Electronic control silicone oil clutch image analysis system based on big data

Abstract

The invention relates to the technical field of image processing, in particular to an electronic control silicone oil clutch image analysis system based on big data, which comprises an image processing module used for acquiring a grayscale image of a flange surface in an electronic control silicone oil clutch; the template acquisition module is used for acquiring a scratch-free template region according to the texture distribution characteristics of the target region subjected to partition processing; the region screening module is used for integrally matching the non-scratch template region with each target region to obtain all suspected scratch regions, and then carrying out local sub-region matching on each suspected scratch region with the non-scratch template region to obtain a scratch sub-region in each suspected scratch region; and the model construction module is used for marking a gray level image according to the scratch sub-region, training a quality detection model by using the marked image, carrying out real-time quality detection on the flange surface by using the trained quality detection model, improving the accuracy of scratch identification through matching of local sub-regions, and further improving the accuracy of flange surface quality detection.

Description

Technical Field

[0001] The present invention relates to the technical field of image processing, and in particular to an electronically controlled silicone oil clutch image analysis system based on big data. Background Art

[0002] Electronically controlled silicone oil clutches are mostly used to adjust the cooling intensity of the engine to adapt to different environmental conditions and operating conditions. The quality of the flange surface affects the performance of the clutch. If the flange surface quality does not meet the standard, such as scratches on the surface, it will lead to insufficient clutch sealing, silicone oil leakage, and increased costs. Therefore, it is necessary to detect scratches on the flange surface.

[0003] In the prior art, the surface scratches of the flange surface are identified and detected by template matching technology. Specifically, the surface image of the flange surface collected in real time is matched with the surface image of the scratch template, and the final scratch area is obtained according to the matching result. However, since the grayscale of the texture and scratches in the flange surface is close, some scratches and textures may overlap, which makes it difficult to directly determine the characteristics of the scratches, thereby reducing the accuracy of model matching. In addition, due to the metal characteristics of the flange surface, some areas will have light reflection, resulting in anomalies in the template matching, so that the illuminated area is recorded as a scratch area, affecting the quality judgment of the flange surface.

[0004] Therefore, how to improve the accuracy of scratch recognition on the flange surface using template matching has become an urgent problem to be solved. Summary of the invention

[0005] In view of this, an embodiment of the present invention provides an electronically controlled silicone oil clutch image analysis system based on big data to solve the problem of how to improve the accuracy of scratch recognition on the flange surface using template matching.

[0006] An embodiment of the present invention provides an electronically controlled silicone oil clutch image analysis system based on big data, which includes the following modules:

[0007] An image processing module is used to obtain a surface image of a flange surface in an electronically controlled silicone oil clutch, and grayscale the surface image to obtain a corresponding grayscale image;

[0008] A template acquisition module, used for performing partition processing on the grayscale image according to the hole positions on the flange surface to obtain at least one target area containing a single hole, and for any target area, obtaining the gradient values ​​of all pixels in the target area, and obtaining the texture distribution feature value of the target area according to the gradient value difference of each pixel in the target area, and obtaining a scratch-free template area according to the texture distribution feature value of each target area;

[0009] A region screening module, used to respectively obtain the overall matching degree between the scratch-free template region and each of the target regions, screen all target regions to obtain suspected scratch regions according to the overall matching degree, and for any suspected scratch region, perform local sub-region matching between the suspected scratch region and the scratch-free template region to obtain a scratch sub-region in the suspected scratch region, and obtain a scratch sub-region in each of the suspected scratch regions;

[0010] The model building module is used to mark the grayscale image according to all scratch sub-areas to obtain a marked image, use the marked image to train the quality detection model to obtain a trained quality detection model, and use the trained quality detection model to perform real-time quality detection on the flange surface in the electronically controlled silicone oil clutch.

