An intelligent identification method for automobile connecting plate processing

Abstract

The present invention belongs to the field of image processing technology, and specifically relates to an intelligent recognition method for processing automotive connector plates. The method comprises: acquiring a surface image of the corner position of a processed automotive connector plate, performing threshold segmentation on the surface image using a target low threshold and a target high threshold, determining a guide line contour from the obtained low-binary image, calculating a first probability of a wrinkling defect in the connector plate based on the curvature variation of contour points on the guide line contour, determining an intersection contour from the obtained high-binary image, calculating a second probability of a wrinkling defect in the connector plate based on the differences in chain code sets of each contour branch point in different directions on the intersection contour, and determining whether the automotive connector plate has a wrinkling defect based on the first and second probabilities. The present invention can more accurately detect wrinkling defects, improving product quality and production efficiency of automotive connector plates.

Description

Technical Field

[0001] The present invention relates to the field of image processing technology, and more particularly to an intelligent recognition method for processing automobile connecting plates. Background Art

[0002] Floor beam connectors are typically installed at the front, center, and rear of the chassis. They connect to other structural components, such as floor panels, longitudinal beams, and floor beams, to form the chassis structure and are key connectors for the vehicle's body frame.

[0003] The connecting plate is mainly made of high-strength steel. Its bottom flange surface, end vertical surface and top connection surface are connected to the floor, side wall and top surface of the body longitudinal beam respectively through welding points.

[0004] To improve load-bearing capacity, impact resistance, and torsional rigidity, and thereby enhance vehicle body safety and chassis stability, the product structural design of floor cross member connecting plates is increasingly leaning toward closed-type connection surfaces to prevent tearing at the notches during collisions. However, this design also increases the difficulty of stamping.

[0005] When the floor beam connecting plate is produced by forming process, the blank at the three-sided intersection position is squeezed and deformed and bulges at the corner position. After the mold is closed, the excess material cannot be eliminated, resulting in wrinkles at the corner position, which in turn affects the

[0006] During the forming process for the floor beam connecting plate, the blank is extruded and deformed at the intersection of three sides, resulting in bulges at the corners; after the mold is closed, the excess material cannot be effectively removed, resulting in wrinkling defects at the corners.

[0007] Wrinkling defects will weaken the structural strength and rigidity of the connecting plate, reduce its fatigue resistance, and thus affect the overall stability and safety of the vehicle body; at the same time, wrinkling problems will also lead to unstable quality of the produced connecting plates, increase the defective rate, reduce production efficiency, and bring additional costs and quality control challenges to the manufacturing process. Summary of the Invention

[0008] In order to solve the above-mentioned technical problem of how to judge whether a processed automobile connecting plate has a wrinkle defect, the present invention provides an intelligent recognition method for automobile connecting plate processing, comprising: collecting a surface image of the corner position of the processed automobile connecting plate; updating the initial low threshold multiple times, and determining a target low threshold based on the changes in the number and connectivity of black connected domains in the binary image corresponding to the low threshold before and after each update; performing threshold segmentation on the surface image using the target low threshold, and extracting the skeleton of the black connected domain with the largest area in the obtained low binary image to obtain a guide line contour; and calculating a first probability of the presence of a wrinkle defect in the connecting plate based on the changes in the curvature of the contour points on the guide line contour. rate; the initial high threshold is updated multiple times, and the target high threshold is determined based on the changes in the number and connectivity of white connected domains in the binary image corresponding to the high threshold before and after each update; the surface image is threshold segmented using the target high threshold, and the skeleton of the white connected domain in the obtained high binary image is extracted to obtain the intersection angle contour of the surface image; contour branch points containing multiple neighborhood points within 8 neighborhoods on the intersection angle contour are obtained, and chain codes are extracted from the intersection angle contour based on the contour branch points. The second probability of the presence of a wrinkle defect in the connecting plate is calculated based on the differences in the chain code sets of the contour branch points in different directions; and the presence of a wrinkle defect in the processed automobile connecting plate is determined based on the first and second probabilities.

