An Appearance Contour Extraction Method for Automobile Part Processing

Abstract

The present invention relates to the technical field of image data. More specifically, the present invention relates to a method for extracting the appearance contour for the machining of automotive parts. The method includes: evenly dividing an automotive wheel hub image into a plurality of fan-shaped sub-blocks, and determining the target object of the fan-shaped sub-block; determining the proximity of the pixel points on the edge between the fan-shaped sub-block and the target object according to the number of overlapping pixel points between the fan-shaped sub-block and the edge of the target object, as well as the gradient and gradient direction between the pixel points; calculating the wheel hub similarity of the edge between the fan-shaped sub-block and the target object; calculating the elimination degree of the edge in the fan-shaped sub-block, and determining the edge elimination result of the fan-shaped sub-block; in response to the edge elimination result of the fan-shaped sub-block, taking the fan-shaped sub-block after the edge is eliminated as the target object of the remaining fan-shaped sub-blocks, and continuing to perform edge elimination on the remaining fan-shaped sub-blocks, so as to obtain the spoke edge in the automotive wheel hub image, effectively improving the accuracy of the spoke contour extraction in the automotive wheel hub image.

Description

Technical Field

[0001] The present invention relates to the technical field of image data. More specifically, the present invention relates to a method for extracting the appearance contour of automobile parts during processing. Background Art

[0002] An automobile wheel hub is one of the core components of an automobile wheel. It connects the wheel to the axle, bears the weight of the vehicle, and transmits power and braking force. The wheel hub spokes (also known as the spoke plates) are important components that connect the rim and the wheel hub. Their main functions are to support the wheel, transmit the weight and torque of the vehicle, and protect the wheel from external damage.

[0003] If the quality of the automobile wheel hub is unqualified, it will not only affect the appearance of the wheel hub, but also may affect the safety and performance of the vehicle. Therefore, after the production of the automobile wheel hub is completed, it is necessary to conduct quality monitoring on the wheel hub spokes to identify and screen the wheel hub spokes that may have defects such as bending, cracking, wear, and uneven distribution. At present, edge detection is mostly used to extract the edges of the target detection object for quality detection. For example, the patent application document with the publication number CN118212255A discloses an image edge extraction method. This method trains a convolutional neural network model by using each original image and the corresponding edge information as samples and expected output results respectively to obtain an image edge extraction model, thereby realizing the edge extraction of the target detection object.

[0004] However, due to the complex structure of the automobile wheel hub, when extracting the edges of the wheel hub spokes through edge detection, non-wheel hub spoke edges may also be extracted, resulting in the non-wheel hub spoke edges in the obtained wheel hub image being mixed with the wheel hub spoke edges as noise edges. When training the convolutional neural network model, if there are noise edges in the original image, the features of the noise edges may be learned simultaneously, resulting in the model being unable to accurately distinguish the true target detection object edges from the noise edges when extracting edges, thereby affecting the accuracy of edge extraction.

[0005] Based on this, how to accurately remove the non-wheel hub spoke edges existing in the wheel hub image, so as to accurately obtain the result of extracting the appearance contour of the automobile wheel hub, is an urgent problem to be solved by those skilled in the art. Summary of the Invention

[0006] To solve the above technical problem of how to accurately remove the non-wheel hub spoke edges existing in the wheel hub image, so as to accurately obtain the result of extracting the appearance contour of the automobile wheel hub, the present invention proposes a method for extracting the appearance contour of automobile parts during processing. The method includes the following steps:

[0007] Divide the wheel hub image into a plurality of fan-shaped sub-blocks according to the number of wheel hub spokes in the automobile wheel hub image;

[0008] Take the other sector sub - blocks outside a sector sub - block as the target objects corresponding to this sector sub - block; overlap the sector sub - block with the edges in the corresponding target objects, and determine the coincidence ratio between the sector sub - block and the target object edge according to the number of coincident pixel points on the edge; determine the proximity of the pixel points on the edge between the sector sub - block and the target object according to the coincidence ratio between the sector sub - block and the target object edge, the gradient between pixel points, and the gradient direction; take the reciprocal of the value obtained by adding 1 to the normalized curvature difference value of the pixel points on the edge between the sector sub - block and its corresponding target object as the similarity index between the sector sub - block and its corresponding target object, and record the product mean of the proximity of the pixel points on the edge between the sector sub - block and the target object and the similarity index as the hub similarity of the edge between the sector sub - block and the target object; calculate the elimination degree of the edge in the sector sub - block, where the elimination degree of the edge is negatively correlated with the hub similarity of the edge, and determine the edge elimination result of this sector sub - block; in response to the edge elimination result of the sector sub - block, take the sector sub - block after edge elimination as the target object of the remaining sector sub - blocks, and continue to perform edge elimination on the remaining sector sub - blocks to obtain the spoke edges in the automotive hub image.

