Intelligent identification method for tooth consistency of bulldozer tooth block

Abstract

The invention relates to the technical field of image processing, in particular to a tooth consistency intelligent identification method for bulldozer tooth blocks, which comprises the following steps of: obtaining a reference connected domain according to the gray distribution of pixel points in a preset local window of all pixel points in each defect connected domain and the difference between data in an edge chain code sequence of each defect connected domain; obtaining a target connected domain according to the gradient change vector of each reference connected domain and the difference between the decline vectors of all edge pixel points in each reference connected domain; obtaining a projection vector of each reference connected domain according to the decline vector; and obtaining a tooth consistency identification result of the bulldozer tooth block according to the difference between the projection vector and the gradient change vector of any two target connected domains and the distance between any two target connected domains. According to the invention, the accuracy of tooth crack detection of the bulldozer tooth block is improved, and the accuracy of the tooth consistency identification result of the bulldozer tooth block is also improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of image processing, in particular to a tooth consistency intelligent identification method for a bulldozer tooth block. BACKGROUND

[0002] Tooth is one of the key components of the bulldozer tooth block, directly contacts and acts on the ground or working material, and undertakes important tasks such as excavation, pushing and loading. Tooth will be subjected to great wear and impact during work. When the tooth of the bulldozer tooth block has a crack, it may cause safety hazards such as breakage and falling during operation of the bulldozer. Therefore, crack identification and detection of the tooth during production is particularly important.

[0003] When identifying and detecting the crack of the tooth in the production process, the crack of the tooth can be detected through edge detection. However, since there are concave and convex regions on the tooth in addition to the crack, the existence of the concave and convex regions will affect the detection of the crack, thereby reducing the accuracy of the crack detection of the tooth of the bulldozer tooth block, and thus reducing the accuracy of the tooth consistency identification result of the bulldozer tooth block. SUMMARY

[0004] The present application provides a tooth consistency intelligent identification method for a bulldozer tooth block to solve the existing problems.

[0005] The tooth consistency intelligent identification method for a bulldozer tooth block of the present application adopts the following technical scheme: One embodiment of the present application provides a tooth consistency intelligent identification method for a bulldozer tooth block, which comprises the following steps: Obtaining a tooth image of a bulldozer tooth block; Obtaining a plurality of defect connected domains in the tooth image, obtaining an edge chain code sequence of each defect connected domain, and screening a reference connected domain from all defect connected domains according to the gray scale distribution of the pixel points in the preset local window of all pixel points in each defect connected domain and the difference between different data in each edge chain code sequence; Obtaining a gradient change vector of each reference connected domain and a decreasing vector of each edge pixel point in each reference connected domain, and screening a target connected domain from all reference connected domains according to the difference between the gradient change vector of each reference connected domain and the decreasing vector of all edge pixel points in each reference connected domain; Obtaining a projection vector of each reference connected domain according to the decreasing vector of all edge pixel points in each reference connected domain, and obtaining a tooth consistency identification result of the bulldozer tooth block by merging all target connected domains according to the difference between the projection vectors of any two target connected domains, the difference between the gradient change vectors of any two target connected domains, and the distance between the center points of any two target connected domains.

[0006] Furthermore, the specific steps for obtaining several defect connected regions in the tooth image and obtaining the edge chain code sequence of each defect connected region are as follows: The edge detection of the bulldozer tooth block image is performed using the Canny edge detection algorithm. Then, the edge detection results of the bulldozer tooth block image are processed by binarization dilation to obtain several defect connected components. The edge chain code sequence of each defective connected component is obtained using the Freeman chain code algorithm.

[0007] Furthermore, the specific steps for selecting reference connected components from all defective connected components based on the grayscale distribution of pixels within a preset local window of all pixels in each defective connected component and the differences between different data in each edge chain code sequence are as follows: Based on the grayscale distribution of pixels within the local window of all pixels in each defect connected region, the probability that each defect connected region belongs to a crack defect can be obtained. The degree to which each defective connected region is a crack defect is obtained by considering the differences between adjacent data in the edge chain code sequence of each defective connected region and the differences between each data in the edge chain code sequence and the mean of all data in the edge chain code sequence. Based on the probability that each defect connected region belongs to a crack defect and the degree to which each defect connected region is a crack defect, the crack degree of each defect connected region is obtained. Among them, the probability that each defect connected region belongs to a crack defect is positively correlated with the crack degree of each defect connected region, and the degree to which each defect connected region is a crack defect is positively correlated with the crack degree of each defect connected region. If the degree of crack is greater than a preset first threshold All defective connected components are denoted as reference connected components.

