A method for detecting rearview mirror mold precision based on machine vision

Abstract

The present invention relates to the field of image processing technology, and more specifically, to a method for detecting the precision of rearview mirror molds based on machine vision. The method comprises: obtaining a grayscale image of the rearview mirror mold; extracting edge information from the grayscale image; filtering the rearview mirror's outline from the edge information; filtering target edges belonging to the reflective area from the outline; and obtaining target pixels; calculating the edge probability of each target pixel, and determining target pixels with an edge probability greater than a set value as edge pixels; determining the edge formed by all edge pixels as the final edge of the reflective area, thereby obtaining the final outline of the rearview mirror; and determining the dimensions of the rearview mirror mold based on the final outline. The present invention thus ensures dimensional precision detection of rearview mirror molds.

Description

Technical Field

[0001] The present invention relates to the field of image processing technology, and more particularly to a method for detecting the precision of a rearview mirror mold based on machine vision. Background Art

[0002] As a critical component for automotive safety and functionality, the dimensional accuracy of the rearview mirror determines the precision of its assembly with other components, such as the vehicle body and brackets. Imprecise mold dimensions can result in an insecure installation of the rearview mirror, potentially compromising driving safety. High-precision molds can reduce processing and adjustment time, minimize rework and inspection, improve production efficiency, and reduce scrap rates, thereby saving time and costs.

[0003] When performing precision testing on automotive rearview mirror molds, various aspects of the mold are typically inspected, including dimensional accuracy, shape accuracy, surface quality, and assembly accuracy. Dimensional accuracy is one of the core indicators of mold manufacturing quality, as it determines the degree of deviation between the mold's actual dimensions and the designed dimensions.

[0004] Currently, optical inspection and image processing technologies are used to inspect the dimensional accuracy of rearview mirror molds. Optical inspection, which relies on automated optical inspection systems, is capable of inspecting the dimensional accuracy of rearview mirror molds. However, it is expensive, poorly adaptable to complex and changing inspection requirements, and difficult to adjust inspection items and parameters. Image processing technology, on the other hand, uses industrial cameras to capture images of the mold and determine the edge contours to determine the corresponding dimensional data. However, due to the unique shape and structure of rearview mirror molds, some edges have localized reflective areas. During canny edge inspection, the outlines of these reflective areas are difficult to detect, resulting in inaccurate contours of the rearview mirror mold.

[0005] Therefore, because the acquired image is sensitive to light and object reflectivity, there may be inaccuracies when capturing the edges of the rearview mirror mold image. Therefore, it is particularly important to improve the dimensional accuracy of the rearview mirror mold while taking into account the cost. Summary of the Invention

[0006] The purpose of the present invention is to propose a rearview mirror mold precision detection method based on machine vision to solve the problems of high cost and inaccurate detection when performing dimensional accuracy detection of rearview mirror molds in the prior art; to this end, the present invention provides a solution in the following aspect.

[0007] The present invention provides a method for detecting the precision of a rearview mirror mold based on machine vision, comprising:

[0008] Acquire a grayscale image of the rearview mirror mold; extract edge information from the grayscale image;

[0009] Filtering the outline of the rearview mirror from the edge information;

[0010] Obtain the target edge in the reflective area of ​​the outline; and obtain the target pixel point in the reflective area; the target pixel point is a pixel point in each edge as the starting point. If the grayscale difference between any pixel point and its adjacent point in the gradient direction is greater than a threshold, the grayscale difference between the adjacent points in the gradient direction is calculated, and so on, until the pixel point with a grayscale difference less than or equal to the threshold is obtained;

[0011] Calculate the edge probability of each target pixel and define the target pixel with an edge probability greater than a set value as an edge pixel; the edge probability is inversely correlated with the grayscale variation mean of the corresponding target pixel; the grayscale variation mean is the mean of the grayscale differences between all two adjacent pixels in a pixel set; the pixel set includes the target pixel and a set number of pixels along the gradient direction of the target pixel;

[0012] The new edge formed by all edge pixels is used as the final edge of the reflective area, thereby obtaining the final outline of the rearview mirror; and the size of the rearview mirror mold is obtained based on the final outline.