[0009] The present invention takes into account that when extracting the spoke contour edges in an automotive hub image, there may be an influence of non - spoke edges. Therefore, eliminating the non - spoke edges in the hub image can accurately extract the spoke contour edges in the hub image. During the process of eliminating non - spoke edges, the present invention divides the hub image according to the symmetry characteristics of the spokes, and accurately calculates the elimination degree of each edge in the current sector sub - block by comprehensively analyzing features such as the gray - level difference and gradient difference between non - spoke edges and spoke edges in the sector sub - block and other sector sub - blocks, so that the elimination of non - spoke edges can be accurately achieved based on this. At the same time, when calculating the elimination degree of each edge in the current sector sub - block, the present invention takes into account that some spoke edges may have differences from normal spoke edges due to the influence of the acquisition environment, and their elimination degree may be too large. Therefore, when determining the elimination degree of the edge, the present invention also combines the curvature characteristics of the spoke edges to exclude the influence of the environment on the elimination result, thereby effectively improving the accuracy of the extraction result of the spoke contour edges in the automotive hub image.

[0010] According to an appearance contour extraction method for automotive parts processing provided by the present invention, before equally dividing the hub image into multiple sector sub - blocks according to the number of spokes in the automotive hub image, it further includes: collecting the original automotive hub image and extracting the edges in the original automotive hub image, locating and eliminating the central hole area of the automotive hub to obtain the automotive hub image.

[0011] An appearance contour extraction method for automobile part processing provided by the present invention, which locates the central hole area of the automobile wheel hub and eliminates it to obtain an automobile wheel hub image, includes: using Hough circle transformation to locate the central hole area in the original automobile wheel hub image and performing convex hull calculation to generate a convex polygon; eliminating the edge of the central hole area contained in the smallest convex polygon to obtain an automobile wheel hub image.

[0012] The present invention takes into account that the characteristics of the central hole area of the automobile wheel hub are relatively close to the characteristics of the outer contour of the wheel hub, and errors may occur when eliminating the edges of non-wheel hub spokes. Therefore, before eliminating the edges of non-wheel hub spokes, the central hole area of the automobile wheel hub is removed through preprocessing to reduce the interference of the central hole area of the automobile wheel hub.

[0013] An appearance contour extraction method for automobile part processing provided by the present invention takes the center of the wheel hub in the wheel hub image as the vertex of the fan-shaped sub-block.

[0014] An appearance contour extraction method for automobile part processing provided by the present invention, which determines the coincidence ratio between the fan-shaped sub-block and the edge of the target object according to the number of coincident pixel points on the edge, includes: taking the ratio of the number of coincident pixel points on one edge of the fan-shaped sub-block and the target object to the total number of pixel points on this edge as the coincidence ratio between the fan-shaped sub-block and this edge of the target object.

[0015] The present invention takes into account that in order to ensure the performance of the automobile, usually the size and shape between each wheel hub spoke of the automobile are highly similar. Therefore, by analyzing the number of coincident pixel points on the same-position edge of different fan-shaped sub-blocks to evaluate the coincidence ratio between the fan-shaped sub-block and the edge of its target object, the proximity degree between the edge in the fan-shaped sub-block and the edge at the same position in other fan-shaped sub-blocks can be initially judged.

[0016] An appearance contour extraction method for automobile part processing provided by the present invention, which determines the proximity degree of the pixel points on the edge between the fan-shaped sub-block and the target object according to the coincidence ratio between the fan-shaped sub-block and the edge of the target object and the gradient and gradient direction between pixel points, includes: calculating the proximity degree of the th pixel point on the th edge between a fan-shaped sub-block and the corresponding th target object :

[0017] ;

[0018] is the coincidence ratio between the fan-shaped sub-block and the th edge between the corresponding th target object, , are respectively the coincidence ratio between the fan-shaped sub-block and the The gradient difference and gradient direction difference of the pixel at the edge between a fan-shaped sub-block and a target object, is the standard normalization function.

[0019] The present invention provides an accurate calculation formula for the proximity of pixel points on the edge between a fan-shaped sub-block and a target object. By combining the coincidence ratio, gradient, and gradient direction of the edge between the fan-shaped sub-block and the target object, the value of the proximity can be accurately obtained.

[0020] According to an appearance contour extraction method for automobile part processing provided by the present invention, calculating the rejection degree of the edge in the fan-shaped sub-block includes: taking the reciprocal of the sum of the mean hub similarity of the edges between the fan-shaped sub-block and all target objects and 1 to obtain the rejection degree of this edge in this fan-shaped sub-block.

[0021] According to an appearance contour extraction method for automobile part processing provided by the present invention, determining the edge rejection result of the fan-shaped sub-block includes: rejecting the edges in the fan-shaped sub-block with a rejection degree greater than the rejection threshold to obtain the edge rejection result of the fan-shaped sub-block.