[0009] Preferably, the method of updating the initial low threshold multiple times and determining the target low threshold according to the changes in the number and connectivity of black connected domains in the binary image corresponding to the low threshold before and after each update includes: updating the initial low threshold multiple times, performing threshold segmentation on the surface image according to the low threshold after each update to obtain the binary image corresponding to the low threshold after each update; performing connected domain analysis on all black pixels in the binary image corresponding to the low threshold after each update to obtain multiple black connected domains; and The two black connected domains with the largest area in the binary image corresponding to the low threshold after the update are When the binary image corresponding to the low threshold after the first update is merged into a black connected domain, stop updating the initial low threshold and set the The updated low threshold is used as the target low threshold.

[0010] Preferably, the method for obtaining the curvature at the contour point includes: for any contour point on the guide line contour, performing linear fitting on the positions of the contour point and the six adjacent points of the contour point, and using the slope of the straight line obtained by fitting as the curvature at the contour point.

[0011] Preferably, the calculating the first probability of the existence of a wrinkling defect in the connecting plate according to the curvature change of the contour points on the guide line contour includes: Where, is the first probability of the existence of wrinkling defects in the connecting plate, is the natural exponential function, For the first The curvature of the contour points, For the first The curvature of the contour points, The number of all contour points contained in the guide line contour.

[0012] Preferably, the method of updating the initial high threshold multiple times and determining the target high threshold according to the changes in the number and connectivity of white connected domains in the binary image corresponding to the high threshold before and after each update includes: updating the initial high threshold multiple times, performing threshold segmentation on the surface image according to the high threshold after each update to obtain the binary image corresponding to the high threshold after each update; performing connected domain analysis on all white pixels in the binary image corresponding to the high threshold after each update to obtain multiple white connected domains; and There is only one white connected domain in the binary image corresponding to the high threshold after the update , and the white connected domain In the When the binary image corresponding to the high threshold after the first update is divided into at least two white connected domains, stop updating the initial high threshold and set the first high threshold to zero. The updated high threshold is used as the target high threshold.

[0013] Preferably, the chain code extraction of the intersection contour based on the contour branch point includes: for any contour branch point, along the direction corresponding to any neighborhood point in the 8-neighborhood of the contour branch point, by judging whether there are other contour points in the 8-neighborhood of each contour point, if there are other contour points in the 8-neighborhood, then obtaining the chain code value of the other contour point compared with the contour point, and obtaining the chain code value of each contour point in turn until there are no other contour points in the 8-neighborhood of the contour point or the other contour points in the 8-neighborhood of the contour point belong to the contour branch point, and stopping, and forming all the chain code values into a chain code set of the contour branch point in this direction.

[0014] Preferably, the calculating the second probability that the connecting plate has a wrinkling defect according to the difference in chain code sets of the contour branch points in different directions includes: Where, is the second probability that the connecting plate has a wrinkling defect, is the number of all contour branch points on the intersection contour, For the The maximum value of the angle between the standard chain code values of each contour branch point in each two different directions, is the natural exponential function.

[0015] Preferably, the method for obtaining the standard chain code value of the contour branch point in each direction includes: for any contour branch point, calculating the mean of all chain code values in the chain code set of the contour branch point in each direction as the standard chain code value of the contour branch point in each direction.

[0016] Preferably, the method of judging whether the processed automobile connecting plate has a wrinkle defect according to the first probability and the second probability includes: when the connecting plate has a wrinkle defect, the first probability and the second probability of the connection plate having wrinkling defects The maximum value in is greater than the tolerance If the connection plate has a wrinkling defect, it is considered that the connection plate has a wrinkling defect; otherwise, the comprehensive probability of the connection plate having a wrinkling defect is calculated based on the first probability and the second probability of the connection plate having a wrinkling defect. When the comprehensive probability of the connection plate having a wrinkling defect is greater than the tolerance When the connection plate is considered to have wrinkling defects.

[0017] Preferably, the calculation formula for the comprehensive probability of the connection plate having wrinkling defects is: Where, is the comprehensive probability of wrinkling defects in the connecting plate, represents the maximum value function, Represents the minimum function; is the first probability of the existence of wrinkling defects in the connecting plate, is the second probability that the link plate has a wrinkle defect.