[0013] The above-mentioned solution extracts the target edge of the reflective area and, based on the grayscale variation between the pixels on any edge and the adjacent pixels in the corresponding gradient direction, determines the suspected edge pixels (target pixels) in the reflective area. The solution then calculates the edge probability of the target pixels to determine whether they are true edge pixels, thereby determining the final edge of the reflective area and the final contour of the rearview mirror mold, thereby obtaining the contour dimensions of the rearview mirror mold. In other words, the present invention can obtain accurate edge pixels through a detailed analysis of the pixels on each edge of the reflective area and the target pixels, providing precise data for the subsequent rearview mirror mold dimensions while reducing the cost of dimension measurement.

[0014] Optionally, the marginal probability is:

[0015] ;in, is the mean grayscale change of the n+1th target pixel, is the edge probability of the n+1th target pixel.

[0016] The above provides a method for accurately calculating the edge probability of a target pixel.

[0017] Optionally, the specific process of obtaining the target edge belonging to the reflective area in the outline is:

[0018] Get the endpoints of all edges on the rearview mirror's outline;

[0019] Calculate the possibility that the edge of each endpoint belongs to the reflective area;

[0020] The edge where the endpoint with a probability greater than the set threshold is located is regarded as the target edge of the reflective area;

[0021] Among them, the possibility is positively correlated with the angle between the corresponding endpoints and the modulus of the second vector; the angle is the angle between the first vector and the second vector, the first vector is a vector pointing from the pixel point adjacent to any endpoint to any endpoint, and the second vector is a vector pointing from any endpoint to the second endpoint; the second endpoint is the endpoint with the smallest distance from any endpoint obtained in a clockwise direction with any endpoint as the starting point.

[0022] In the above solution, by analyzing the shape of the rearview mirror mold and the characteristics of the collected image, the target edge belonging to the reflective area can be accurately determined.

[0023] Optionally, the possibility is:

[0024] ; represents the angle between the mth endpoints, is the mth endpoint, is the endpoint with the smallest distance from the mth endpoint selected in the clockwise direction, represents the magnitude of the second vector, It indicates the possibility that the edge where the mth endpoint is located is a reflective area, and norm() is the normalization function.

[0025] Optionally, the specific process of obtaining the target edge belonging to the reflective area in the outline is:

[0026] The endpoints of all edges on the rearview mirror's outline are obtained, and the endpoints are clustered to obtain multiple clusters. The edges in the area where the number of clusters with more endpoints than a set number is located are selected as the target edges of the reflective area.

[0027] The above scheme can determine the target edge belonging to the reflective area by clustering the endpoints on the edge.

[0028] Optionally, filtering out the outline of the rearview mirror from the edge information includes: selecting an edge corresponding to a maximum circularity as the outline of the rearview mirror;

[0029] Alternatively, a contour detection algorithm is used to obtain the boundary in the edge information to obtain the contour of the rearview mirror.

[0030] Optionally, the reflective area is a minimum circumscribed rectangle corresponding to an edge of the target.

[0031] Optionally, extracting edge information from the grayscale image includes: using a Canny algorithm to obtain edge information from the grayscale image.

[0032] Optionally, taking the new edge formed by all edge pixels as the final edge of the reflective area includes: connecting all edge pixels in sequence to obtain the final edge.

[0033] Optionally, obtaining the size of the rearview mirror mold according to the final contour includes:

[0034] Respectively obtaining the maximum and minimum values ​​of the horizontal coordinates and the maximum and minimum values ​​of the vertical coordinates of the pixel points in the final contour;

[0035] The absolute value of the difference between the maximum and minimum values ​​of the horizontal coordinate is used as the length of the rearview mirror mold; the absolute value of the difference between the maximum and minimum values ​​of the vertical coordinate is used as the width of the rearview mirror mold.

[0036] The above solution can obtain more accurate dimensions of the rearview mirror mold.

[0037] The beneficial effects of the present invention are:

[0038] The solution of the present invention analyzes the target edge belonging to the reflective area in the grayscale image of the acquired rearview mirror mold based on the shape characteristics of the acquired rearview mirror mold, determines the final edge of the reflective area, and then obtains the final contour of the rearview mirror mold, thereby solving the problem of inaccurate contour shape obtained when the rearview mirror mold has a complex shape, and at the same time reduces the error generated when obtaining the rearview mirror mold size. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 A flowchart schematically illustrates a method for detecting rearview mirror mold accuracy based on machine vision in this embodiment;

[0040] Figure 2 The edge image of the rearview mirror mold in this embodiment is schematically shown;

[0041] Figure 3 The outline of the rearview mirror mold in this embodiment is schematically shown;

[0042] Figure 4 The reflective area in this embodiment is schematically shown. DETAILED DESCRIPTION

[0043] The technical solutions in the embodiments of the present invention will be described clearly and completely below with reference to the accompanying drawings in the embodiments of the present invention.