[0011] Compared with the prior art, the embodiments of the present invention have the following beneficial effects:

[0012] The present invention provides an electronically controlled silicone oil clutch image analysis system based on big data, comprising an image processing module, which is used to obtain a surface image of a flange surface in an electronically controlled silicone oil clutch, and grayscale the surface image to obtain a corresponding grayscale image; a template acquisition module, which is used to partition the grayscale image according to the hole position on the flange surface, to obtain at least one target area containing a single hole, and for any target area, to obtain the gradient values ​​of all pixel points in the target area, and to obtain the texture distribution feature value of the target area according to the gradient value difference of each pixel point in the target area, and to obtain the scratch-free template area according to the texture distribution feature value of each target area; and an area screening module, which is used to partition the grayscale image according to the hole position on the flange surface, to obtain at least one target area containing a single hole, and for any target area, to obtain the gradient values ​​of all pixel points in the target area, and to obtain the texture distribution feature value of the target area according to the gradient value difference of each pixel point in the target area, and to obtain the scratch-free template area according to the texture distribution feature value of each target area; and The overall matching degree between the scratch-free template area and each of the target areas is obtained separately, and suspected scratch areas are screened in all target areas according to the overall matching degree. For any suspected scratch area, the suspected scratch area is matched with the scratch-free template area in a local sub-area to obtain the scratch sub-area in the suspected scratch area, and the scratch sub-area in each of the suspected scratch areas is obtained; a model building module is used to mark the grayscale image according to all scratch sub-areas to obtain a marked image, use the marked image to train the quality detection model to obtain the trained quality detection model, and use the trained quality detection model to perform real-time quality detection on the flange surface in the electronically controlled silicone oil clutch. Among them, since the texture and scratch grayscale of the flange surface in the electronically controlled silicone oil clutch are close, some scratches and textures may overlap, making it difficult to directly determine the characteristics of the scratches, and the texture distribution rules of the flange surface are clear and the design edges are clear. Therefore, the obtained flange surface grayscale image is partitioned to obtain at least one target area, and a scratch-free template area is obtained according to the texture distribution characteristics of each target area; due to the metal characteristics of the flange surface, some areas will have light reflection, resulting in a decrease in the matching result, thereby marking the area as a scratch area, affecting the quality judgment of the flange surface, therefore, each target area and the scratch-free template area are first matched as a whole to obtain a suspected scratch area, and then each suspected scratch area and the scratch-free template area are matched locally in a sub-area to obtain the final scratch sub-area, so as to eliminate the matching errors caused by lighting, thereby improving the accuracy of scratch recognition on the flange surface using template matching, and then improving the accuracy of flange surface quality detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0014] Figure 1 It is a structural block diagram of an electronically controlled silicone oil clutch image analysis system based on big data provided by the first embodiment of the present invention;

[0015] Figure 2 It is a schematic diagram of partitioning a grayscale image of a flange surface in an electronically controlled silicone oil clutch provided in the first embodiment of the present invention;

[0016] Figure 3 It is a schematic diagram of obtaining lost reference points of a local sub-region belonging to a suspected scratch region in a target local sub-region, provided in the first embodiment of the present invention. DETAILED DESCRIPTION

[0017] Embodiments of the present disclosure are described in detail below, and examples of the embodiments are shown in the accompanying drawings. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present disclosure, but should not be understood as limiting the present disclosure.

[0018] It should be noted that the terms "first", "second", etc. in the specification of the present disclosure and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present disclosure described herein can be implemented in an order other than those illustrated or described herein. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present disclosure. Instead, they are merely examples of devices and methods consistent with some aspects of the present disclosure.

[0019] In order to illustrate the technical solution of the present invention, specific embodiments are provided below for illustration.

[0020] See also Figure 1 , is a structural block diagram of an electronically controlled silicone oil clutch image analysis system based on big data provided by the first embodiment of the present invention, such as Figure 1 As shown, the system may include the following modules:

[0021] The image processing module 11 is used to obtain a surface image of the flange surface in the electronically controlled silicone oil clutch, and grayscale the surface image to obtain a corresponding grayscale image.

[0022] Since the quality of the flange surface affects the performance of the electronically controlled silicone oil clutch, if the flange surface quality is not up to standard, such as scratches on the surface, it will lead to insufficient clutch sealing, silicone oil leakage, and increased costs. Therefore, in an embodiment of the present invention, a surface image of the flange surface in the electronically controlled silicone oil clutch is first obtained, and then the obtained surface image of the flange surface in the electronically controlled silicone oil clutch is grayed to obtain a corresponding grayscale image, which is convenient for subsequent detection and identification of scratches on the flange surface.

[0023] It should be noted that since the technology of grayscale processing of the surface image of the flange surface is a well-known technology, it will not be elaborated here.