[0018] The beneficial effects of the present invention are:

[0019] The present invention is based on the two characteristics that wrinkling defects will affect guide lines corresponding to other assembly relationships of the connecting plate near the corner position and will generate new protrusions near the corner position of the connecting plate. The surface image is threshold segmented using a target low threshold and a target high threshold respectively, and the guide line contour is determined from the obtained low binary image. According to the curvature change of the contour points on the guide line contour, a first probability of the connecting plate having a wrinkling defect is calculated. The intersection contour is determined from the obtained high binary image, and according to the difference in chain code sets of each contour branch point in different directions on the intersection contour, a second probability of the connecting plate having a wrinkling defect is calculated. Then, based on the first and second probabilities of the connecting plate having a wrinkling defect, it is judged whether the processed automobile connecting plate has a wrinkling defect. The present invention can detect wrinkling defects more accurately, improve the product quality and production efficiency of automobile connecting plates, reduce the defective rate and production cost, and is suitable for online detection and quality control of automobile connecting plates. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1is a flow chart schematically illustrating an intelligent identification method for processing automobile connecting plates in the present invention;

[0021] Figure 2 is a schematic diagram schematically showing the corner position at the intersection of three sides of the connecting plate;

[0022] Figure 3 is a surface image diagram schematically showing the corner positions of a normal automobile connecting plate;

[0023] Figure 4 is a schematic diagram of a surface image of a corner of an automobile connecting plate having a wrinkle defect;

[0024] Figure 5 is a flowchart schematically illustrating step S2;

[0025] Figure 6 It is schematically shown Figure 3 The surface image shown is a schematic diagram of the black connected domain in the binary image corresponding to the low threshold after different updates;

[0026] Figure 7 It is schematically shown Figure 4 The schematic diagram of the binary image corresponding to the low threshold value after different updates of the surface image shown;

[0027] Figure 8 It is schematically shown Figure 3 The schematic diagram of the binary image corresponding to the high threshold value after different updates of the surface image shown;

[0028] Figure 9 It is schematically shown Figure 4 The shown diagram is a schematic diagram of the binary image corresponding to the high threshold after different updates of the surface image. DETAILED DESCRIPTION

[0029] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work shall fall within the scope of protection of the present invention.

[0030] The specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0031] The embodiment of the present invention discloses an intelligent identification method for processing automobile connecting plates, referring to Figure 1 , including steps S1 to S3:

[0032] S1. Collecting a surface image of a corner position of a processed automobile connecting plate.

[0033] It should be noted that the production of floor beam connecting plates adopts a forming process scheme, in which the middle area is formed in a free state on both sides of the pressing material. Since the two side walls with an angle of about 90 degrees are formed downward at the same time, and the pressing core at the R corner position of the product is in a closed state, the blank at the three-sided intersection position is squeezed and deformed and bulges at the corner position. After the mold is closed, the excess material cannot be eliminated, resulting in wrinkling defects at the corner position. Therefore, in order to detect whether the processed automobile connecting plates have wrinkling defects, it is necessary to collect surface images of the corner positions of the connecting plates, and determine whether the automobile connecting plates have wrinkling defects by analyzing the surface images.

[0034] Specifically, on the production line for producing automobile floor crossbeam connecting plates, after the stamping parts are formed and separated from the die, the surface image of the corner position of the processed automobile connecting plate at the three-sided intersection is captured by a camera, wherein the schematic diagram of the corner position of the three-sided intersection of the connecting plate is as shown in FIG. Figure 2 shown.

[0035] S2. Perform threshold segmentation on the surface image using the target low threshold and the target high threshold, determine the guide line contour from the obtained low binary image, calculate the first probability of the connection plate having a wrinkling defect based on the curvature change of the contour points on the guide line contour, determine the intersection contour from the obtained high binary image, and calculate the second probability of the connection plate having a wrinkling defect based on the differences in the chain code sets of each contour branch point in different directions on the intersection contour.