[0008] Furthermore, the specific calculation method for determining the probability that each defect connected region belongs to a crack defect based on the grayscale distribution of pixels within a local window of all pixels in each defect connected region is as follows: In the formula, Indicates the first The first defective connected component The pixels within the local window of the nth pixel belong to the nth pixel. The average grayscale value of all pixels in a defective connected region. Indicates the first The first defective connected component Pixels within a local window of pixel number 1 do not belong to the 1st pixel. The average grayscale value of all pixels in a defective connected region. Indicates the first The number of all pixels in a defective connected domain. It is the absolute value symbol. This represents an exponential function with the natural constant as its base. Indicates the first The probability that a defective connected domain belongs to a crack defect.

[0009] Furthermore, the specific calculation method for determining the degree to which each defective connected region is a crack defect based on the differences between adjacent data in the edge chain code sequence of each defective connected region and the differences between each data in the edge chain code sequence of each defective connected region and the mean of all data in the edge chain code sequence is as follows: In the formula, Indicates the first The first edge chain code sequence of the defective connected component One data point, Indicates the first The first edge chain code sequence of the defective connected component One data point, Indicates the first The mean of all data in the edge chain code sequence of a defective connected component. Indicates the first The number of all data in the edge chain code sequence of a defective connected component It is the absolute value symbol. Represents a linear normalization function. Indicates the first Each defect connected region represents the degree of the crack defect.

[0010] Furthermore, the specific steps for obtaining the gradient change vector of each reference connected component are as follows: Obtain the minimum bounding rectangle of each reference connected component; draw a line segment through the center point of the minimum bounding rectangle that is parallel to the short side and has the same length as the short side, and denote it as the first line segment of the minimum bounding rectangle of each reference connected component; draw a line segment through the center point of the minimum bounding rectangle that is parallel to the long side and has the same length as the long side, and denote it as the second line segment of the minimum bounding rectangle of each reference connected component. Divide the minimum bounding rectangle of each reference connected component into two target regions based on the first line segment of the minimum bounding rectangle. Calculate the mean gradient magnitude of all pixels in each target region. The target region with the largest mean gradient magnitude is designated as the first target region, and the target region with the smallest mean gradient magnitude is designated as the second target region. According to the second line segment of the minimum circumscribed rectangle of each reference connected domain, the first target region and the second target region, a gradient change vector of each reference connected domain is obtained, wherein a length of the gradient change vector of each reference connected domain is equal to the length of the second line segment of the minimum circumscribed rectangle of each reference connected domain, and a direction of the gradient change vector of each reference connected domain is from the first target region to the second target region.

[0011] Further, the obtaining of the decreasing vector of each edge pixel point in each reference connected domain comprises the following specific steps: An edge pixel point in each reference connected domain is obtained, a pixel point with a gradient amplitude smaller than each edge pixel point and with a minimum gradient amplitude in the eight-neighbor domain of each edge pixel point is obtained and recorded as a target pixel point of each edge pixel point, and a direction of each edge pixel point pointing to the corresponding target pixel point is taken as a direction of the decreasing vector of each edge pixel point, wherein if the target pixel point of each edge pixel point belongs to the 4-neighbor pixel point of the edge pixel point, a length of the decreasing vector of each edge pixel point is a unit length 1, and if the target pixel point of each edge pixel point belongs to the D-neighbor pixel point of the edge pixel point, the length of the decreasing vector of each edge pixel point is ; When there is no target pixel point meeting the condition in the eight-neighbor domain of each edge pixel point, the decreasing vector of each edge pixel point is replaced by a 0 vector.

[0012] Further, the screening of the target connected domain from all the reference connected domains according to the gradient change vector of each reference connected domain and the difference between the decreasing vectors of all the edge pixel points in each reference connected domain comprises the following specific calculation method: In the formula, the decreasing vector of the i-th edge pixel point in the j-th reference connected domain is represented as the gradient change vector of the j-th reference connected domain is represented as the length of the decreasing vector of the i-th edge pixel point in the j-th reference connected domain is represented as the cosine similarity between the decreasing vector of the i-th edge pixel point in the j-th reference connected domain and the gradient change vector of the j-th reference connected domain is represented as the number of all the edge pixel points in the j-th reference connected domain is represented as the linear normalization function is represented as ​​​​​​​​​ Indicates the first The probability that a reference connected region is a cracked connected region; The probability that the reference connected component is a fractured connected component is greater than a preset second threshold. All reference connected components are denoted as the target connected component.

[0013] Furthermore, the specific steps for obtaining the projection vector of each reference connected region based on the decreasing vector of all edge pixels in each reference connected region are as follows: Will Recorded as the number The first reference connected component The decreasing projection vector of the i-th edge pixel, for the i-th edge pixel Perform vector operations on the decreasing projection vectors of all edge pixels in the nth reference connected region, and denote the result of the vector operation as the nth... Projection vectors of each reference connected component; In the formula, Indicates the first The magnitude of the gradient change vector of a reference connected region.