[0024] At this point, a grayscale image of the flange surface in the electronically controlled silicone oil clutch is obtained.

[0025] The template acquisition module 12 is used to partition the grayscale image according to the hole positions on the flange surface to obtain at least one target area containing a single hole. For any target area, the gradient values ​​of all pixels in the target area are obtained, and the texture distribution characteristic value of the target area is obtained according to the difference in gradient values ​​of each pixel in the target area. According to the texture distribution characteristic value of each target area, a scratch-free template area is obtained.

[0026] After obtaining the grayscale image, considering that the flange surface is a standard circle and the positions of the holes for mounting the parts are evenly distributed during design, it is necessary to set a partitioning rule to make it easier to match so that the texture is consistent in the direction of each partition, has matching properties, and is not interfered by factors such as the designed holes. Therefore, in the embodiment of the present invention, the grayscale image is partitioned according to the positions of the holes on the surface of the flange surface to obtain at least one target area containing a single hole, so as to improve the efficiency and accuracy of subsequent template matching and ensure more accurate positioning of the target object. The specific partitioning processing method is as follows:

[0027] The inner edge line of the flange surface and the outer edge line of the flange surface in the grayscale image are obtained to obtain a circular area formed by the inner edge line of the flange surface and the outer edge line of the flange surface, the number of holes in the grayscale image is counted, and the circular area is evenly divided into N target areas equal to the number of the holes, and one of the target areas contains one hole.

[0028] In one implementation, the inner edge line of the flange surface and the outer edge line of the flange surface in the grayscale image are obtained according to the edge detection technology, the center point of the flange surface and the center point of the hole in the flange surface are obtained, and for any three adjacent holes, the center points of the holes and the center points of the flange surface are connected respectively, and the inner edge arc and the outer edge arc intersecting with the inner edge line and the outer edge line of the flange surface are correspondingly obtained, and the obtained inner edge arc and the outer edge arc are evenly divided into four parts, and the middle part arcs of the inner edge arc and the outer edge arc are respectively divided, that is, the arcs To The bottom and the top of the target area are obtained by locating the position of the target area, and the bottom and the top of the target area are connected by a straight line. The connected area obtained is recorded as the target area, so that the target area contains a hole and the hole position is the middle position of the target area. Repeat the above steps to obtain the target area to obtain N target areas.

[0029] For example: Refer to Figure 2 , which is a schematic diagram of partitioning the grayscale image of the flange surface in the electronically controlled silicone oil clutch, Figure 2 The large circle in the figure is the outer edge of the flange surface, the small circle is the inner edge of the flange surface, and the area between the outer edge of the flange surface and the inner edge of the flange surface is the annular area, that is, all the holes on the flange surface ( Figure 2 The area where the small ball in the figure is located is located, and the corresponding bold area is taken as a target area, and each target area has the inner edge line of the flange surface as the bottom and the outer edge line of the flange surface as the top, thereby obtaining N target areas with the same number of holes as the flange surface, and each target area has a hole, and the middle position of the target area is the hole position.

[0030] Since the texture distribution rules of the flange surface are clear, the design edges are clear, and scratches exist only in a few areas, it is more accurate to determine the template according to the uniformity of the texture distribution of the flange surface, and the grayscale values ​​of the texture-free positions in the flange surface are similar, and the gradient values ​​of the texture edge points are significantly different from the gradient values ​​of the surrounding points. The texture edge points have the characteristic of large gradient values, while the gradient values ​​of the remaining non-texture edge points in the flange surface are relatively small. Therefore, in an embodiment of the present invention, for any target area, the gradient values ​​of all pixel points in the target area are first obtained, and the texture distribution characteristic value of the target area is obtained according to the difference in the gradient values ​​of each pixel point in the target area. Then, according to the texture distribution characteristic value of each target area, a scratch-free template area is obtained.

[0031] The texture distribution feature value of the target area is obtained according to the gradient value difference of each pixel point in the target area, including:

[0032] If the gradient value of any pixel point in the target area is greater than a preset gradient threshold, the pixel point is marked as a reference point to obtain all reference points in the target area;

[0033] All reference points are clustered to obtain at least one cluster, the total number of clusters and the number of reference points in each cluster are counted, the vertical distance between the cluster center point of each cluster and the inner edge line of the flange surface in the target area is calculated respectively, and all clusters are arranged in descending order of the vertical distance to obtain a cluster sequence.