[0036] It should be noted that wrinkling defects have two effects on the corner position of the connecting plate:

[0037] (1) On the one hand, the wrinkling defect will affect the guide lines corresponding to other assembly relationships of the connecting plate near the corner position, causing the guide lines to become more compact, and ultimately manifested as a smaller angle at the turning point of the guide line, such as Figure 3 As shown in FIG, a schematic diagram of a surface image of a normal automobile connecting plate at a corner position is shown, wherein the area marked by the white box is the guide line corresponding to other assembly relationships of the connecting plate near the corner position; Figure 4 As shown, it is a schematic diagram of the surface image of the corner position of the automobile connecting plate with a wrinkling defect, wherein the area marked with a white box is the guide line affected by the wrinkling defect.

[0038] (2) On the other hand, the wrinkling defect will generate a new bulge near the corner of the connecting plate, which will be connected to the contour of the three-sided intersection at the corner, and finally appear as an increase in branches on the contour formed by brighter pixels, such as Figure 3As shown in FIG, a schematic diagram of a surface image of a normal automobile connecting plate at a corner position, wherein, due to the presence of two side walls with an angle of approximately 90° and a bottom wall with an angle of approximately 90° with the two side walls, the angle of the branch on the outline formed by the brighter pixel point is close to 90°; Figure 4 As shown, it is a schematic diagram of the surface image of the corner position of the automobile connecting plate with a wrinkling defect. The wrinkling defect will cause an increase in the branches on the contour formed by the brighter pixels, and the angles of the increased branches do not meet the condition of being close to 90°.

[0039] In summary, whether wrinkling defects occur at the corners of the connecting plate can be determined by the angle of the turning point of the guide line and the changes in the number and angle of branches on the outline formed by the brighter pixels.

[0040] See the flowchart of step S2 Figure 5 , including steps S201 to S202, specifically:

[0041] S201 , performing threshold segmentation on the surface image using a target low threshold, determining a guide line contour from the obtained low binary image, and calculating a first probability of a wrinkle defect existing on the connecting plate based on a curvature change of contour points on the guide line contour.

[0042] 1. Update the initial low threshold multiple times, and determine the target low threshold based on the changes in the number and connectivity of black connected domains in the binary image corresponding to the low threshold before and after each update.

[0043] Specifically, the initial low threshold is updated multiple times, and the surface image is threshold segmented according to the low threshold after each update to obtain the binary image corresponding to the low threshold after each update; all black pixels in the binary image corresponding to the low threshold after each update are analyzed for connected domains to obtain multiple black connected domains; the two black connected domains with the largest area are recorded as and ; until The two black connected regions with the largest area in the binary image corresponding to the low threshold after the update and , in When the binary image corresponding to the low threshold after the first update is merged into a black connected domain, stop updating the initial low threshold and set the The updated low threshold is used as the target low threshold.

[0044] The specific value of the initial low threshold can be set according to the actual application scenario and requirements, and the value range of the initial low threshold is [85,93]. The present invention sets the initial low threshold to 90.

[0045] For example, for Figure 3 The surface image of the corner position of the normal car connecting plate shown in the figure, in the binary image corresponding to the low threshold after the fourth update, the two black connected domains with the largest area are as follows Figure 6 As shown in (1), these two black connected domains are merged into one black connected domain in the binary image corresponding to the low threshold after the fifth update, as shown in Figure 6 As shown in (2), the low threshold after the fifth update is used as the target low threshold.

[0046] For example, for Figure 4 The surface image of the corner of the automobile connecting plate with wrinkling defects is shown in the figure. In the binary image corresponding to the low threshold after the fifth update, the two black connected domains with the largest area are as follows: Figure 7 As shown in (1), these two black connected domains are merged into one black connected domain in the binary image corresponding to the low threshold after the sixth update, as shown in Figure 7 As shown in (2), the low threshold after the sixth update is used as the target low threshold.

[0047] 2. Perform threshold segmentation on the surface image using the target low threshold to obtain a low binary image.

[0048] Specifically, pixels with grayscale values less than or equal to the target low threshold are marked as black pixels, and pixels with grayscale values greater than the target low threshold are marked as white pixels, thereby obtaining a low binary image.