[0022] According to an appearance contour extraction method for automobile part processing provided by the present invention, after continuing to perform edge rejection on the remaining fan-shaped sub-blocks, obtaining the spoke edges in the automobile hub image includes: connecting the broken regions in the automobile hub image after edge rejection to obtain the complete spoke edges in the automobile hub image.

[0023] The present invention takes into account that some spoke edges of the hub may be missing or wrongly rejected during the process of edge extraction. Therefore, connecting the broken regions after obtaining the edge rejection result of the hub image can effectively ensure the continuity and integrity of the edge contour.

[0024] According to an appearance contour extraction method for automobile part processing provided by the present invention, connecting the broken regions in the automobile hub image after edge rejection includes: obtaining the endpoint gray values of the broken regions in the automobile hub image and the gray values of the pixel points in the blank region in the middle of the broken regions, and taking the pixel points in the blank region with a difference less than the gray threshold from the mean value of the endpoint gray values as target pixel points; connecting the endpoints of the broken regions with the target pixel points in the middle blank region.

[0025] The present invention has the following beneficial effects:

[0026] Based on the above technical solution, a method for extracting the appearance contour of automotive parts provided by the present invention, when extracting the outer contour of the automotive wheel hub, by removing the non-wheel-spoke edges in the wheel hub image, the contour edges of the wheel spokes in the wheel hub image can be accurately extracted, reducing the influence of non-wheel-spoke edges on the extraction result. During the process of removing non-wheel-spoke edges, the present invention divides the wheel hub image according to the symmetry characteristics of the wheel spokes, and accurately calculates the removal degree of each edge in the current fan-shaped sub-block by comprehensively analyzing the gray difference and gradient difference and other characteristics between the non-wheel-spoke edges and the wheel-spoke edges in the fan-shaped sub-block and other fan-shaped sub-blocks, so that the removal of non-wheel-spoke edges can be accurately realized based on this. At the same time, when calculating the removal degree of each edge in the current fan-shaped sub-block, the present invention takes into account that some wheel-spoke edges may have differences from normal wheel-spoke edges due to the influence of the acquisition environment, and their removal degree may be too large. Therefore, when determining the removal degree of the edge, the present invention also combines the curvature characteristics of the wheel-spoke edges to eliminate the influence of the environment on the removal result, thereby effectively improving the accuracy of the extraction result of the wheel-spoke contour edges in the automotive wheel hub image. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 It is a flowchart of the steps of a method for extracting the appearance contour of automotive parts provided by an embodiment of the present invention;

[0028] Figure 2 It is an automotive wheel hub image provided by an embodiment of the present invention;

[0029] Figure 3 It is a detection result of the edges of the wheel hub image provided by an embodiment of the present invention;

[0030] Figure 4 It is a schematic diagram of an automotive wheel hub image after removing the central hole area provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0031] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all of the embodiments.

[0032] In order to reduce the interference of noise edges when extracting the wheel-spoke edges in the automotive wheel hub image, an embodiment of the present invention provides a method for extracting the appearance contour of automotive parts. By analyzing the characteristic differences between the wheel-spoke edges and the non-wheel-spoke edges, the non-wheel-spoke edges can be accurately removed from the automotive wheel hub image, effectively improving the accuracy of the extraction of the automotive wheel hub appearance contour.

[0033] Specifically, please refer to Figure 1 , Figure 1The flowchart of steps for an appearance contour extraction method for automobile part processing provided by an embodiment of the present invention. This method includes the following steps:

[0034] S1: Obtain an image of an automobile wheel hub.

[0035] Exemplarily, in the embodiment of the present invention, obtaining an image of an automobile wheel hub includes: collecting an original image of an automobile wheel hub and extracting the edges in the original image of the automobile wheel hub, locating the central hole area of the automobile wheel hub and removing it to obtain an image of the automobile wheel hub.

[0036] Exemplarily, the image of the automobile wheel hub can also be grayscale processed.

[0037] Specifically, in the shooting environment, use a uniform annular light source or a combined light source. Install the camera directly above the wheel hub, adjust the camera's perspective to ensure that the entire wheel hub enters the camera's field of view, set the image resolution and pixels, and take an image of the automobile wheel hub. For details, please refer to Figure 2 as shown Figure 2 An image of an automobile wheel hub provided by an embodiment of the present invention. Combining Figure 2 it can be seen that the overall shape of the automobile wheel hub is a circular contour, and there are 5 groups of wheel hub spokes installed in the middle.

[0038] Among them, the image pixels can be set to 1024×768 pixels, and the image resolution can be set to 100 DPI; the image resolution and pixels can be specifically set according to actual needs, and the embodiments of the present invention do not limit this too much here.

[0039] Exemplarily, the image of the automobile wheel hub can be processed by the Canny edge detection algorithm to extract the edges in the original image of the automobile wheel hub. For details, please refer to Figure 3 as shown Figure 3 The edge detection result of a wheel hub image provided by an embodiment of the present invention. Combining Figure 3 it can be seen that each wheel hub spoke of the automobile wheel hub is approximately centrosymmetric. However, the central hole area in the image of the automobile wheel hub is not the same as the axis of symmetry of the wheel hub spokes, and there may be interference during subsequent image analysis.