[0044] The present invention addresses the problem that the complex shape of the mold edge easily leads to the appearance of reflective areas, and the grayscale of the reflective area is relatively close to the background area. Therefore, there is no obvious boundary when transitioning from the reflective area to the background area. At this time, Canny edge detection cannot accurately identify the edge of the rearview mirror mold in the reflective area, and thus cannot accurately detect the size of the rearview mirror mold. Therefore, a rearview mirror mold precision detection method based on machine vision is proposed, that is, by accurately obtaining the complete rearview mirror contour and obtaining the length and width of the mold according to the distance between the edge pixels in the contour.

[0045] Specifically, if Figure 1 As shown, a rearview mirror mold precision detection method based on machine vision in this embodiment includes the following steps:

[0046] Step S1, obtaining a grayscale image of a rearview mirror mold.

[0047] Specifically, in this embodiment, when capturing images of the rearview mirror mold, a high-resolution digital camera or industrial camera is selected to ensure image clarity and detail capture. Furthermore, the rearview mirror mold must be placed stably and kept horizontal or vertical to avoid errors caused by angles.

[0048] After acquiring the image, a filtering algorithm (such as Gaussian filtering) is used to remove noise from the captured image to enhance image quality. The filtered image is then grayscaled to obtain a grayscale image.

[0049] Step S2: determining the contour of the rearview mirror in the grayscale image of the rearview mirror mold, and correcting the contour to obtain a final contour.

[0050] according to Figure 2 As can be seen, due to the relatively complex appearance of some rearview mirror molds, some edge areas may reflect light when performing edge detection on the grayscale image. In this case, the edges are not accurately extracted, resulting in errors in the obtained rearview mirror mold edges, which in turn leads to errors in the obtained rearview mirror mold dimensional data. Therefore, it is necessary to obtain the contour of the rearview mirror mold and analyze the contour to obtain the final contour.

[0051] The process of obtaining the final contour includes steps S21-S24, specifically:

[0052] Step S21: extract edge information from the grayscale image.

[0053] Specifically, in this embodiment, the canny algorithm is used to obtain the edge image in the image, such as Figure 2 As shown; and then obtaining edge information in the edge image; wherein the edge information is the edge in the edge image.

[0054] Step S22: Filter out the outline of the rearview mirror mold from the edge information.

[0055] In one embodiment, a contour detection algorithm is used to obtain the boundary in the edge information to obtain the contour of the rearview mirror.

[0056] In another embodiment, the edge corresponding to the maximum circularity is selected as the outline of the rearview mirror. Since the calculation of circularity is a prior art, it will not be described in detail here.

[0057] The reason for calculating circularity above is that Figure 2 It can be seen that the contour of the rearview mirror mold is a smooth curve that is approximately circular, and this curve is in a nearly closed state. That is, the edge of the rearview mirror mold in the edge detection image is extracted based on the mold's shape features.

[0058] Step S23: Filter out the target edge belonging to the reflective area from the outline.

[0059] This embodiment takes into account the complex edge morphology of the rearview mirror mold when capturing an image of the entire rearview mirror mold. This may result in light reflection, which can cause blurring of reflective areas at the edges, resulting in complex and uneven edges in the obtained edge image. Therefore, this embodiment utilizes the morphological characteristics of the target edge in the reflective area (which is more irregular and less smooth) to capture the target edge in the reflective area within the outline.

[0060] In one embodiment, the process of obtaining the target edge belonging to the reflective area from the outline is as follows:

[0061] First, get the endpoints of all edges on the mirror's outline.

[0062] In this embodiment, it is considered that the outline of the rearview mirror may not be a closed outline, and it may be composed of multiple consecutive edges, the edges are disconnected, and the distance between the end points is relatively close.

[0063] Secondly, calculate the possibility that the edge where each endpoint is located belongs to the reflective area.