[0049] For example, for Figure 3 The surface image of the corner position of the normal automobile connecting plate shown in FIG is threshold segmented by the target low threshold, and the schematic diagram of the low binary image obtained is as shown in FIG. Figure 6 As shown in (2), for Figure 4 The surface image of the corner position of the automobile connecting plate with wrinkle defects is shown in FIG. 1 . The surface image is threshold segmented by the target low threshold, and the schematic diagram of the low binary image obtained is shown in FIG. Figure 7 As shown in (2).

[0050] 3. Perform skeleton extraction on the black connected domain with the largest area in the low binary image to obtain the guide line outline of the surface image.

[0051] Among them, skeleton extraction is a technology in image processing, which is used to extract the center line or outline of an object from a binary image. Since wrinkling defects will cause the turning angle of the guide line to become smaller, it is necessary to perform skeleton extraction on the black connected domain with the largest area in the low binary image to obtain the guide line outline of the surface image.

[0052] The methods for skeleton extraction include but are not limited to: K3M algorithm (Kirschner-Mosteller-Kalaba algorithm), Medial Axis Transform (MAT), Medial Axis Transform (MAT), Distance Transform with Thinning, Topological Skeleton Extraction, and Fast Marching Method (FMM).

[0053] 4. Calculate the first probability of wrinkling defects on the connecting plate based on the curvature change of the contour points on the guide line contour.

[0054] Specifically, for any contour point on the guide line contour, linear fitting is performed on the positions of the contour point and its six adjacent points, and the slope of the fitted straight line is used as the curvature at the contour point.

[0055] Furthermore, the first probability of the connection plate having a wrinkle defect is calculated based on the curvature change of the contour points on the guide line contour. The calculation formula for the first probability of the connection plate having a wrinkle defect is:

[0056] ;

[0057] Where, is the first probability of the existence of wrinkling defects in the connecting plate, is the natural exponential function, For the first The curvature of the contour points, For the first The curvature of the contour point, the first contour points and contour points belong to adjacent contour points, is the number of all contour points contained in the guide line contour, represents the maximum value function, Represents the inverse tangent function, which is used to determine the angle from the curvature.

[0058] It should be noted that in the surface image of the automobile connecting plate, the angle of the contour point representing the turning point of the guide line will change suddenly. Therefore, by calculating the difference in the angle between two adjacent contour points, that is, , by obtaining the maximum value of the difference between the angles of each two adjacent contour points, that is , the contour points representing the turning points of the guide lines are obtained from the contours of the guide lines; on this basis, since the wrinkling defect will cause the guide lines to become more compact, it will eventually appear as a smaller angle at the turning point of the guide lines. Therefore, the smaller the angle of the contour points representing the turning points of the guide lines in the surface image of the automobile connecting plate, the greater the probability that the connecting plate has a wrinkling defect, that is, the first probability that the connecting plate has a wrinkling defect In addition, the denominator Used for Perform normalization.

[0059] S202. Perform threshold segmentation on the surface image using a target high threshold, determine the intersection profile from the obtained high binary image, and calculate a second probability of a wrinkle defect existing in the connecting plate based on the differences in chain code sets of each profile branch point in different directions on the intersection profile.

[0060] 1. Update the initial high threshold multiple times, and determine the target high threshold based on the changes in the number and connectivity of white connected domains in the binary image corresponding to the high threshold before and after each update.

[0061] Specifically, the initial high threshold is updated multiple times, and the surface image is threshold segmented according to the high threshold after each update to obtain a binary image corresponding to the high threshold after each update; all white pixels in the binary image corresponding to the high threshold after each update are analyzed for connected domains to obtain multiple white connected domains; until the There is only one white connected domain in the binary image corresponding to the high threshold after the update , and the white connected domain In the When the binary image corresponding to the high threshold after the first update is divided into at least two white connected domains, stop updating the initial high threshold and set the first high threshold to zero. The updated high threshold is used as the target high threshold.

[0062] The specific value of the initial high threshold can be set according to actual application scenarios and requirements, and the value range of the initial high threshold is [95,115]. The present invention sets the initial high threshold to 110.

[0063] For example, for Figure 3 The surface image of the corner position of the normal car connecting plate shown in FIG. 1 has only one white connected domain in the binary image corresponding to the high threshold after the 6th update. Figure 8 As shown in (1), the white connected domain is segmented into at least two white connected domains in the binary image corresponding to the high threshold after the 7th update, as shown in Figure 8 As shown in (1), the high threshold after the sixth update is used as the target high threshold.