[0014] Furthermore, the specific steps for merging all target connected components based on the difference between the projection vectors of any two target connected components, the difference between the gradient change vectors of any two target connected components, and the distance between the center points of any two target connected components to obtain the tooth consistency recognition result of the bulldozer tooth block are as follows: The probability of merging any two target connected components is obtained based on the difference between the projection vectors of any two target connected components, the difference between the gradient change vectors of any two target connected components, and the distance between the center points of any two target connected components. The specific calculation method is as follows: In the formula, Indicates the first The target connected component and the first The distance between the center points of each target connected component Indicates the first The target connected component and the first The angle between the projection vectors of the target connected components. Indicates the first The target connected component and the first The angle between the gradient change vectors of the target connected components. This represents an exponential function with the natural constant as its base. Indicates the first The target connected component and the first The possibility of merging between target connected components; The probability of merging is greater than the preset third threshold. Merge the two target connected components and denote all the target connected components after the merger as the crack connected components; The ratio between the area of ​​all connected regions of cracks and the area of ​​the bulldozer tooth image is denoted as the crack area ratio; when the crack area ratio is less than a preset fourth threshold... If the identified bulldozer teeth are consistent, then the crack area ratio is greater than or equal to the preset fourth threshold. If so, the identified bulldozer teeth are inconsistent.

[0015] The beneficial effects of the technical solution of this invention are as follows: This invention obtains the crack degree of each defective connected region based on the difference between the gray values ​​of pixels in each defective connected region and the gray values ​​of surrounding pixels, and the smoothness of the shape of each defective connected region. This reduces the influence of uneven regions on the detection of crack defects. Based on the difference between the gradient change vector of each reference connected region and the decreasing vector of all edge pixels within each reference connected region, the probability that each reference connected region is a cracked connected region is obtained. This eliminates the influence of some connected regions on the tooth surface related to crack types on the cracked connected region. Based on any two target connected regions... The difference between the projection vector and the gradient change vector, and the distance between the center points of any two target connected regions are used to obtain the merging probability between any two target connected regions. The merging probability between any two target connected regions reduces the interference of noise. All target connected regions are merged according to the merging probability between any two target connected regions to obtain all crack connected regions, thus completing the crack defect detection in the teeth of the bulldozer tooth block. The final consistency recognition result is obtained according to the proportion of the area of ​​all crack connected regions in the tooth image area, which improves the accuracy of crack detection in the teeth of the bulldozer tooth block and also improves the accuracy of the consistency recognition result of the bulldozer tooth block. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 This is a flowchart illustrating the steps of the intelligent identification method for tooth consistency of bulldozer tooth blocks according to the present invention. Figure 2 This is a flowchart for identifying the tooth consistency of bulldozer tooth blocks. DETAILED DESCRIPTION

[0018] In order to further clarify the technical means and effects taken by the present application to achieve the predetermined inventive purpose, the specific embodiments, structures, features and effects of the tooth consistency intelligent identification method for bulldozer tooth blocks according to the present application are described in detail below in combination with the drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. In addition, the specific features, structures or characteristics in one or more embodiments can be combined in any suitable form.

[0019] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs.

[0020] The specific scheme of the tooth consistency intelligent identification method for bulldozer tooth blocks provided by the present application is described below in combination with the drawings.

[0021] Please refer to Figure 1 which shows the step flowchart of the tooth consistency intelligent identification method for bulldozer tooth blocks provided by one embodiment of the present application, which includes the following steps: Step S001: Collecting the tooth image of the bulldozer tooth block.

[0022] It should be noted that in order to improve the durability of the teeth of the bulldozer tooth block and ensure the safety of the bulldozer during operation, crack defect detection is performed on the teeth of the bulldozer tooth block during production, and the image of the teeth needs to be collected before the crack detection of the teeth.

[0023] Specifically, the color image of the teeth of the bulldozer tooth block is collected by an industrial camera, the color image of the teeth is preprocessed by grayscale, the tooth grayscale image is obtained, and the foreground area of the tooth in the tooth grayscale image is extracted by a semantic segmentation algorithm, which is recorded as a tooth image.

[0024] Among them, the grayscale preprocessing of the tooth color image is a known technology, which will not be described in detail here; It should be noted that in this embodiment, DeepLabV3 neural network is used as the neural network model of the semantic segmentation algorithm, and cross-entropy loss function is used as the loss function of DeepLabV3 neural network. The neural network is a known technology, and the specific results and training method of the network will not be described in this embodiment.