[0034] For any reference point in the target area, all reference points contained in the neighborhood of the reference point and the reference point form a reference point set of the reference point, and according to the reference point set of each reference point in the target area, the total number of reference point sets and the number of reference points in each reference point set are counted, and all reference point sets are sorted according to the traversal order of the reference points in the target area to obtain a reference point set sequence;

[0035] For an r-th reference point set in the reference point set sequence, calculate an absolute value of a difference between the number of reference points in the r-th reference point set and the number of reference points in the r-th cluster in the cluster sequence, calculate an average of all absolute values ​​of the differences according to the total number of reference point sets, and obtain a first addition result between a constant 1 and the average;

[0036] Calculating a first absolute value of a difference between the total number of the reference point set and the total number of clusters, and obtaining a second addition result between a constant 1 and the first absolute value of the difference;

[0037] The product of the first addition result and the second addition result is used as the texture distribution feature value of the target area.

[0038] In one implementation, based on the obtained N target areas, taking the nth target area obtained by division as an example, the gradient values ​​of all pixels in the nth target area are obtained using the sobel operator, and a preset gradient threshold is obtained. The gradient threshold of this solution is 40. The preset gradient threshold is not limited in the embodiment of the present invention and can be set according to the implementation scenario. If the gradient value of any pixel is greater than the gradient threshold, the pixel is marked as a reference point, and all reference points in the nth target area are obtained according to the above method. It is worth noting that the sobel operator belongs to the prior art and will not be repeated here.

[0039] Density clustering is performed on all reference points. Areas with sufficiently high density are clustered into a cluster according to the density of the reference points. Density clustering belongs to the prior art and will not be described here. The total number of clusters M and the number of reference points in each cluster are statistically obtained by clustering. The vertical distance between the cluster center point of each cluster and the inner edge line of the flange surface in the nth target area is calculated respectively. All clusters are arranged in descending order according to the vertical distance to obtain a cluster sequence. Then, a four-neighborhood traversal is performed on each reference point. Any reference point is selected as the initial reference point. The neighboring areas in the four directions of up, down, left, and right of the initial reference point are traversed. The reference point is marked as the second reference point, and then each second reference point is used to continue traversing to obtain adjacent reference points in the four directions of the second reference point, namely, above, below, left, and right. The reference points are traversed according to the above method until there are no reference points in the four neighborhoods of the reference point. The traversal is terminated, and the initial reference point and all the reference points in the four neighborhoods obtained by traversal are stored in the same reference point set. Similarly, the reference point set corresponding to each reference point in the nth target area is obtained, and the total number of reference point sets H and the number of reference points in each reference point set are obtained by counting. The reference point sets are sorted according to the order of reference point traversal to obtain a reference point set sequence.

[0040] For the rth reference point set in the reference point set sequence, the absolute value of the difference between the number of reference points in the rth reference point set and the number of reference points in the rth cluster in the cluster sequence is calculated. According to the total number H of the reference point set, the average of the H absolute values ​​of the difference is calculated. The absolute value of the number difference between the total number of the reference point set and the total number of clusters is calculated. According to the absolute value and the average of the number difference, the texture distribution feature value of the nth target area is obtained. Then, the expression of the texture distribution feature value is:

[0041]

[0042] Among them, R n represents the texture distribution feature value of the nth target area, H n represents the total number of reference points in the nth target area, M n represents the total number of clusters in the nth target area, G r represents the number of reference points in the rth reference point set in the reference point set sequence, C r It represents the number of reference points in the rth cluster in the cluster sequence, 1 represents a constant, and || represents the absolute value sign.

[0043] It should be noted that |H n -M n | reflects the difference between the total number of reference points and the total number of clusters, |H n -M n The smaller the value of |, the closer the total number of clusters is to the total number of reference points, the fewer discrete points are, and the closer the corresponding texture distribution is; It reflects the distribution between clusters and reference point sets. The smaller the value of is, the closer the distribution of the clusters is to the distribution of the reference point set, and the closer the corresponding texture distribution is, that is, the more the distribution conforms to the regional characteristics of the flange surface with only texture.