[0040] Based on this, in the embodiment of the present invention, the central hole area in the original image of the automobile wheel hub can be removed first.

[0041] Exemplarily, in the embodiment of the present invention, locating the central hole area of the automobile wheel hub and removing it to obtain an image of the automobile wheel hub includes: using the Hough (also known as the Hough) circle transformation to locate the central hole area in the original image of the automobile wheel hub and performing a convex hull calculation to generate a convex polygon; removing the edges of the central hole area contained within the smallest convex polygon to obtain an image of the automobile wheel hub.

[0042] Among them, in order to avoid the incorrect elimination of other circles (such as the outer shape circle of the wheel hub) other than the central hole area when the Hough circle transform locates the central hole area in the original automobile wheel hub image, it is necessary to set the parameter radius of the Hough circle transform.

[0043] Specifically, the radius of the central hole of the automobile wheel hub is usually between 50 mm and 200 mm, and the corresponding pixel point range in a 1024×768 pixel image is approximately 197 to 787 pixels. In order to convert the range of 50 mm to 200 mm into the pixel point range in a 1024×768 pixel image, the parameter radius of the Hough circle transform for positioning can be set to [200, 800] pixels, so as to accurately locate the central hole area in the original automobile wheel hub image.

[0044] It can be understood that the convex hull is a convex polygon containing a set of points. Through convex hull calculation, the contour boundary of the target area can be generated, so as to describe the shape of the target area and accurately obtain the central hole area in the original automobile wheel hub image.

[0045] Exemplarily, when locating the central hole area in the original automobile wheel hub image and performing convex hull calculation to generate a convex polygon, the contour points in the original grayscale image of the automobile wheel hub can be extracted through an algorithm to obtain the minimum convex polygon, where the minimum convex polygon completely contains the central hole area in the original automobile wheel hub image.

[0046] Among them, the algorithm for extracting the contour points in the original grayscale image of the automobile wheel hub can be the Jarvis March algorithm, the Quick Hull algorithm, etc., and can be specifically set according to actual needs.

[0047] Specifically, reference can be made to Figure 4 as shown Figure 4 which is a schematic diagram of an automobile wheel hub image after removing the central hole area provided by an embodiment of the present invention. Combining Figure 3 and Figure 4 it can be seen that after removing the edge of the central hole area, the edge including the automobile wheel hub and the wheel hub spokes can be clearly seen in the automobile wheel hub image, so that the following steps can be continued based on this to extract the appearance contour of the automobile wheel hub.

[0048] S2: Divide the wheel hub image into a plurality of fan-shaped sub-blocks according to the number of wheel hub spokes in the automobile wheel hub image, and the vertex of the fan-shaped sub-block is the center of the wheel hub in the wheel hub image.

[0049] It should be noted that, in order to ensure the balanced mechanical properties and the dynamic balance of the wheel hub, usually, the wheel hub spoke area of the automobile consists of a plurality of identical spokes, and the size and shape of these spokes are exactly the same. The spoke patterns in the non-wheel hub spoke areas that belong to the noise edges in the wheel hub image are different from the normal wheel hub spokes.

[0050] Based on this, in the embodiments of the present invention, the hub image can be evenly divided into multiple fan-shaped sub-blocks according to the symmetry characteristics of the hub spokes in the hub image, and the non-hub spoke edges can be accurately screened according to the differences on the edges at the same position in different fan-shaped sub-blocks.

[0051] It can be understood that the contour edge of an automobile hub is a concentric circle, which is located in the outermost circle of the entire hub image and has the largest radius. And the hub spokes are usually connected to the hub center. Therefore, based on the shape characteristics of the automobile hub, the edge of the automobile hub can be located, so as to accurately divide the hub image.

[0052] Exemplarily, in the embodiments of the present invention, when evenly dividing the hub image into multiple fan-shaped sub-blocks according to the number of hub spokes in the automobile hub image, the Hough circle transform can be used to locate all the circular edges in the hub image and obtain the radii of all the circular edges. The area with the largest radius is used as the edge of the automobile hub, and the center of the circle of the automobile hub edge is used as the center of the hub in the hub image.

[0053] Exemplarily, in combination with Figure 4 It can be seen that when segmenting the schematic diagram of the hub image provided by the embodiments of the present invention based on the above method, the number of hub spoke edges in the automobile hub image is 5. Therefore, the automobile hub image can finally be divided into 5 fan-shaped sub-blocks.

[0054] After segmenting the hub image based on the above steps, the screening of the non-hub spoke edges can be realized according to the feature differences on the edges at the same position in different fan-shaped sub-blocks, that is, continue to execute the following steps.

