.

[0004] The second marking method is more commonly used. The colored line formed by it is called the tire color mark line, also commonly known as the landing line. The color mark line allows the tread to be easily identified by comparison during the molding and vulcanization process, thereby achieving an error-proofing function.

[0005] The existing color code line drawing method is to draw the line after the tire crown is extruded, and most of the line drawing work is manually controlled. Although the workload of manually operating the line drawing mechanism is not large, it may also cause problems such as material errors or inaccurate mechanism adjustment due to manual operation, which will lead to incorrect tire line color type and spacing, resulting in material returns in subsequent processes.

[0006] In summary, it is also necessary to add an automatic detection mechanism during the conveying process of the semi-finished tread, so as to automatically detect the quality of the color mark line drawing online and report errors in time to correct production. Summary of the invention

[0007] The purpose of the present invention is to solve the deficiencies of the above-mentioned technology and provide an automatic tire line detection method based on machine vision with low hardware cost.

[0008] To this end, the present invention provides a tire line detection method based on machine vision, comprising the following steps:

[0009] S1, image capture, using a camera to continuously capture the tread color mark line on the conveyor line to obtain a color mark line image;

[0010] S2, image preprocessing, preprocessing the color mark line image and obtaining a color image of the color mark line after color enhancement;

[0011] S3, image analysis, extracting the features of the image obtained in S2, matching and obtaining the color identification of each color marking line processing; obtaining the spacing of each color marking line according to geometric conversion, and recording the processing position identification of each color marking line;

[0012] S4, obtain the color mark line identification information, connect to the PLM system, and request the database to obtain the process color identification and process position identification of the current processing color mark line;

[0013] S5, comparing identification information, continuously comparing the processing color identification and the processing position identification with the process color identification and the process position identification;

[0014] S6. Sending instructions. When the comparison result in S5 is abnormal, sending preset instructions to the alarm system and the MES system.

[0015] Preferably, the camera in S1 is a color smart camera, and in the continuous shooting, N photos are taken within a unit time T as a group, and the S2 is sent to the subsequent steps;

[0016] T is greater than 0s, and N is a positive integer.

[0017] Preferably, the pretreatment method in S2 includes:

[0018] S21, denoising the image in S1 through a filter;

[0019] S22. The denoised image is converted into an HSV color model to obtain a color enhanced image.

[0020] Preferably, the method for obtaining the processing color identification in S3 includes:

[0021] S31, converting the color enhanced image into a grayscale image, and processing the grayscale image by Gaussian blur;

[0022] S32, binarizing the image obtained in S31, and extracting the contours of each color mark line in the image;

[0023] S33, extracting feature points from the area within the contour line and marking the positions of the feature points; extracting color components of pixel points in the color enhanced image according to the positions of the feature points;

[0024] S34, judging the processing color identification of the color mark line according to the color component in S33.

[0025] Preferably, the feature point extraction method in S33 includes:

[0026] The area within the contour line is segmented, and points are selected in each segmented block in a sequential or random manner; the set of points within the contour line is the point set of the feature points.

[0027] Preferably, the determination method in S34 includes:

[0028] The extracted standard deviation and variance of the pixel color component are brought into the machine learning model to obtain the pixel color recognition result, and marked as the processing color identifier.

[0029] Preferably, the method for obtaining the processing position identification in S3 includes:

[0030] S35, fitting the contour area of ​​each color-coded line segment to form a center line;

[0031] S36, calculating the distance between the center lines of two adjacent color mark lines according to the geometric conversion relationship, and marking it as the processing position mark.

[0032] Preferably, the centerline acquisition method of S35 includes:

[0033] A plurality of dividing lines are arranged at equal intervals in the contour area, and the dividing lines are arranged in a direction perpendicular to the tread conveying direction; each of the dividing lines intersects with the contour line to generate two intersection points, and the midpoint of the two intersection points is the centerline point; the centerline point set in the contour area is fitted by the least squares method to form the centerline.