[0044] Similarly, referring to the above method for obtaining texture distribution feature values, the texture distribution feature values ​​of each target area are obtained respectively, and then the scratch-free template area is obtained according to the texture distribution feature values ​​of each target area for subsequent template matching. The specific method for obtaining the scratch-free template area is:

[0045] According to the texture distribution characteristic value of each target area, the target area corresponding to the minimum value of the texture distribution characteristic value is used as the scratch-free template area.

[0046] At this point, the scratch-free template area in the divided target area is obtained according to the texture distribution characteristics of the flange surface.

[0047] The area screening module 13 is used to obtain the overall matching degree between the scratch-free template area and each target area respectively, and screen the suspected scratch area in all target areas according to the overall matching degree. For any suspected scratch area, the suspected scratch area is matched with the scratch-free template area in a local sub-area to obtain the scratch sub-area in the suspected scratch area, and the scratch sub-area in each suspected scratch area is obtained.

[0048] After obtaining the scratch-free template area, each target area is matched with the scratch-free template area respectively. Since the flange surface is a uniform and symmetrical object, and the design anomalies of the flange surface are also taken into account when partitioning, if there is no abnormal interference such as scratches, the distribution of each target area should be similar to the distribution of the scratch-free template area, and the corresponding template matching degree will be higher. On the contrary, the target area with a lower matching degree is likely to be a scratch area. Since local processing can better capture the features in the target area and can more accurately perform template matching on the target area, in the embodiment of the present invention, the overall matching degree between the scratch-free template area and each target area is first obtained respectively, and then the suspected scratch area is screened in all target areas according to the overall matching degree. The specific method for obtaining the suspected scratch area is:

[0049] (1) For any target area, the target area and the scratch-free template area are respectively divided into at least one sub-area of ​​a preset size, and the sub-areas of the target area and the scratch-free template area are matched at the same position to obtain a matching sub-area pair.

[0050] In one embodiment, based on the N target areas obtained, taking the nth target area obtained by division as an example, the nth target area and the scratch-free template area are respectively divided into at least one sub-area of ​​size 3*3, and the number of sub-areas in the nth target area is equal to the number of sub-areas in the scratch-free template area, and the sub-areas corresponding to the same position in the nth target area and the scratch-free template area are matched to obtain K matched sub-area pairs, wherein one sub-area pair contains two sub-areas at the same position in the nth target area and the scratch-free template area.

[0051] It should be noted that the embodiment of the present invention does not impose any requirements on the size of the divided sub-areas, which can be set according to the implementation scenario.

[0052] (2) Obtaining the local matching degree of each of the sub-region pairs according to the difference in the region grayscale value and the number of reference points in the sub-region pairs.

[0053] Specifically, for any pair of sub-regions, respectively obtain the number ratio between the number of pixel points in each sub-region of the pair of sub-regions and the number of reference points, obtain the subtraction absolute value of the number ratio, and obtain the subtraction result between a constant 1 and the subtraction absolute value;

[0054] Using a structural similarity index to obtain the local similarity of the sub-region pair;

[0055] A weighted sum is performed on the local similarity and the subtraction result of the sub-region pair to obtain a corresponding weighted sum result as the local matching degree of the sub-region pair.

[0056] In one embodiment, based on the obtained K sub-region pairs, taking the matched k-th sub-region pair as an example, the number of reference points and the number of pixels in each sub-region of the k-th sub-region pair are obtained, a first ratio between the number of reference points t and the number of pixels T in the sub-region belonging to the target region in the k-th sub-region pair is calculated, a second ratio between the number of reference points t and the number of pixels T in the sub-region belonging to the scratch-free template region in the k-th sub-region pair is calculated, a ratio difference between the first ratio and the second ratio is obtained, an SSIM value in the k-th sub-region pair is calculated, and a local matching degree of the k-th sub-region pair is obtained according to the SSIM value and the ratio difference. Then, the expression of the local matching degree of the k-th sub-region pair is:

[0057]

[0058] Among them, T k represents the local matching degree of the kth sub-region pair, SSIM k represents the local similarity of the kth sub-region pair, F kt represents the number of reference points in the sub-region belonging to the scratch-free template region in the k-th sub-region pair, F kT represents the number of pixels in the sub-region belonging to the scratch-free template region in the k-th sub-region pair, Q kt represents the number of reference points in the sub-region belonging to the target region in the k-th sub-region pair, Q kT It represents the number of pixels in the sub-region belonging to the target region in the k-th sub-region pair, 1 represents a constant, and || represents the absolute value sign.