[0025] At this point, the tooth image of the bulldozer tooth block is obtained.

[0026] Step S002: Obtain a plurality of defect connected domains in the tooth image, and obtain an edge chain code sequence of each defect connected domain; according to the gray scale distribution of the pixel points in the preset local window of all pixel points in each defect connected domain and the difference between different data in each edge chain code sequence, a reference connected domain is selected from all defect connected domains.

[0027] It should be noted that when detecting the cracks in the teeth of the bulldozer tooth block, due to the interference of the concave-convex region, when detecting the crack defects in the tooth image of the bulldozer tooth block by edge detection, the detected may be the concave-convex region. Therefore, according to the gray scale difference and shape difference between the crack region and the concave-convex region, the crack region and the concave-convex region are distinguished.

[0028] Specifically, the edge of the tooth image of the bulldozer tooth block is detected by the Canny edge detection algorithm, and then the edge detection result of the tooth image of the bulldozer tooth block is processed by binary dilation to obtain a plurality of defect connected domains.

[0029] Among them, Canny edge detection algorithm and binary dilation are all known technologies, which will not be described in detail here.

[0030] It should be noted that the protruding part of the concave-convex region may have a higher gray value, and the recessed part may have a lower gray value, so the gray value of the concave-convex region is greatly different from the gray value of the pixel points of the tooth image of the bulldozer tooth block. The crack usually represents the break or crack of the surface of the object. Since the crack may cause light scattering or shadow, the gray value of the crack region is low, because they may reflect or absorb less light, so the gray value of the crack region is less different from the gray value of the pixel points of the tooth image of the bulldozer tooth block.

[0031] Specifically, a parameter is preset, and the embodiment is described taking as an example, which is not limited specifically, wherein may be determined according to the specific implementation. Taking any one pixel point in the tooth image of the bulldozer tooth block as a local window center point, and taking as the size of the local window, the local window of any one pixel point in the tooth image of the bulldozer tooth block is obtained.

[0032] According to the gray scale distribution of the pixel points in the local window of all pixel points in each defect connected domain, the possibility that each defect connected domain belongs to the crack defect is obtained, and as an embodiment, the specific calculation method is as follows: In the formula, represents the first defect connected domain, and The pixels within the local window of the nth pixel belong to the nth pixel. The average grayscale value of all pixels in a defective connected region. Indicates the first The first defective connected component Pixels within a local window of pixel number 1 do not belong to the 1st pixel. The average grayscale value of all pixels in a defective connected region. Indicates the first The number of all pixels in a defective connected domain. It is the absolute value symbol. This represents an exponential function with the natural constant as its base. Indicates the first The probability that a defective connected domain belongs to a crack defect.

[0033] in, This represents the difference between the gray values ​​of all pixels in each defective connected region and the gray values ​​of the pixels surrounding the defective connected region. The larger the difference, the greater the probability that the defective connected region is a concave-convex region, and the smaller the probability that it is a crack defect; the smaller the difference, the smaller the probability that the defective connected region is a concave-convex region, and the greater the probability that it is a crack defect.

[0034] Thus, we obtain the probability that each defect connected domain belongs to a crack defect.

[0035] It should be noted that cracked areas are usually linear or irregular, and they may have a slender shape or a tortuous outline; while uneven areas are usually protrusions or depressions on the surface of an object, and may form various shapes such as circles, ovals, and squares. The edges of uneven areas are usually relatively smooth and do not show characteristics of breakage or forking; therefore, uneven areas are relatively smooth compared to cracked areas.

[0036] Specifically, the edge chain code sequence of each defective connected domain is obtained through the Freeman chain code algorithm; the Freeman chain code algorithm is a well-known technique and will not be described in detail here.

[0037] The degree to which each defective connected region is a crack defect is obtained based on the differences between adjacent data in the edge chain code sequence of each defective connected region, and the differences between each data in the edge chain code sequence of each defective connected region and the mean of all data in the edge chain code sequence. As an example, the specific calculation method is as follows: In the formula, Indicates the first The first edge chain code sequence of the defective connected component One data point, the i-th data in the edge chain code sequence of the i-th defect connected domain, the i-th data in the edge chain code sequence of the i-th defect connected domain, the mean value of all data in the edge chain code sequence of the i-th defect connected domain, the number of all data in the edge chain code sequence of the i-th defect connected domain, the absolute value symbol, the linear normalization function, the degree of the i-th defect connected domain being a crack defect. the degree of the i-th defect connected domain being a crack defect.