[0064] For example, for Figure 3 The surface image of the corner of the automobile connecting plate with wrinkling defects is shown in FIG. 1 . In the binary image corresponding to the high threshold after the third update, there is only one white connected domain as shown in FIG. Figure 9 As shown in (1), the white connected domain is segmented into at least two white connected domains in the binary image corresponding to the high threshold after the 7th update, as shown in Figure 9 As shown in (2), the high threshold after the third update is used as the target high threshold.

[0065] 2. Perform threshold segmentation on the surface image using the target high threshold to obtain a high binary image.

[0066] Specifically, pixels with grayscale values less than or equal to the target high threshold are marked as black pixels, and pixels with grayscale values greater than the target high threshold are marked as white pixels, so as to obtain a high binary image, and the obtained high binary image has only one white connected domain.

[0067] For example, for Figure 3 The surface image of the corner position of the normal automobile connecting plate shown in FIG is threshold segmented by the target high threshold, and the schematic diagram of the high binary image obtained is as follows Figure 8 As shown in (1), for Figure 4 The surface image of the corner position of the automobile connecting plate with wrinkle defects is shown in FIG. 4 . The surface image is threshold segmented by the target high threshold, and the schematic diagram of the high binary image obtained is shown in FIG. Figure 9 As shown in (1).

[0068] 3. Extract the skeleton of the white connected domain in the high binary image to obtain the intersection angle contour of the surface image.

[0069] Among them, since wrinkling defects will produce new bulges near the corners of the connecting plate, resulting in an increase in branches on the contour formed by brighter pixels, it is necessary to perform skeleton extraction on the white connected domain in the high binary image to obtain the guide line contour of the surface image.

[0070] 4. Obtain all contour branch points on the intersection contour, extract chain codes from the intersection contour based on the contour branch points, and obtain the chain code sets of the contour branch points in each direction.

[0071] It should be noted that on the automobile connecting plate, the contours of the three-sided intersection angle converge at the corner position, and the angle between the two side walls is approximately 90°, and the angle between the bottom wall and the two side walls is approximately 90°. Therefore, the contour branch point is first obtained based on the number of neighboring points in the 8-neighborhood of the contour point on the intersection angle contour to represent the confluence of the contours of the three-sided intersection angle; then, based on the confluence of the contours of the three-sided intersection angle, that is, the contour branch point, the chain code of the intersection angle contour is extracted to obtain the chain code set of the contour branch point in each direction to represent the multiple branch contours constituting the three-sided intersection angle.

[0072] Specifically, the number of neighboring points in the 8-neighborhood of each contour point on the intersection contour is obtained. If the number of neighboring points in the 8-neighborhood of the contour point is greater than 1, it means that the 8-neighborhood of the contour point contains multiple neighboring points. The contour point that contains multiple neighboring points in the 8-neighborhood is recorded as a contour branch point.

[0073] Furthermore, for any contour branch point, along the direction corresponding to any neighborhood point within the 8-neighborhood of the contour branch point, by judging whether there are other contour points within the 8-neighborhood of each contour point, if there are other contour points within the 8-neighborhood, the chain code values of the other contour points compared with the contour point are obtained, and the chain code values of each contour point are obtained in turn until there are no other contour points within the 8-neighborhood of the contour point or the other contour points within the 8-neighborhood of the contour point belong to the contour branch point. All chain code values are combined into a chain code set of the contour branch point in this direction, and the chain code set of the contour branch point in this direction represents a branch contour constituting a three-sided intersection angle.

[0074] In this way, the chain code set of each contour branch point in the direction corresponding to each neighborhood point in the 8-neighborhood is obtained.

[0075] 6. According to the chain code set of the contour branch point in each direction, the standard chain code value of the contour branch point in each direction is obtained.