[0055] S3: Respectively use the other fan-shaped sub-blocks outside a fan-shaped sub-block as the target object corresponding to this fan-shaped sub-block; overlap the edges in the fan-shaped sub-block and the corresponding target object, and calculate the proximity of the pixel points on the edges between the fan-shaped sub-block and each target object.

[0056] It should be noted that due to the symmetry characteristics of the automobile hub spokes, in the hub image without noise edge interference, the hub spoke edges in the fan-shaped sub-block and other fan-shaped sub-blocks can completely overlap, and the pixel points on the edges can also completely overlap. However, the appearance of noise edges is relatively random, and the possibility of appearing at the same position in other fan-shaped sub-blocks is lower. If the pixel points on the edge in the fan-shaped sub-block can find pixel points with the same position on the same edge in the corresponding target object, then the pixel point in the fan-shaped sub-block is the overlapping pixel point on the edge between the fan-shaped sub-block and the target object. The more the number of overlapping pixel points, the higher the possibility that this edge is the hub spoke edge and the lower the possibility that it is a random noise edge.

[0057] Illustrate the process of determining the coincident pixel points: When the current fan-shaped sub-block is Z1 and obtaining the coincident pixel points in the fan-shaped sub-block S1 during the process of comparing the fan-shaped sub-block Z1 with the corresponding target object Z2, the edge coinciding with the fan-shaped sub-block Z1 can be obtained at the same position in the target object Z2 first. If the third edge in the fan-shaped sub-block Z1 can find the third edge at the same position in the target object Z2, and the fourth pixel point on the third edge of the fan-shaped sub-block Z1 can find the corresponding fourth pixel point at the same position on the third edge of the target object Z2, then the fourth pixel point on the third edge of the fan-shaped sub-block Z1 is the coincident pixel point. By looping like this, the number of coincident pixel points on the edge between the fan-shaped sub-block Z1 and its target object Z2 can be finally obtained.

[0058] Based on this, in the embodiment of the present invention, other fan-shaped sub-blocks except the current fan-shaped sub-block can be used as the target object of the current fan-shaped sub-block for edge analysis, and the proximity of the pixel points on the edge between the fan-shaped sub-block and the target object can be determined according to the characteristics of the coincident pixel points on the edge between the fan-shaped sub-block and the target object.

[0059] Exemplarily, in the embodiment of the present invention, determining the proximity of the pixel points on the edge between the fan-shaped sub-block and the target object includes: determining the coincidence ratio of the fan-shaped sub-block and the edge of the target object according to the number of coincident pixel points on the edge between the fan-shaped sub-block and the target object, and determining the proximity of the pixel points on the edge between the fan-shaped sub-block and the target object according to the coincidence ratio of the fan-shaped sub-block and the edge of the target object and the gradient and gradient direction between the pixel points.

[0060] Exemplarily, in the embodiment of the present invention, determining the coincidence ratio of the fan-shaped sub-block and the edge of the target object according to the number of coincident pixel points on the edge includes: taking the ratio of the number of coincident pixel points on one edge between the fan-shaped sub-block and the target object to the total number of pixel points on this edge as the coincidence ratio of the fan-shaped sub-block and this edge of the target object.

[0061] It can be understood that the more the number of coincident pixel points on one edge between the current fan-shaped sub-block and the target object, the higher the coincidence degree of the current fan-shaped sub-block and this edge of the target object, and the higher the possibility that this edge conforms to the edge characteristics of the wheel hub spoke.

[0062] It should be further noted that some randomly generated non-wheel hub spoke edges in two fan-shaped sub-blocks may coincide with the wheel hub spoke edge, and the obtained coincidence ratio will also be relatively high. When the edge lines at the same position on different fan-shaped sub-blocks are all wheel hub edges, their gray-scale characteristics and extension directions will be highly similar. Therefore, in the embodiment of the present invention, by further analyzing the gray-scale characteristics and gradient characteristics of the pixel points on the edges at the same position on different fan-shaped sub-blocks, the proximity of the pixel points on the edge between the fan-shaped sub-block and the target object can be accurately obtained.

[0063] Among them, if a pixel point on the edge in a fan-shaped sub-block has no corresponding pixel point at the same position in the target object, the proximity of this pixel point can be recorded as 0.

[0064] Exemplarily, in the embodiment of the present invention, to determine the proximity of pixel points on the edge between a fan-shaped sub-block and the target object, the following relational expression can be specifically referred to:

[0065] ;

[0066] represents the proximity of the th pixel point on the th edge between a fan-shaped sub-block and the corresponding th target object, is the coincidence ratio of the th edge between the fan-shaped sub-block and the corresponding th target object, is the gradient difference of the th pixel point on the th edge between the fan-shaped sub-block and the corresponding th target object, is the gradient direction difference of the th pixel point on the th edge between the fan-shaped sub-block and the corresponding th target object, is the standard normalization function.