[0064] In this embodiment, after obtaining all endpoints, it is necessary to determine the possibility that each endpoint belongs to the reflective area. The specific process is as follows:

[0065] First, the first vector and second vector of each endpoint are obtained, and then the angle between the first vector and the second vector is obtained. According to the angle and the modulus of the second vector, the possibility that the corresponding endpoint belongs to the reflective area is calculated.

[0066] The above-mentioned first vector is a vector pointing from a pixel point adjacent to any endpoint to any endpoint; the second vector is a vector pointing from any endpoint to the second endpoint; the second endpoint is the endpoint with the smallest distance from any endpoint obtained in a clockwise direction starting from any endpoint.

[0067] For example, select an endpoint on any edge , and in the clockwise direction, get the distance endpoint The nearest endpoint , at this point, connect the two endpoints and generate the second vector ; Get endpoint Adjacent pixels on the same edge , and get the first vector . The second vector With the first vector The angle between .

[0068] The specific expression formula of possibility is:

[0069] ; represents the angle between the mth endpoints, is the mth endpoint, is the endpoint with the smallest distance from the mth endpoint selected in the clockwise direction, represents the magnitude of the second vector, It indicates the possibility that the edge where the mth endpoint is located is a reflective area, and norm() is the normalization function.

[0070] The angle above represents the angular change of the second vector relative to the first vector (the first vector represents the edge between the two pixels). The smaller the angular change, the more likely the edge corresponding to that endpoint is a normal area. Conversely, the larger the angular change, the more likely the edge corresponding to that endpoint is a reflective area. Furthermore, because the distance between endpoints in reflective areas is relatively large, the modulus of the second vector is used to represent the distance between two endpoints on two different edges. That is, the larger the modulus, the more likely the edge containing the two endpoints is a reflective area.

[0071] The above normalization function may be maximum and minimum value normalization, tangent hyperbolic normalization, etc.

[0072] In this embodiment, due to the shape of the rearview mirror mold and light reflection, the edges in the reflective area of ​​the grayscale image appear irregular, with forks and large breaks. The distance between the second vector formed by the trailing endpoint of the edge and the leading endpoint of the next edge is generally large, and the angle between the first vector formed by the trailing endpoint of the edge and its adjacent pixel points and the second vector is also large. In contrast, edges in normal areas tend to be smooth, single arc segments, with the trailing endpoint of the edge and the leading endpoint of the next edge being close, and the angle of the vector formed by the line connecting the two endpoints changes less than that of the edge. Therefore, the likelihood that the edge at the corresponding endpoint belongs to the reflective area is calculated by using the angle and the modulus of the second vector.

[0073] In another embodiment, the process of obtaining the edge of the target belonging to the reflective area is as follows:

[0074] Get the endpoints of all edges on the rearview mirror's outline, cluster the endpoints, and select the edges in the area where the number of clusters exceeds a set number of endpoints as the edges of the reflective area. The set number is set to 4.

[0075] Then, the edge where the endpoint whose probability is greater than the set threshold is located is regarded as the edge of the reflective area.

[0076] In this embodiment, the threshold is set to 0.85, that is, endpoints with a probability greater than 0.85 are extracted, and the area where the edge of the corresponding endpoint is located is used as the reflective area.

[0077] It should be noted that when there are two endpoints on an edge, as long as there is an endpoint with a probability greater than 0.85, the area where the corresponding edge is located will be used as the reflective area.

[0078] After determining the target edges belonging to the reflective area, the minimum bounding rectangle of the determined one or more target edges is used as the reflective area, such as Figure 4 The area corresponding to the white box.

[0079] It should be noted that, in this embodiment, there is only one reflective area. Of course, in other implementations, there may be multiple reflective areas, in which case each reflective area corresponds to a target edge.

[0080] When there are multiple reflective areas, the target edges can be clustered to obtain clusters of different categories, and the minimum circumscribed rectangle of the target edge in each cluster is used as the corresponding reflective area.

[0081] Step S24, obtain the target pixel points in the reflective area, calculate the edge probability of each target pixel point, take the target pixel points with an edge probability greater than the set value as the edge pixel points, and take the new edge formed by all edge pixels as the final edge of the reflective area, thereby obtaining the final outline of the rearview mirror.

[0082] After obtaining the target edge belonging to the reflective area, in order to obtain a more realistic final edge in the reflective area, this embodiment analyzes each pixel in the target edge of the reflective area to determine the true edge pixel, thereby obtaining the final edge.