[0076] It should be noted that the chain code set of the contour branch point in each direction can represent multiple branch contours that constitute a three-sided intersection angle. The directional trends of the multiple branch contours that converge at the contour branch point are represented by the average of all chain code values in the chain code set of the contour branch point in each direction. Furthermore, whether the angles between the directional trends of the multiple branch contours meet the condition of being close to 90° is used to judge whether the multiple branch contours obtained are indeed branch contours that constitute a three-sided intersection angle, and further whether the contour branch point is indeed the confluence of the contours of the three-sided intersection angle.

[0077] It should be further explained that wrinkling defects will produce new protrusions near the corners of the connecting plate, which will be connected to the contours of the three-sided intersection at the corners, and ultimately manifest as an increase in branches on the contour formed by brighter pixels. Therefore, if there is a contour branch point that does not belong to the confluence of the contours of the three-sided intersection, it means that the automobile connecting plate has a wrinkling defect.

[0078] Specifically, for any contour branch point, the mean of all chain code values in the chain code set of the contour branch point in each direction is calculated as the standard chain code value of the contour branch point in each direction.

[0079] In this way, the standard chain code value of each contour branch point in the direction corresponding to each neighborhood point in the 8-neighborhood is obtained.

[0080] 7. According to the difference of chain code sets of each contour branch point in different directions on the intersection contour, the second probability of the existence of wrinkling defects in the connecting plate is calculated.

[0081] The specific calculation formula is:

[0082] ;

[0083] Where, is the second probability that the connecting plate has a wrinkling defect, is the number of all contour branch points on the intersection contour, For the The maximum value of the angle between the standard chain code values of each contour branch point in each two different directions, is the natural exponential function.

[0084] It should be noted that the closer the maximum value of the angle between the standard chain code values of the contour branch point in each two different directions is to 90°, the more likely the contour branch point is the contour branch point at the confluence of the contours of the three-face intersection angle. The closer it is to 1, on the contrary, the more the maximum value of the angle between the standard chain code values of the contour branch point in each two different directions deviates from 90°, the less likely the contour branch point is to belong to the contour branch point at the confluence of the contours of the three-face intersection angle, and the greater the probability of wrinkling defects in the connecting plate. The closer to 0, The closer it is to 1, the higher the second probability that the connecting plate has a wrinkling defect. The bigger.

[0085] S3. Determine whether the automobile connecting plate has a wrinkle defect based on the first probability and the second probability that the connecting plate has a wrinkle defect.

[0086] 1. The first probability of wrinkling defects on the connecting plate and the second probability of the connection plate having wrinkling defects The maximum value in is greater than the tolerance When the connection plate is considered to have wrinkling defects.

[0087] Among them, the tolerance The specific value can be set according to the actual application scenario and requirements, and the tolerance The value range of is [0.85, 1), and the present invention sets the preset first threshold to 0.9.

[0088] 2. Otherwise, the comprehensive probability of the connection plate wrinkling defect is calculated based on the first probability and the second probability of the connection plate wrinkling defect. When the comprehensive probability of the connection plate wrinkling defect is greater than the tolerance When the connection plate is considered to have wrinkling defects.

[0089] The calculation formula for the comprehensive probability of wrinkling defects in the connecting plate is:

[0090] ;

[0091] Where, is the comprehensive probability of wrinkling defects in the connecting plate, represents the maximum value function, represents the minimum value function; is the first probability of the existence of wrinkling defects in the connecting plate, is the second probability that the link plate has a wrinkle defect.

[0092] It should be noted that the present invention determines whether a processed automobile connecting plate has a wrinkling defect based on the first probability and the second probability of the connecting plate having a wrinkling defect. This can more accurately detect wrinkling defects, improve the product quality and production efficiency of automobile connecting plates, reduce the defective rate and production costs, and is suitable for online detection and quality control of automobile connecting plates.