[0034] Preferably, in S5, the processing color mark and processing position mark of each color mark line in a single image or a single group of images are compared with the process color mark and process position mark in sequence;

[0035] When the color identifications are different or the position identification difference is greater than the threshold, continue with step S6;

[0036] When the color identifications are consistent and the position identification difference is less than the threshold, the subsequent steps are stopped and the cycle is restarted from the S1 step.

[0037] Preferably, the tread conveying line is further provided with a line drawing mechanism downstream along the conveying direction; and S6 further includes:

[0038] Send an action instruction to the line drawing mechanism, calculate the line drawing position according to the conveying speed, and mark the points or sections with abnormal comparison results in S5.

[0039] The beneficial effects of the present invention are:

[0040] The present invention realizes automatic detection of line drawing quality by online shooting of tread color marking lines combined with machine vision application, improves production efficiency, reduces defective product rate, and thus reduces cost expenditure. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 is a flow chart of the present invention;

[0042] Figure 2 It is a grayscale image processing comparison diagram of the present invention;

[0043] Figure 3 It is a comparison diagram after the binary image processing of the present invention;

[0044] Figure 4 It is a schematic diagram of the contour center line fitting of the present invention. DETAILED DESCRIPTION

[0045] The present invention is further described below in conjunction with the accompanying drawings and specific embodiments to help understand the content of the present invention. The methods used in the present invention are all conventional methods unless otherwise specified; the raw materials and devices used are all conventional commercial products unless otherwise specified.

[0046] It should be noted that the embodiment of the present invention is aimed at the detection of line marking on the tread of a semi-finished product, wherein the tread is continuously conveyed on a belt conveyor after extrusion, and a line marking device is set up upstream of the conveyor. Correspondingly, in this embodiment, a high-resolution color smart camera and a fill light component are set up downstream of the line marking device as shooting hardware; the shooting hardware is specifically suspended above the conveyor belt through a bracket.

[0047] like Figure 1 As shown, the present invention provides a tire line detection method based on machine vision, which mainly includes the following steps:

[0048] S1. Image capture: Continuously capture the tread color mark line on the conveyor line with a camera to obtain a color mark line image; preferably, the camera is a high-resolution color smart camera as described above, and the continuous capture takes 2 seconds as a unit time, and 10 tread photos are taken as a group during the period, and further sent to the image processing module in the system.

[0049] S2, image preprocessing, preprocessing the image of the tread photo in the image processing module, and obtaining a color image of the color mark line after color enhancement; preferably, in the actual shooting process, image distortion may occur due to camera placement or installation problems, so on this basis, distortion correction is also required before the image can be used. Since image distortion correction is a routine operation, it will not be described here.

[0050] The specific processing methods are:

[0051] S21, firstly, denoising the image by filtering, wherein the filter may be selected from a mean filter, a median filter or a bilateral filter, so as to improve the image quality;

[0052] S22. Considering color judgment, the denoised image is converted into an HSV color model to obtain a color enhanced image.

[0053] S3, image analysis, extracting the features of the image obtained in S2, matching and obtaining the color identification of each color marking line processing; obtaining the spacing of each color marking line according to geometric conversion, and recording the processing position identification of each color marking line;

[0054] Preferably, the method for obtaining the processing color identification specifically includes:

[0055] S31, convert the color enhanced image in the previous step into a grayscale image, and process the grayscale image by Gaussian blur; the processing method is mainly to consider that the tire tread as the background color is darker, and the color of the line is not all bright, and the workshop is not bright. Even if the fill light is added, the color mark line will be unclear, so the image needs to be processed, refer to Figure 2 As shown in Figure 2 a is the grayscale image of the unprocessed color scale line area. Figure 2 b is the grayscale image of the processed color-coded line area. It can be clearly seen that the edge of the line segment can be displayed, which is beneficial for subsequent processing;

[0056] S32, binarizing the image obtained above, as follows: Figure 3 As shown, if the previous step is not performed, Figure 3 c image, it is obvious that some color scale lines cannot be displayed correctly, so after the previous step, we can get Figure 3 d, and on this basis, the contour extraction algorithm can be used to extract the contours of the color marking lines in the image;