[0038] wherein, the mean value of all differences between adjacent data in the edge chain code sequence of each defect connected domain, when the mean value of the differences is greater, it indicates that the defect connected domain is less smooth, and the degree of the defect connected domain being a crack defect is greater; when the mean value of the differences is smaller, it indicates that the defect connected domain is more smooth, and the degree of the defect connected domain being a crack defect is smaller, i.e. the degree of the defect connected domain being a concave-convex region is greater. the accumulated sum of all differences between all data and the mean value of all data in the edge chain code sequence of each defect connected domain, when the accumulated sum of the differences is greater, it indicates that the defect connected domain is less smooth, and the degree of the defect connected domain being a crack defect is greater; when the accumulated sum of the differences is smaller, it indicates that the defect connected domain is more smooth, and the degree of the defect connected domain being a crack defect is smaller, i.e. the degree of the defect connected domain being a concave-convex region is greater.

[0039] Thus, the degree of each defect connected domain being a crack defect is obtained.

[0040] According to the possibility of each defect connected domain belonging to a crack defect and the degree of each defect connected domain being a crack defect, the crack degree of each defect connected domain is obtained, and as an embodiment, the specific calculation method is as follows: wherein, the possibility of the i-th defect connected domain belonging to a crack defect, the degree of the i-th defect connected domain being a crack defect, the crack degree of the i-th defect connected domain. A first threshold value is preset , and as an example, the embodiment is described, and the embodiment is not specifically limited, wherein

[0041] , and as an example, the embodiment is described, and the embodiment is not specifically limited, wherein ​​​​​​The first threshold value can be set according to specific implementation. All the defect connected domains with the crack degree greater than the preset first threshold value are recorded as reference connected domains.

[0042] At this point, all the reference connected domains are obtained.

[0043] Step S003: obtaining a gradient change vector of each reference connected domain, obtaining a decreasing vector of each edge pixel point in each reference connected domain; according to the difference between the gradient change vector of each reference connected domain and the decreasing vector of all edge pixel points in each reference connected domain, a target connected domain is screened out from all the reference connected domains.

[0044] It should be noted that the tooth surface structure causes the tooth surface to form some connected domains with small gray difference and flat edges similar to cracks, so it is necessary to screen from all the reference connected domains to obtain the accurate crack connected domain.

[0045] Further, it should be noted that the gradient change of the edge pixel points of the crack region presents a decreasing regularity, so the gray change of the edge pixel points of the reference connected domain is further analyzed.

[0046] Specifically, a minimum circumscribed rectangle of each reference connected domain is obtained; a line segment parallel to the short side and equal in length to the short side is drawn through the center point of the minimum circumscribed rectangle, which is recorded as the first line segment of the minimum circumscribed rectangle of each reference connected domain; a line segment parallel to the long side and equal in length to the long side is drawn through the center point of the minimum circumscribed rectangle, which is recorded as the second line segment of the minimum circumscribed rectangle of each reference connected domain; the minimum circumscribed rectangle is divided into two target regions according to the first line segment of the minimum circumscribed rectangle of each reference connected domain, the mean value of the gradient amplitudes of all pixel points in each target region is calculated, the target region with the maximum mean value of the gradient amplitudes is recorded as the first target region, and the target region with the minimum mean value of the gradient amplitudes is recorded as the second target region.

[0047] According to the second line segment of the minimum circumscribed rectangle of each reference connected domain, the first target region and the second target region, a gradient change vector of each reference connected domain is obtained, wherein the length of the gradient change vector of each reference connected domain is equal to the length of the second line segment of the minimum circumscribed rectangle of each reference connected domain, and the direction of the gradient change vector of each reference connected domain is from the first target region to the second target region.

[0048] Obtain the edge pixels at the edges of each reference connected component. Find the pixel with the smallest gradient magnitude in its eight-neighborhood that is less than the edge pixel itself, and denote it as the target pixel of that edge pixel. Use the direction from each edge pixel to its corresponding target pixel as the direction of the decreasing vector for that edge pixel. If the target pixel of each edge pixel belongs to one of its four neighboring pixels, the magnitude of the decreasing vector is 1 unit. If the target pixel belongs to one of its D neighboring pixels, the magnitude of the decreasing vector is... Thus, the decreasing vector of each edge pixel is obtained. When there is no matching target pixel within the eight neighbors of an edge pixel, the decreasing vector of that edge pixel is replaced with a vector of 0.

[0049] Based on the difference between the gradient change vector of each reference connected region and the decreasing vector of all edge pixels within each reference connected region, the probability that each reference connected region is a cracked connected region is obtained. As an example, the specific calculation method is as follows: In the formula, Indicates the first The first reference connected component A decreasing vector of edge pixels, Indicates the first Gradient change vectors of each reference connected component Indicates the first The first reference connected component The magnitude of the decreasing vector of each edge pixel. Indicates the first The first reference connected component The decreasing vector of the edge pixel and the first Cosine similarity between gradient transformation vectors of reference connected components Indicates the first The number of all edge pixels in a reference connected region Represents a linear normalization function. Indicates the first The probability that a reference connected region is a cracked connected region.