[0059] It should be noted that SSIM k It reflects the similarity between the two sub-regions in the k-th sub-region pair. k The larger the value of is, the more similar the two sub-regions in the k-th sub-region pair are, and the higher the local matching degree of the corresponding k-th sub-region pair is; reflects the distribution of reference points between the two sub-regions in the k-th sub-region pair, The smaller the value of is, the closer the reference point distribution between the two sub-regions in the k-th sub-region pair is, and the higher the local matching degree of the corresponding sub-region pair is.

[0060] (3) Calculating the average of all local matching degrees as the overall matching degree between the target area and the scratch-free template area; based on the overall matching degree between each target area and the scratch-free template area, if any overall matching degree is less than a preset matching degree threshold, marking the target area corresponding to the overall matching degree as a suspected scratch area, thereby obtaining all suspected scratch areas.

[0061] In one embodiment, according to the method for obtaining the local matching degree of the kth sub-region pair, the local matching degree of each sub-region pair is obtained respectively, and then according to the obtained local matching degrees of the K sub-region pairs, the average of the local matching degrees of the K sub-region pairs is calculated to obtain the overall matching degree between the nth target region and the scratch-free template region, and the expression of the overall matching degree is:

[0062]

[0063] Among them, S n represents the overall matching degree between the nth target region and the scratch-free template region, K represents the number of sub-region pairs, and T k represents the local matching degree of the kth sub-region pair.

[0064] Repeat the above method of obtaining local matching and overall matching to obtain the overall matching between each target area and the scratch-free template area. In the embodiment of the present invention, the matching threshold is set to 0.9. There is no restriction on the preset matching threshold, which can be set according to the implementation scenario. If any overall matching degree is less than 0.9, the target area corresponding to the overall matching degree is marked as a suspected scratch area, thereby obtaining all suspected scratch areas.

[0065] Due to the possible existence of light reflection, the texture features of the local area in the target area disappear and the grayscale value of the local area changes, resulting in a reduction in the number of reference points in the local area of ​​the target area, resulting in mismatch between the local area of ​​some target areas and the scratch-free template area. Therefore, it is necessary to re-obtain the local matching degree of the local area of ​​the suspected scratch area and perform outlier analysis to obtain the final matching degree. Based on the final matching degree, the scratch sub-area in the suspected scratch area is screened again. Specifically, taking any suspected scratch area as an example, the suspected scratch area is matched with the scratch-free template area for local sub-areas to obtain the scratch sub-areas in the suspected scratch area. The specific method for obtaining the scratch sub-area is as follows:

[0066] (1) According to the local matching degree of each sub-region pair between the suspected scratch region and the scratch-free template region, the sub-region pair corresponding to the local matching degree less than the matching degree threshold is marked as a target sub-region pair, and all target sub-region pairs are obtained; for any target sub-region pair, each sub-region in the target sub-region pair is divided into at least one local sub-region of a preset size, and each sub-region in the target sub-region pair is matched with the local sub-regions at the same position to obtain a matching local sub-region pair.

[0067] In one implementation, taking the obtained mth suspected scratch region as an example, the local matching degree of each sub-region pair between the mth suspected scratch region and the scratch-free template region is known. If the local matching degree is less than the matching value threshold, the sub-region pair corresponding to the local matching degree is marked as the target sub-region pair, and all target sub-region pairs corresponding to the mth suspected scratch region are obtained according to the local matching degree of each sub-region pair between the mth suspected scratch region and the scratch-free template region. It should be noted that the matching degree threshold of this solution is 0.9, and the matching degree threshold is the same as the matching degree threshold for the above-mentioned comparison with the overall matching degree.

[0068] Taking the i-th target sub-region pair obtained as an example, each sub-region in the i-th target sub-region pair is divided into at least one local sub-region with a size of 3*3, and the local sub-regions at the same position in each sub-region in the i-th target sub-region pair are matched to obtain Z matched local sub-region pairs.

[0069] It should be noted that the embodiment of the present invention does not impose any requirements on the size of the divided local sub-regions, which can be set according to the implementation scenario.