[0057] S33, Figure 3 d as an example, since there are fewer white pixels in the entire picture, the set of white pixels can be determined as the color mark line, so the feature points can be simply extracted from the area within the contour line and the feature points can be marked; further, the specific extraction method is to segment the area within the contour line into Z segments, each segment block has M×M pixels, and the segments containing black pixels are removed. Then, points are randomly selected in each segment block, and these points are used as feature points. The set of points in each contour line is the point set of feature points;

[0058] According to the position of the feature points in the point set, the color components of the corresponding pixels are extracted from the previously obtained color enhanced image;

[0059] S34, judging the processing color identification of the color-coded line according to the color component. Furthermore, since there are many types of line drawing colors, in order to increase the recognition accuracy, the color of each color-coded line cannot be simply determined according to the color component. Therefore, the standard deviation and variance of the extracted pixel color component are brought into the machine learning model. Specifically, the pixel color recognition result can be obtained by the SVM model (support vector machine) trained with the standard color, and the processing color identification is marked for each color-coded line in the image.

[0060] Furthermore, the method for obtaining the processing position identification includes:

[0061] S35, using the contour lines of the aforementioned color-coded line segments, fitting to form a center line in the region of each contour line; the specific method is combined with Figure 4 As shown, multiple dividing lines are set at equal intervals in each contour area, wherein the dividing line setting direction is perpendicular to the tread conveying direction. Taking the example dividing line in the figure as an example, it intersects with the contour line to generate two intersection points a1 and a2, and the midpoint a0 of the two intersection points is taken as the centerline point on the dividing line, and b0 is obtained in sequence accordingly... A series of centerline points form a collection, and the centerline point set in the contour area is fitted by the least squares method to form the first centerline l1. The corresponding adjacent color mark line contour area also calculates the second centerline l2. Since the two center lines are not necessarily strictly parallel, the total number of pixels between the two center lines is counted and the average is taken, so that the number of pixels of the lateral spacing between the two center lines can be obtained;

[0062] S36. Based on the camera installation position, camera parameters, center line spacing D (number of horizontal pixels) and other information, the actual spacing between the center lines of two adjacent color marking lines can be calculated by projection method and marked as a processing position mark.

[0063] S4. Obtain the color mark line identification information, connect to the PLM system (product life cycle management system), and request the database to obtain the process color identification and process position identification of the current processing color mark line.

[0064] S5, comparing identification information, continuously comparing the processing color identification and processing position identification with the process color identification and process position identification; preferably, in this step, the processing color identification and processing position identification of each color mark line in a single group of images are compared with the corresponding process color identification and process position identification in sequence;

[0065] When the color identification is different or the position identification difference is greater than the threshold, it means that the photographed section color mark line is abnormal and there is a problem in drawing the line, so continue to step S6;

[0066] When the color identification is consistent and the position identification difference is less than the threshold, it means that there is no abnormality in the photographed segment color mark line and the line drawing is correct, so the subsequent steps are stopped and the cycle is restarted from step S1.

[0067] S6, send instructions. When the comparison result in S5 is abnormal, send preset instructions to the alarm system and MES system (manufacturing execution system); preferably, the content is to send the abnormal message to the MES system, and the MES system will continuously alarm the on-site machine and record the alarm log until the abnormality is resolved. According to the needs of on-site management and control, it can perform automatic control such as alarm, prohibition of output, and shutdown. Further, considering that the line drawing problem is not necessarily a continuous problem, it is necessary to mark the erroneous points or sections. Therefore, in this embodiment, a separate line drawing mechanism is also provided downstream of the tread conveyor line along the conveying direction. When a line drawing abnormality is detected, the system can calculate when the problem position moves to the line drawing mechanism according to the conveying speed, and send an action instruction to the line drawing mechanism. Before shutting down, the points or sections with abnormal comparison results in S5 are marked with lines, so that the operator can make on-site judgments.