[0067] In the above formula The larger it is, the higher the coincidence degree of the th edge between the current fan-shaped sub-block and the corresponding th target object. However, if and are also large, it means that the gradient and gradient direction of the th pixel point on the th edge between the current fan-shaped sub-block and the corresponding th target object have a greater difference. The possibility that the th edge in the current fan-shaped sub-block is a non-wheel hub spoke edge is greater. Therefore, it is necessary to reduce the proximity of the th pixel point on the th edge between the current fan-shaped sub-block and the corresponding th target object, so as to accurately evaluate the possibility that the th edge between the fan-shaped sub-block and the corresponding th target object is a wheel hub spoke edge.

[0068] The proximity between the fan-shaped sub-block and the corresponding th target object on the The proximity of the th pixel point on the edge is used to characterize the similarity between the fan-shaped sub-block and the corresponding th target object at the th edge and the possibility that the th pixel point is a pixel point on the edge of the wheel hub spoke.

[0069] S4: Determine the wheel hub similarity of the edge between the fan-shaped sub-block and the target object according to the proximity and curvature difference of the pixel points on the edge between the fan-shaped sub-block and each target object; calculate the rejection degree of the edge in the fan-shaped sub-block, and determine the edge rejection result of the fan-shaped sub-block.

[0070] Among them, the rejection degree of the edge is negatively correlated with the wheel hub similarity of the edge.

[0071] It should be noted that due to the influence of the acquisition illumination, there will also be differences in the gray levels between some edges of the wheel hub spokes, but the curvatures between the edges of the wheel hub spokes are consistent. Therefore, in the embodiments of the present invention, the proximity of the pixel points can be further verified by obtaining the curvature characteristics of the edge between the fan-shaped sub-block and the target object, so as to accurately obtain the wheel hub similarity of the edge between the fan-shaped sub-block and the corresponding target object.

[0072] Exemplarily, in the embodiments of the present invention, the similarity of the pixel points on the edge between the fan-shaped sub-block and each target object can be determined according to the proximity and curvature difference of the pixel points on the edge between the fan-shaped sub-block and each target object, and the average value of the similarities of all pixel points on the edge between the fan-shaped sub-block and the corresponding target object is used as the wheel hub similarity of the edge between the fan-shaped sub-block and the corresponding target object.

[0073] Exemplarily, when calculating the wheel hub similarity of the edge between the fan-shaped sub-block and the corresponding target object, the reciprocal of the value obtained by adding 1 to the normalized curvature difference value of the pixel points on the edge between the fan-shaped sub-block and its corresponding target object can be used as the similarity index between the fan-shaped sub-block and its corresponding target object, and the product mean of the proximity of the pixel points on the edge between the fan-shaped sub-block and the target object and the similarity index is denoted as the wheel hub similarity of the edge between the fan-shaped sub-block and the target object.

[0074] Among them, the product of the proximity of the pixel points on the edge between the fan-shaped sub-block and the target object and the similarity index is the similarity of the pixel points on the edge between the fan-shaped sub-block and the target object.

[0075] Exemplarily, when calculating the similarity of the pixel points on the edge between the fan-shaped sub-block and the target object, the following relational expression can be specifically referred to:

[0076] ;

[0077] is the similarity of the th pixel point on the th edge between the fan-shaped sub-block and the corresponding th target object, is the proximity of the th pixel point on the th edge between the fan-shaped sub-block and the corresponding th target object, is the curvature of the th pixel point on the th edge of the fan-shaped sub-block, is the curvature of the th pixel point on the th edge of the th target object corresponding to the fan-shaped sub-block, is the absolute value symbol, is the standard normalization function.

[0078] Among them, the curvature on the edge where the th pixel point is located can be used as the curvature of the th pixel point.

[0079] In the above formula, represents the curvature difference between the th pixel points on the th edge between the fan-shaped sub-block and the corresponding th target object. The greater the difference, the lower the possibility that the th edge between the fan-shaped sub-block and the corresponding th target object is the edge of the wheel hub spoke.

[0080] represents the normalized curvature difference value of the th pixel points on the th edge between the current fan-shaped sub-block and the corresponding th target object, represents the similarity index of the th pixel points on the th edge between the current fan-shaped sub-block and the corresponding th target object.

[0081] Based on the above steps, after obtaining the similarities of all pixel points on the th edge between the fan-shaped sub-block and the corresponding th target object, the average value of the similarities of all pixel points on the th edge between the fan-shaped sub-block and the corresponding th target object can be used as the similarity between the fan-shaped sub-block and the corresponding th target object for the Similarity of the edges of a hub

[0082] It should be noted that there may be non-hub spoke edges in the current sector sub-block and the target object itself. If the edges in the current sector sub-block are determined to be removed only by a single target object, the hub spoke edges in the current sector sub-block may be removed.

[0083] Based on this, the embodiments of the present invention can obtain the hub similarity between the edges of the sector sub-block and the edges of all corresponding target objects, and obtain the removal degree of the edge in the sector sub-block, so as to accurately remove the non-hub spoke edges.