[0050] in, Indicates the first The first reference connected component The projection of the magnitude of the decreasing vector of the edge pixels on the th edge pixel is... The length of the gradient change vector in the direction corresponding to each reference connected region is such that the longer the length, the more the gradient change of the edge pixels of the reference connected region conforms to the decreasing trend, and the greater the probability that the reference connected region is a cracked connected region; the shorter the length, the less the gradient change of the edge pixels of the reference connected region conforms to the decreasing trend, and the smaller the probability that the reference connected region is a cracked connected region.

[0051] Thus, we obtain the probability that each reference connected region is a cracked connected region.

[0052] Preset a second threshold In this embodiment, This example is used for illustration; no specific limitations are set in this embodiment. It depends on the specific implementation situation.

[0053] The probability that the reference connected component is a fractured connected component is greater than a preset second threshold. All reference connected components are denoted as the target connected component.

[0054] At this point, all target connected components have been obtained.

[0055] Step S004: Obtain the projection vector of each reference connected component based on the decreasing vector of all edge pixels in each reference connected component; merge all target connected components based on the difference between the projection vectors of any two target connected components, the difference between the gradient change vectors of any two target connected components, and the distance between the center points of any two target connected components to obtain the tooth consistency recognition result of the bulldozer tooth block.

[0056] It should be noted that, due to the possibility of noise interference, the cracked connected domains belonging to the same connected domain may be divided into two. Therefore, it is necessary to merge them based on the differences between the two target connected domains.

[0057] Specifically, Recorded as the number The first reference connected component The decreasing projection vector of the i-th edge pixel, for the i-th edge pixel Perform vector operations on the decreasing projection vectors of all edge pixels in the nth reference connected region, and denote the result of the vector operation as the nth... The projection vectors of the reference connected components. Where, in the formula, Indicates the first The first reference connected component A decreasing vector of edge pixels, Indicates the first Gradient change vectors of each reference connected component Indicates the first The first reference connected component The magnitude of the decreasing vector of each edge pixel. Indicates the first The magnitude of the gradient change vector of each reference connected region. Indicates the first The first reference connected component The decreasing vector of the edge pixel and the first Cosine similarity between gradient transformation vectors of a reference connected region.

[0058] Thus, the projection vector of each reference connected component is obtained; and the projection vector of each target connected component is also obtained.

[0059] Based on the difference between the projection vectors of any two target connected components, the difference between the gradient change vectors of any two target connected components, and the distance between the center points of any two target connected components, the merging probability between any two target connected components is obtained. As an example, the specific calculation method is as follows: In the formula, Indicates the first The target connected component and the first The distance between the center points of each target connected component Indicates the first The target connected component and the first The angle between the projection vectors of the target connected components. Indicates the first The target connected component and the first The angle between the gradient change vectors of the target connected components. This represents an exponential function with the natural constant as its base. Indicates the first The target connected component and the first The possibility of merging between target connected components.

[0060] Specifically, the smaller the distance between the center points of two target connected components, the greater the likelihood of them merging; conversely, the greater the distance, the less likely they are to merge. Similarly, the smaller the angle between the projection vectors of two target connected components, the greater the likelihood of them merging; and the greater the angle, the less likely they are to merge. Finally, the smaller the angle between the gradient change vectors of two target connected components, the greater the likelihood of them merging; and the greater the angle, the less likely they are to merge.

[0061] Preset a third threshold In this embodiment, This example is used for illustration; no specific limitations are set in this embodiment. It depends on the specific implementation situation.

[0062] The probability of merging is greater than the preset third threshold. The two target connected components are merged, and all the target connected components after the merger are denoted as the crack connected components.

[0063] At this point, all the connected domains with cracks have been obtained.

[0064] Preset a fourth threshold In this embodiment, This example is used for illustration; no specific limitations are set in this embodiment. This can be determined based on the specific implementation. The ratio between the area of ​​all connected regions of cracks and the area of ​​the bulldozer tooth image is recorded as the crack area ratio; when the crack area ratio is less than a preset fourth threshold... If the identified bulldozer teeth are consistent, then the crack area ratio is greater than or equal to the preset fourth threshold. If so, the identified bulldozer teeth are inconsistent.

[0065] At this point, the tooth consistency identification of the bulldozer tooth blocks is complete. The flowchart for the bulldozer tooth consistency identification is as follows: Figure 2 As shown.