[0068] In the description of the present invention, it is necessary to understand that the terms "left", "right", "up", "down", "top", "bottom", "front", "back", "inside", "outside", "back", "middle", etc., indicating the orientation or position relationship are based on the orientation or position relationship shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on the present invention.

[0069] However, what is described above is only a specific embodiment of the present invention and should not be used to limit the scope of implementation of the present invention. Therefore, the replacement of equivalent components, or equivalent changes and modifications made according to the scope of protection of the present invention should still fall within the scope covered by the claims of the present invention.

Claims

1. A tire line detection method based on machine vision, characterized in that: The following steps are involved: S1, image capture, using a camera to continuously capture the tread color mark line on the conveyor line to obtain a color mark line image; S2, image preprocessing, preprocessing the color mark line image and obtaining a color image of the color mark line after color enhancement; S3, image analysis, extracting the features of the image obtained in S2, matching and obtaining the color identification of each color marking line processing; obtaining the spacing of each color marking line according to geometric conversion, and recording the processing position identification of each color marking line; S4, obtain the color mark line identification information, connect to the PLM system, and request the database to obtain the process color identification and process position identification of the current processing color mark line; S5, comparing identification information, continuously comparing the processing color identification and the processing position identification with the process color identification and the process position identification; S6, sending instructions, when the result of the comparison in S5 is abnormal, sending preset instructions to the alarm system and the MES system; The pre-processing method in S2 includes: S21, denoising the image in S1 through a filter; S22, converting the denoised image into an HSV color model to obtain a color enhanced image; The method for obtaining the processing color identification in S3 includes: S31, converting the color enhanced image into a grayscale image, and processing the grayscale image by Gaussian blur; S32, binarizing the image obtained in S31, and extracting the contours of each color mark line in the image; S33, extracting feature points from the area within the contour line and marking the positions of the feature points; extracting color components of pixel points in the color enhanced image according to the positions of the feature points; S34, judging the processing color identification of the color mark line according to the color component in S33; The method for obtaining the processing position identification in S3 includes: S35, fitting the contour area of ​​each color-coded line segment to form a center line; S36, calculating the distance between the center lines of two adjacent color mark lines according to the projection method, and marking it as the processing position mark.

2. The tire line detection method based on machine vision according to claim 1, characterized in that: The camera in S1 is a color smart camera, and in the continuous shooting, N photos are taken within a unit time T as a group, and sent to S2 for subsequent steps; T is greater than 0s, and N is a positive integer.

3. The tire line detection method based on machine vision according to claim 1, characterized in that: The feature point extraction method in S33 includes: The area within the contour line is segmented, and points are selected in each segmented block in a sequential or random manner; the set of points within the contour line is the point set of the feature points.

4. The tire line detection method based on machine vision according to claim 1, characterized in that: The determination method in S34 includes: The extracted standard deviation and variance of the pixel color component are brought into the machine learning model to obtain the pixel color recognition result, and marked as the processing color identifier.

5. The tire line detection method based on machine vision according to claim 1, characterized in that: The centerline acquisition method of S35 includes: A plurality of dividing lines are arranged at equal intervals in the contour area, and the dividing lines are arranged in a direction perpendicular to the tread conveying direction; each of the dividing lines intersects with the contour line to generate two intersection points, and the midpoint of the two intersection points is the centerline point; the centerline point set in the contour area is fitted by the least squares method to form the centerline.

6. The tire line detection method based on machine vision according to claim 1, characterized in that: In S5, the processing color mark and processing position mark of each color mark line in a single image or a single group of images are sequentially compared with the process color mark and process position mark; When the color identifications are different or the position identification difference is greater than the threshold, continue with step S6; When the color identifications are consistent and the position identification difference is less than the threshold, the subsequent steps are stopped and the cycle is restarted from step S1.

7. The tire line detection method based on machine vision according to claim 1, characterized in that: The conveying line is also provided with a line drawing mechanism downstream along the conveying direction; the S6 also includes: Send an action instruction to the line drawing mechanism, calculate the line drawing position according to the conveying speed, and mark the points or sections with abnormal comparison results in S5.

Images