[0084] Exemplarily, in the embodiments of the present invention, calculating the removal degree of the edge in the sector sub-block includes: taking the reciprocal of the sum value of the mean value of the hub similarity between the edges of the sector sub-block and all target objects and 1, to obtain the removal degree of the edge in the sector sub-block.

[0085] An example of the method for obtaining the removal degree of the 3rd edge in the current sector sub-block: Obtain the hub similarity between the current sector sub-block and the 3rd edge in the corresponding target object in the target object corresponding to the current sector sub-block; Calculate the removal degree of the 3rd edge in the current sector sub-block according to the mean value of the hub similarity between the current sector sub-block and the 3rd edge in the corresponding target object.

[0086] Exemplarily, in the embodiments of the present invention, determining the edge removal result of the sector sub-block includes: removing the edges in the sector sub-block with a removal degree greater than the removal threshold, to obtain the edge removal result of the sector sub-block.

[0087] Among them, the removal threshold can be set to 0.8; The removal threshold can be specifically set according to actual needs, and the embodiments of the present invention do not limit it too much here.

[0088] After accurately removing the non-hub spoke edges in the current sector sub-block based on the above steps, the current sector sub-block not interfered by noise edges can be used as the comparison object of the remaining sector sub-blocks, and all non-hub spoke edges in the hub image can be dynamically and cyclically removed.

[0089] S5: In response to the edge removal result of the sector sub-block, taking the sector sub-block after removing the edges as the target object of the remaining sector sub-blocks, and continuing to remove the edges of the remaining sector sub-blocks, to obtain the hub spoke edges in the automotive hub image.

[0090] It should be noted that when removing the non-spoke edges in the automotive wheel hub image, the spoke edges in the automotive wheel hub image may be defectively removed due to factors such as excessive environmental brightness and incorrect edge recognition. Based on this, the embodiments of the present invention can connect the broken regions in the finally completed removed wheel hub image to connect the possibly broken spoke edges into complete spoke edges.

[0091] Exemplarily, in the embodiments of the present invention, after continuing to remove the edges of the remaining sector sub-blocks, the spoke edges in the automotive wheel hub image are obtained, including: connecting the broken regions in the automotive wheel hub image after edge removal is completed to obtain the complete spoke edges in the automotive wheel hub image.

[0092] Exemplarily, in the embodiments of the present invention, connecting the broken regions in the automotive wheel hub image after edge removal is completed includes: obtaining the grayscale values of the end points of the broken regions in the automotive wheel hub image and the grayscale values of the pixel points in the blank region in the middle of the broken regions, and taking the pixel points in the blank region with a difference less than the grayscale threshold from the average value of the grayscale values of the end points as target pixel points; connecting the end points of the broken regions with the target pixel points in the middle blank region.

[0093] Wherein, the difference is the absolute value of the grayscale difference between the pixel points in the blank region and the average value of the grayscale values of the end points. The grayscale threshold can be set to 50 and can be specifically set according to needs.

[0094] Exemplarily, the broken region repair can be completed through morphological operations and dynamic programming algorithms, which can be specifically set according to actual needs, and the embodiments of the present invention do not limit this too much here.

[0095] It can be seen that in the embodiment of the present invention, when extracting the appearance contour of an automotive wheel hub, the wheel hub image can be evenly divided into multiple fan-shaped sub-blocks according to the number of spokes in the automotive wheel hub image; the other fan-shaped sub-blocks except one fan-shaped sub-block are respectively used as the target objects corresponding to the fan-shaped sub-block; the fan-shaped sub-block is overlapped with the edge in the corresponding target object, and the coincidence ratio between the fan-shaped sub-block and the edge of the target object is determined according to the number of coincident pixel points on the edge; according to the coincidence ratio between the fan-shaped sub-block and the edge of the target object, as well as the gradient and gradient direction between pixel points, the proximity of pixel points on the edge between the fan-shaped sub-block and the target object is determined; the reciprocal of the value obtained by adding 1 to the normalized curvature difference value of pixel points on the edge between the fan-shaped sub-block and its corresponding target object is used as the similarity index between the fan-shaped sub-block and its corresponding target object, and the product mean of the proximity of pixel points on the edge between the fan-shaped sub-block and the target object and the similarity index is recorded as the wheel hub similarity of the edge between the fan-shaped sub-block and the target object; the elimination degree of the edge in the fan-shaped sub-block is calculated, and the elimination degree of the edge is negatively correlated with the wheel hub similarity of the edge, and the edge elimination result of the fan-shaped sub-block is determined; in response to the edge elimination result of the fan-shaped sub-block, the fan-shaped sub-block after eliminating the edge is used as the target object of the remaining fan-shaped sub-blocks, and after continuing to perform edge elimination on the remaining fan-shaped sub-blocks, the spoke edges in the automotive wheel hub image are obtained, effectively improving the accuracy of spoke contour extraction in the automotive wheel hub image.