[0083] Before determining the edge pixel points, each pixel point in the target edge belonging to the reflective area is analyzed to obtain the target pixel point, and then the edge probability of the target pixel point is analyzed.

[0084] Specifically, the method for selecting the target pixel point is:

[0085] Obtain any pixel point in any edge of the reflective area, take the any pixel point as the starting point, and obtain the first pixel point adjacent to it along the gradient direction of any pixel point. Calculate the grayscale difference between any pixel point and the first pixel point. If the grayscale difference is greater than the threshold, continue to locate the second pixel point along the gradient direction of the first pixel point, and continue to judge the grayscale difference until the grayscale difference is less than or equal to the threshold. Stop and take the pixel point when the grayscale difference is less than or equal to the threshold as the target pixel point.

[0086] For example, based on any pixel point O1 on any edge extracted, locate the next pixel point O2 along its gradient direction, and express the grayscale difference between the two pixels as I 12 , if I 12 If the value is greater than the threshold, continue to locate the next pixel O3 along the gradient direction of pixel O2 and calculate the grayscale difference I between the two pixels. 23 , compare it with the threshold until I n,n+1 The pixel point On+1 corresponding to the value less than or equal to the threshold is used as the target pixel point; continue to traverse other pixel points on any edge to obtain all target pixel points.

[0087] The threshold value is set to 5, which can of course be determined according to actual conditions. The grayscale difference is the absolute value of the grayscale difference between two adjacent pixels.

[0088] Since the grayscale value of the pixel point gradually decreases along the direction of the maximum gradient change when transitioning from the reflective area to the background area, and the grayscale value is approximately unchanged when finally reaching the background area, this embodiment locates the target pixel point (suspected edge pixel point) based on the approximately unchanged characteristic of the grayscale value of the pixel point on the target edge of the reflective area along the gradient direction.

[0089] It should be noted that when determining the target pixel points mentioned above, all target edges in the reflective area are targeted; of course, as other implementation methods, the target edge where the pixel point with the largest gradient is located can also be determined, and the target edge can be analyzed without analyzing other target edges, which can simplify the calculation.

[0090] In this embodiment, after obtaining the target pixel point, a set number of pixel points in the extension direction of the gradient direction are obtained along the gradient direction of the target pixel point, the set number of pixel points and the corresponding target pixel point constitute a pixel point set, and the grayscale difference between two adjacent pixel points in the pixel point set is obtained, the average of all grayscale differences is used as the average grayscale change amount, and based on the average grayscale change amount, the edge probability of the corresponding target pixel point is calculated.

[0091] Specifically, the marginal probability is:

[0092] ;

[0093] in, is the mean grayscale change of the n+1th target pixel, is the edge probability of the n+1th target pixel.

[0094] In this embodiment, after obtaining the edge probability of each target pixel point, the corresponding edge probability is also compared with the set value. That is, when the edge probability of the target pixel point is greater than the set value, the target pixel point is an edge pixel point, and then all edge pixel points are obtained; all edge pixel points are connected in sequence to obtain the final edge of the reflective area.

[0095] The above setting value is 0.85, and of course it can be determined according to actual conditions as other implementation methods.

[0096] The reason for calculating the edge probability of edge pixels above is that, due to the special prismatic structure of the rearview mirror mold, the grayscale values ​​of the pixels in the edge area of ​​the grayscale image of the captured image are close to the grayscale values ​​of the background area, making it impossible to accurately determine the true edge during edge detection. Therefore, it is necessary to calculate the edge probability of the target pixel.

[0097] In this embodiment, the final edge of the obtained reflective area is connected with the edges of other normal areas to obtain the final outline of the rearview mirror.

[0098] Step S5: obtaining the size of the rearview mirror mold according to the final contour.

[0099] Specifically, in this embodiment, the maximum and minimum values ​​of the horizontal coordinates and the maximum and minimum values ​​of the vertical coordinates of the pixel points in the final contour are respectively obtained; the absolute value of the difference between the maximum and minimum values ​​of the horizontal coordinates is used as the length of the rearview mirror mold; and the absolute value of the difference between the maximum and minimum values ​​of the vertical coordinates is used as the width of the rearview mirror mold.

[0100] For example, the coordinate values ​​of all pixels on the edge of the mold are extracted, and the four pixels with the largest and smallest horizontal and vertical coordinates are located. The width of the mold is expressed as W and the length is expressed as L. The specific calculation method is as follows:

[0101] ;

[0102] .