[0070] (2) for any local sub-region pair, obtaining a reference point quantity difference according to the number of reference points of each local sub-region in the local sub-region pair, and if the reference point quantity difference is less than or equal to a preset value, marking the local sub-region in the local sub-region pair belonging to the suspected scratch region as a scratch sub-region;

[0071] If the reference point quantity difference is greater than a preset value, a final matching degree of the local sub-region pair is obtained, and when the final matching degree is less than the matching degree threshold, a local sub-region in the local sub-region pair belonging to the suspected scratch region is marked as a scratch sub-region;

[0072] In one embodiment, based on the obtained Z local sub-region pairs, taking the zth local sub-region pair obtained by matching as an example, the number of reference points in each local sub-region in the zth local sub-region pair is counted, and the difference D in the number of reference points between the local sub-regions in the zth local sub-region pair is calculated, and the expression D=F zt -Qzt , where F zt represents the number of reference points in the local sub-region belonging to the scratch-free template region in the z-th local sub-region pair, Q zt It represents the number of reference points in the local sub-region belonging to the m-th suspected scratch region in the z-th local sub-region pair.

[0073] Since the edge points of the scratch area include not only the texture edge points of the flange surface, but also the scratch edge points, the number of reference points will increase. Therefore, if D≤0, the local sub-region belonging to the m-th suspected scratch region in the z-th local sub-region pair is marked as a scratch sub-region.

[0074] If D>0, the final matching degree of the zth local sub-region pair is calculated. The matching degree threshold set in the embodiment of the present invention is 0.9, which is the same matching degree threshold as the matching degree threshold for comparison with the overall matching degree. If the final matching degree of the zth local sub-region pair is less than 0.9, the local sub-region in the zth local sub-region pair belonging to the suspected scratch region is marked as a scratch sub-region.

[0075] Wherein, obtaining the final matching degree of the local sub-region pair includes:

[0076] 1) Marking a local subregion belonging to the suspected scratch region in the local subregion pair as a first subregion, obtaining an illuminated edge point in the first subregion, and dividing the first subregion into an illuminated region and a non-illuminated region according to the illuminated edge point.

[0077] Specifically, a pixel point corresponding to the maximum gradient value in the first sub-region is obtained as an illumination edge point, and according to the reference point position in the local sub-region belonging to the scratch-free template region in the local sub-region pair, a pixel point at the same position as the reference point is obtained in the first sub-region as a first pixel point, and a first pixel point that does not belong to the reference point in the first pixel point is marked as a second pixel point, and according to the grayscale value of each pixel point within the local neighborhood of each second pixel point, a grayscale value range corresponding to each second pixel point is obtained, and a union operation is performed on all grayscale value ranges and the grayscale value of the illumination edge point to obtain a final grayscale value range;

[0078] For any second pixel point, obtain a non-reference point within a local neighborhood of the second pixel point, and if the grayscale value of any non-reference point is within the final grayscale value range, mark the non-reference point as an illuminated pixel point, take any illuminated pixel point as the second pixel point, repeat the steps for the second pixel point, and obtain all illuminated pixels points;

[0079] All illuminated pixel points corresponding to each of the second pixel points are obtained to form an illuminated area of ​​the first sub-area, and the remaining area except the illuminated area is used as a non-illuminated area of ​​the first sub-area.

[0080] In one embodiment, taking the obtained zth local sub-region pair as an example, the local sub-region belonging to the suspected scratch region in the zth local sub-region pair is marked as the first sub-region, and the local sub-region belonging to the scratch-free template region in the zth local sub-region pair is marked as the second sub-region, and the Sobel operator is used to obtain the gradient values ​​of all pixels in the first sub-region, wherein the Sobel operator belongs to the prior art and will not be described herein, and the pixel point corresponding to the maximum gradient value is taken as the illuminated edge point, and according to the position of the reference point in the second sub-region, the pixel point with the same position as the reference point in the second sub-region is obtained in the first sub-region as the first pixel point.

[0081] For example, let the coordinates of all reference points in the second sub-region be (x 1 ,y 1 ),(x 2 ,y 2 )…(x i ,y i ), according to the coordinates of each reference point in the second sub-region, obtain the pixel point with the same position coordinates in the first sub-region as the first pixel point, and the coordinates of the first pixel point are (x 1 ',y 1 '),(x 2 ',y 2 ')…(x i ',y i '), where x i =x i ',y i =y i '.