Claims

1. An intelligent identification method for automobile connecting plate processing, characterized in that: include: Collecting surface images of the corners of the processed automobile connecting plates; The initial low threshold is updated multiple times, and the surface image is threshold segmented according to the low threshold after each update, so as to obtain a binary image corresponding to the low threshold after each update; Perform connected domain analysis on all black pixels in the binary image corresponding to the low threshold after each update to obtain multiple black connected domains; until the The two black connected domains with the largest area in the binary image corresponding to the low threshold after the update are When the binary image corresponding to the low threshold after the first update is merged into a black connected domain, stop updating the initial low threshold and set the The updated low threshold is used as the target low threshold; the surface image is segmented by the target low threshold, and the skeleton is extracted from the black connected domain with the largest area in the obtained low binary image to obtain the guide line contour; the first probability of the connection plate having a wrinkle defect is calculated based on the curvature change of the contour points on the guide line contour , is the natural exponential function, For the first The curvature of the contour points, For the first The curvature of the contour points, is the number of all contour points contained in the guide line contour; The initial high threshold is updated multiple times, and the surface image is threshold segmented according to the high threshold after each update, so as to obtain a binary image corresponding to the high threshold after each update; Perform connected domain analysis on all white pixels in the binary image corresponding to the high threshold after each update to obtain multiple white connected domains; until the There is only one white connected domain in the binary image corresponding to the high threshold after the update , and the white connected domain In the When the binary image corresponding to the high threshold after the first update is divided into at least two white connected domains, stop updating the initial high threshold and set the first high threshold to zero. The high threshold after the update is used as the target high threshold; the surface image is threshold segmented using the target high threshold, and the skeleton of the white connected domain in the obtained high binary image is extracted to obtain the intersection contour of the surface image; the contour branch points containing multiple neighborhood points in the 8 neighborhoods on the intersection contour are obtained, and the chain code of the intersection contour is extracted based on the contour branch points. According to the difference in the chain code sets of the contour branch points in different directions, the second probability of the existence of wrinkling defects in the connecting plate is calculated , is the number of all contour branch points on the intersection contour, For the The maximum value of the angle between the standard chain code values of each contour branch point in each two different directions; It is determined whether the processed automobile connecting plate has a wrinkle defect according to the first probability and the second probability.

2. The intelligent identification method for automobile connecting plate processing according to claim 1, characterized in that: The method for obtaining the curvature at the contour point includes: For any contour point on the guide line contour, a linear fit is performed on the position of the contour point and its six adjacent points, and the slope of the fitted straight line is used as the curvature at the contour point.

3. The intelligent identification method for automobile connecting plate processing according to claim 1, characterized in that: The chain code extraction of the intersection angle contour based on the contour branch point includes: For any contour branch point, along the direction corresponding to any neighborhood point in the 8-neighborhood of the contour branch point, by judging whether there are other contour points in the 8-neighborhood of each contour point, if there are other contour points in the 8-neighborhood, the chain code values of other contour points compared with the contour point are obtained, and the chain code values of each contour point are obtained in turn until there are no other contour points in the 8-neighborhood of the contour point or the other contour points in the 8-neighborhood of the contour point belong to the contour branch point, and all chain code values are combined into a chain code set of the contour branch point in this direction.

4. The intelligent identification method for automobile connecting plate processing according to claim 1, characterized in that: The method for obtaining the standard chain code value of the contour branch point in each direction includes: For any contour branch point, the mean of all chain code values in the chain code set of the contour branch point in each direction is calculated as the standard chain code value of the contour branch point in each direction.

5. The intelligent identification method for automobile connecting plate processing according to claim 1, characterized in that: The determining, based on the first probability and the second probability, whether the processed automobile connecting plate has a wrinkle defect includes: The first probability when the connecting plate has wrinkling defects and the second probability of the connection plate having wrinkling defects The maximum value in is greater than the tolerance When , it is considered that the connecting plate has wrinkling defects; Otherwise, the comprehensive probability of the connection plate having a wrinkle defect is calculated based on the first probability and the second probability of the connection plate having a wrinkle defect. When the comprehensive probability of the connection plate having a wrinkle defect is greater than the tolerance When the connection plate is considered to have wrinkling defects.

6. The intelligent identification method for automobile connecting plate processing according to claim 1, characterized in that: The calculation formula for the comprehensive probability of the connection plate having wrinkling defects is: ； Where, is the comprehensive probability of wrinkling defects in the connecting plate, represents the maximum value function, Represents the minimum function; is the first probability of the existence of wrinkling defects in the connecting plate, is the second probability that the link plate has a wrinkle defect.

Images