[0103] The solution of the present invention obtains the real outline of the rearview mirror mold in the image of the rearview mirror mold and then obtains the distance between the pixel points in the outline to obtain the size of the rearview mirror mold with high precision without contact.

[0104] While several embodiments of the present invention have been shown and described herein, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Numerous modifications, variations, and alternatives will occur to those skilled in the art without departing from the concept and spirit of the present invention. It should be understood that various alternatives to the embodiments of the present invention described herein may be employed in practicing the present invention.

Claims

1. A rearview mirror mold precision detection method based on machine vision, characterized in that: include: Obtain a grayscale image of the rearview mirror mold; Extracting edge information from the grayscale image; Filtering the outline of the rearview mirror from the edge information; Obtain the target edge in the outline that belongs to the reflective area, including: obtaining the endpoints of all edges on the outline of the rearview mirror; calculating the probability that the edge where each endpoint is located belongs to the reflective area: , represents the angle between the mth endpoints, is the mth endpoint, is the endpoint with the smallest distance from the mth endpoint selected in the clockwise direction, represents the magnitude of the second vector, Indicates the possibility that the edge where the mth endpoint is located is a reflective area, norm() is a normalization function; the edge where the endpoint with a possibility greater than the set threshold is located is taken as the target edge of the reflective area; the angle is the angle between the first vector and the second vector, the first vector is a vector from a pixel point adjacent to any endpoint to any endpoint, and the second vector is a vector from any endpoint to the second endpoint; the second endpoint is the endpoint with the smallest distance from any endpoint obtained in a clockwise direction with any endpoint as the starting point; and the target pixel point in the reflective area is obtained; the target pixel point is any pixel point in each edge as the starting point, if the grayscale difference between any pixel point and its adjacent point in the gradient direction is greater than the threshold, then the grayscale difference between the adjacent point and the adjacent point in the gradient direction is calculated, and so on, until the pixel point with a grayscale difference less than or equal to the threshold is obtained; Calculate the edge probability of each target pixel and define the target pixel with an edge probability greater than a set value as an edge pixel; the edge probability is inversely correlated with the grayscale variation mean of the corresponding target pixel; the grayscale variation mean is the mean of the grayscale differences between all two adjacent pixels in a pixel set; the pixel set includes the target pixel and a set number of pixels along the gradient direction of the target pixel; The new edge formed by all edge pixels is used as the final edge of the reflective area, thereby obtaining the final outline of the rearview mirror; and the size of the rearview mirror mold is obtained based on the final outline.

2. The method for detecting rearview mirror mold precision based on machine vision according to claim 1, characterized in that: The marginal probability is: ;in, is the mean grayscale change of the n+1th target pixel, is the edge probability of the n+1th target pixel.

3. The method for detecting rearview mirror mold precision based on machine vision according to claim 1, characterized in that: The step of filtering out the outline of the rearview mirror from the edge information includes: selecting an edge corresponding to a maximum circularity as the outline of the rearview mirror; Alternatively, a contour detection algorithm is used to obtain the boundary in the edge information to obtain the contour of the rearview mirror.

4. The method for detecting rearview mirror mold precision based on machine vision according to claim 1, characterized in that: The reflective area is the minimum circumscribed rectangle corresponding to the edge of the target.

5. The method for detecting rearview mirror mold precision based on machine vision according to claim 1, characterized in that: The extracting edge information from the grayscale image includes: using a Canny algorithm to obtain the edge information from the grayscale image.

6. The method for detecting rearview mirror mold precision based on machine vision according to claim 1, characterized in that: The method of using the new edge formed by all edge pixels as the final edge of the reflective area includes: connecting all edge pixels in sequence to obtain the final edge.

7. The method for detecting rearview mirror mold precision based on machine vision according to claim 1, characterized in that: The step of obtaining the size of the rearview mirror mold according to the final contour includes: Respectively obtaining the maximum and minimum values ​​of the horizontal coordinates and the maximum and minimum values ​​of the vertical coordinates of the pixel points in the final contour; The absolute value of the difference between the maximum and minimum values ​​of the horizontal coordinate is used as the length of the rearview mirror mold; the absolute value of the difference between the maximum and minimum values ​​of the vertical coordinate is used as the width of the rearview mirror mold.

Images