[0066] It should be noted that the embodiments used in this example The model is only used to represent negative correlations and the results of the constraint model output are in Within this range, in specific implementations, other models with the same purpose can be substituted; this embodiment is merely an example. The description will be based on a model, without making specific limitations on it. This refers to the input of the model.

[0067] This concludes the embodiment.

[0068] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for intelligent recognition of tooth consistency in bulldozer tooth blocks, characterized in that, The method includes the following steps: Obtain images of the teeth of the bulldozer's tooth block; Obtain several defect connected components in the tooth image, and obtain the edge chain code sequence for each defect connected component; based on the grayscale distribution of pixels within a preset local window of all pixels in each defect connected component and the differences between different data in each edge chain code sequence, select reference connected components from all defect connected components; Obtain the gradient change vector of each reference connected component, and obtain the decreasing vector of each edge pixel in each reference connected component; based on the difference between the gradient change vector of each reference connected component and the decreasing vector of all edge pixels in each reference connected component, select the target connected component from all reference connected components. The projection vector of each reference connected component is obtained by using the decreasing vector of all edge pixels in each reference connected component. Based on the difference between the projection vectors of any two target connected components, the difference between the gradient change vectors of any two target connected components, and the distance between the center points of any two target connected components, all target connected components are merged to obtain the tooth consistency recognition result of the bulldozer tooth block.

2. The intelligent recognition method for tooth consistency of bulldozer tooth blocks according to claim 1, characterized in that, The specific steps for obtaining several defect connected regions in the tooth image and obtaining the edge chain code sequence of each defect connected region are as follows: The edge detection of the bulldozer tooth block image is performed using the Canny edge detection algorithm. Then, the edge detection results of the bulldozer tooth block image are processed by binarization dilation to obtain several defect connected components. The edge chain code sequence of each defective connected component is obtained using the Freeman chain code algorithm.

3. The intelligent recognition method for tooth consistency of bulldozer tooth blocks according to claim 1, characterized in that, The specific steps for selecting reference connected components from all defective connected components based on the grayscale distribution of pixels within a preset local window of all pixels in each defective connected component and the differences between different data in each edge chain code sequence are as follows: Based on the grayscale distribution of pixels within the local window of all pixels in each defect connected region, the probability that each defect connected region belongs to a crack defect can be obtained. The degree to which each defective connected region is a crack defect is obtained by considering the differences between adjacent data in the edge chain code sequence of each defective connected region and the differences between each data in the edge chain code sequence and the mean of all data in the edge chain code sequence. Based on the probability that each defect connected region belongs to a crack defect and the degree to which each defect connected region is a crack defect, the crack degree of each defect connected region is obtained. Among them, the probability that each defect connected region belongs to a crack defect is positively correlated with the crack degree of each defect connected region, and the degree to which each defect connected region is a crack defect is positively correlated with the crack degree of each defect connected region. If the degree of crack is greater than a preset first threshold All defective connected components are denoted as reference connected components.

4. The intelligent recognition method for tooth consistency of bulldozer tooth blocks according to claim 3, characterized in that, The method for determining the probability that each connected region of a defect belongs to a crack defect based on the grayscale distribution of pixels within a local window of all pixels in each defect connected region includes the following specific calculation method: In the formula, Indicates the first The first defective connected component The pixels within the local window of the nth pixel belong to the nth pixel. The average grayscale value of all pixels in a defective connected region. Indicates the first The first defective connected component Pixels within a local window of pixel number 1 do not belong to the 1st pixel. The average grayscale value of all pixels in a defective connected region. Indicates the first The number of all pixels in a defective connected domain. It is the absolute value symbol. This represents an exponential function with the natural constant as its base. Indicates the first The probability that a defective connected domain belongs to a crack defect.

5. The intelligent recognition method for tooth consistency of bulldozer tooth blocks according to claim 3, characterized in that, The degree to which each defective connected region is a crack defect is obtained based on the differences between adjacent data in the edge chain code sequence of each defective connected region, and the differences between each data in the edge chain code sequence of each defective connected region and the mean of all data in the edge chain code sequence. The specific calculation method is as follows: In the formula, Indicates the first The first edge chain code sequence of the defective connected component One data point, Indicates the first The first edge chain code sequence of the defective connected component One data point, Indicates the first The mean of all data in the edge chain code sequence of a defective connected component. Indicates the first The number of all data in the edge chain code sequence of a defective connected component It is the absolute value symbol. Represents a linear normalization function. Indicates the first Each defect connected region represents the degree of the crack defect.