[0096] The above are only the preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. An appearance contour extraction method for automobile part processing, characterized in that, Including: Dividing the wheel hub image into multiple fan-shaped sub-blocks evenly according to the number of spokes in the automotive wheel hub image; Regarding the other fan-shaped sub-blocks except one fan-shaped sub-block as the target object corresponding to this fan-shaped sub-block; overlapping the edge of the fan-shaped sub-block with the corresponding target object, and determining the coincidence ratio between the fan-shaped sub-block and the edge of the target object according to the number of overlapping pixel points on the edge; determining the proximity of the pixel points on the edge between the fan-shaped sub-block and the target object according to the coincidence ratio between the fan-shaped sub-block and the edge of the target object and the gradient and gradient direction between pixel points, including: Calculate the proximity of the th pixel on the th edge between a fan-shaped sub-block and the corresponding th target object : ； is the coincidence ratio of the th edge between the fan-shaped sub-block and the corresponding th target object, , are respectively the gradient difference and the gradient direction difference of the th pixel on the th edge between the fan-shaped sub-block and the corresponding th target object, and is the standard normalization function; Taking the reciprocal of the value obtained by adding 1 to the normalized curvature difference value of the pixel points on the edge between the fan-shaped sub-block and its corresponding target object as the similarity index between the fan-shaped sub-block and its corresponding target object, and recording the product mean of the proximity of the pixel points on the edge between the fan-shaped sub-block and the target object and the similarity index as the wheel hub similarity of the edge between the fan-shaped sub-block and the target object; calculating the elimination degree of the edge in the fan-shaped sub-block, where the elimination degree of the edge is negatively correlated with the wheel hub similarity of the edge, and determining the edge elimination result of this fan-shaped sub-block, including: Eliminating the edges in the fan-shaped sub-block whose elimination degree is greater than the elimination threshold to obtain the edge elimination result of this fan-shaped sub-block; In response to the edge elimination result of the fan-shaped sub-block, regarding the fan-shaped sub-block after eliminating the edges as the target object of the remaining fan-shaped sub-blocks, and continuing to perform edge elimination on the remaining fan-shaped sub-blocks to obtain the spoke edges in the automotive wheel hub image.

2. The appearance contour extraction method for automobile part processing according to claim 1, characterized in that, Before the step of dividing the wheel hub image into multiple fan-shaped sub-blocks evenly according to the number of spokes in the automotive wheel hub image, it further includes: Collecting the original automotive wheel hub image and extracting the edges in the original automotive wheel hub image, positioning and eliminating the central hole area of the automotive wheel hub to obtain the automotive wheel hub image.

3. A method for extracting the appearance contour of an automotive part according to claim 2, characterized in that, The step of positioning and eliminating the central hole area of the automotive wheel hub to obtain the automotive wheel hub image includes: Using the Hough circle transformation to locate the central hole area in the original automotive wheel hub image and performing convex hull calculation to generate a convex polygon; eliminating the edge of the central hole area included in the smallest convex polygon to obtain the automotive wheel hub image.

4. A method for extracting the appearance contour of an automotive part according to claim 1, characterized in that Taking the center of the wheel hub in the wheel hub image as the vertex of the fan-shaped sub-block.

5. A method for extracting the appearance contour for machining automotive parts according to claim 1, characterized in that, The step of determining the coincidence ratio between the fan-shaped sub-block and the edge of the target object according to the number of overlapping pixel points on the edge includes: Taking the ratio of the number of overlapping pixel points on one edge between the fan-shaped sub-block and the target object to the total number of pixel points on this edge as the coincidence ratio between the fan-shaped sub-block and this edge of the target object.

6. The appearance contour extraction method for automotive part processing according to claim 1, wherein The step of calculating the elimination degree of the edge in the fan-shaped sub-block includes: Taking the reciprocal of the sum value of the mean of the wheel hub similarities of the edges between the fan-shaped sub-block and all target objects and 1 to obtain the elimination degree of this edge in the fan-shaped sub-block.

7. A method for extracting the appearance contour of an automotive part according to claim 1, characterized in that The step of continuing to perform edge elimination on the remaining fan-shaped sub-blocks to obtain the spoke edges in the automotive wheel hub image includes: Connecting the broken areas in the automotive wheel hub image after completing edge elimination to obtain the complete spoke edges in the automotive wheel hub image.

8. A method for extracting the appearance contour for machining automotive parts according to claim 7, characterized in that, The step of connecting the broken areas in the automotive wheel hub image after completing edge elimination includes: Obtain the gray values of the end points of the fracture region and the gray values of the pixel points in the blank region in the middle of the fracture region of the automobile wheel hub image, and take the pixel points in the blank region whose difference from the average value of the gray values of the end points is less than the gray threshold as the target pixel points; connect the end points of the fracture region with the target pixel points in the middle blank region.

Images