6. The intelligent recognition method for tooth consistency of bulldozer tooth blocks according to claim 1, characterized in that, The specific steps for obtaining the gradient change vector of each reference connected component are as follows: Obtain the minimum bounding rectangle of each reference connected component; draw a line segment through the center point of the minimum bounding rectangle that is parallel to the short side and has the same length as the short side, and denote it as the first line segment of the minimum bounding rectangle of each reference connected component; draw a line segment through the center point of the minimum bounding rectangle that is parallel to the long side and has the same length as the long side, and denote it as the second line segment of the minimum bounding rectangle of each reference connected component. Divide the minimum bounding rectangle of each reference connected component into two target regions based on the first line segment of the minimum bounding rectangle. Calculate the mean gradient magnitude of all pixels in each target region. The target region with the largest mean gradient magnitude is designated as the first target region, and the target region with the smallest mean gradient magnitude is designated as the second target region. Based on the second line segment of the minimum bounding rectangle of each reference connected region, the first target region, and the second target region, the gradient change vector of each reference connected region is obtained. The magnitude of the gradient change vector of each reference connected region is equal to the length of the second line segment of the minimum bounding rectangle of each reference connected region, and the direction of the gradient change vector of each reference connected region points from the first target region to the second target region.

7. The intelligent recognition method for tooth consistency of bulldozer tooth blocks according to claim 1, characterized in that, The specific steps for obtaining the decreasing vector of each edge pixel in each reference connected component are as follows: Obtain the edge pixels in each reference connected component. Find the pixel with the smallest gradient magnitude in its eight-neighborhood that is less than the edge pixel itself, and denote it as the target pixel of that edge pixel. Use the direction from each edge pixel to its corresponding target pixel as the direction of its decreasing vector. If the target pixel of each edge pixel belongs to one of its four neighboring pixels, the magnitude of the decreasing vector is 1 unit. If the target pixel belongs to one of its D neighboring pixels, the magnitude of the decreasing vector is... ; If there is no matching target pixel in the eight neighborhoods of each edge pixel, then the decreasing vector of each edge pixel is replaced with a 0 vector.

8. The intelligent recognition method for tooth consistency of bulldozer tooth blocks according to claim 1, characterized in that, The specific calculation method for selecting the target connected component from all reference connected components based on the difference between the gradient change vector of each reference connected component and the decreasing vector of all edge pixels within each reference connected component is as follows: In the formula, Indicates the first The first reference connected component A decreasing vector of edge pixels, Indicates the first Gradient change vectors of each reference connected component Indicates the first The first reference connected component The magnitude of the decreasing vector of each edge pixel. Indicates the first The first reference connected component The decreasing vector of the nth edge pixel and the nth edge pixel Cosine similarity between gradient transformation vectors of reference connected components Indicates the first The number of all edge pixels in a reference connected region Represents a linear normalization function. Indicates the first The probability that a reference connected region is a cracked connected region; The probability that the reference connected component is a fractured connected component is greater than a preset second threshold. All reference connected components are denoted as the target connected component.

9. The intelligent recognition method for tooth consistency of bulldozer tooth blocks according to claim 8, characterized in that, The specific steps for obtaining the projection vector of each reference connected region based on the decreasing vector of all edge pixels in each reference connected region are as follows: Will Recorded as the number The first reference connected component The decreasing projection vector of the i-th edge pixel, for the i-th edge pixel Perform vector operations on the decreasing projection vectors of all edge pixels in the nth reference connected region, and denote the result of the vector operation as the nth... Projection vectors of each reference connected component; In the formula, Indicates the first The magnitude of the gradient change vector of a reference connected region.

10. The intelligent recognition method for tooth consistency of bulldozer tooth blocks according to claim 1, characterized in that, The process of merging all target connected components based on the differences between the projection vectors of any two target connected components, the differences between the gradient change vectors of any two target connected components, and the distance between the center points of any two target connected components to obtain the tooth consistency recognition result of the bulldozer tooth block includes the following specific steps: The probability of merging any two target connected components is obtained based on the difference between the projection vectors of any two target connected components, the difference between the gradient change vectors of any two target connected components, and the distance between the center points of any two target connected components. The specific calculation method is as follows: In the formula, Indicates the first The target connected component and the first The distance between the center points of each target connected component Indicates the first The target connected component and the first The angle between the projection vectors of the target connected components. Indicates the first The target connected component and the first The angle between the gradient change vectors of the target connected components. This represents an exponential function with the natural constant as its base. Indicates the first The target connected component and the first The possibility of merging between target connected components; The probability of merging is greater than the preset third threshold. Merge the two target connected components and denote all the target connected components after the merger as the crack connected components; The ratio between the area of ​​all connected regions of cracks and the area of ​​the bulldozer tooth image is denoted as the crack area ratio; when the crack area ratio is less than a preset fourth threshold... If the identified bulldozer teeth are consistent, then the crack area ratio is greater than or equal to the preset fourth threshold. If so, the identified bulldozer teeth are inconsistent.

Images