[0082] Mark all first pixel points that are not reference points as second pixel points, perform eight-neighborhood traversal for each second pixel point, obtain non-reference points within the eight-neighborhood of each second pixel point, and for the eight-neighborhood of any second pixel point, obtain the grayscale values ​​of all non-reference points in the eight-neighborhood, count the maximum and minimum grayscale values, and obtain the grayscale value range of the eight-neighborhood. Perform a union operation on the grayscale value range within the eight-neighborhood of each second pixel point and the grayscale value of the illuminated edge point to obtain the final grayscale value range.

[0083] For any second pixel point, perform eight-neighborhood traversal on the second pixel point to obtain non-reference points within the local neighborhood of the second pixel point. If the grayscale value of any non-reference point is within the final grayscale value range, the non-reference point is marked as an illuminated pixel point, and then any illuminated pixel point is used as the second pixel point, and the above steps for the second pixel point are repeated to obtain all illuminated pixel points after non-reference point traversal with the second pixel point as the starting point. Similarly, obtain all illuminated pixel points after non-reference point traversal with each second pixel point as the starting point.

[0084] All illuminated pixel points in the first sub-region are obtained after non-reference point traversal based on each second pixel point as the starting point, the area composed of all the illuminated pixel points is taken as the illuminated area in the first sub-region, and the remaining area except the illuminated area is taken as the non-illuminated area in the first sub-region.

[0085] 2) Obtaining the number of lost reference points in the first sub-area.

[0086] Specifically, a reference point closest to the illuminated edge point in the horizontal direction in the first sub-region is obtained as a candidate point, a distance between the candidate point and the nearest non-illuminated edge point to the candidate point in the horizontal direction is calculated as a first standard distance, and a distance between the candidate point and the nearest reference point to the candidate point in the lower vertical direction is calculated as a second standard distance;

[0087] The reference point closest to the candidate point in the lower vertical direction is taken as the candidate target point, and whether there are reference points on both horizontal sides of the candidate target point is detected. If not, the candidate target point is taken as the candidate point, and the steps of the candidate target point are repeated until it is detected that there are reference points on both horizontal sides of the candidate target point. Then, the candidate target point is taken as the target point, and any reference point on both horizontal sides of the target point is taken as the candidate target point, and the steps of the candidate target point are repeated to obtain all target points;

[0088] For any target point, detect whether the distance difference between the target point and its left and right adjacent reference points and the first standard distance is within a preset distance difference range; if both are within the preset distance difference range, select a reference point on the same side as the illuminated edge point from the left adjacent reference point and the right adjacent reference point of the target point as a base point, obtain a pixel point in the upper vertical direction of the base point that is equal to the second standard distance as a reference base point, calculate the angle difference of the gradient direction angle between the base point and the reference base point, and if the angle difference is within a preset angle difference threshold range and the reference base point belongs to a pixel point in the illuminated area, mark the reference base point as a lost reference point, take the lost reference point as the base point, repeat the base point step, and obtain all lost reference points corresponding to the target point;

[0089] All lost reference points corresponding to each of the target points are acquired as lost reference points in the first sub-area, and the number of lost reference points in the first sub-area is counted.

[0090] In one embodiment, taking the obtained zth local sub-region pair as an example, referring to Figure 3 , which is a schematic diagram of obtaining the lost reference points of the local sub-region (that is, the first sub-region) in the suspected scratch region in the z-th local sub-region pair. In the horizontal direction of the first sub-region, by calculating the illumination edge point ( Figure 3 The Euclidean distance between the edge point in the image and its adjacent reference point is used to obtain the reference point closest to the illuminated edge point ( Figure 3 The reference point 1 in is taken as a candidate point, and the Euclidean distance between the candidate point and its nearest non-illuminated edge point is calculated as the first standard distance ( Figure 3 S1 in the figure), calculate the reference point that is closest to the candidate point in the vertical direction below it ( Figure 3 The Euclidean distance between the reference points 2 in the Figure 3 h1 in the text file.

[0091] In the lower vertical direction of the first sub-region, the reference point closest to the candidate point is taken as the candidate target point ( Figure 3 ), detect whether there are reference points on both sides of reference point 2. If not, take reference point 2 as a candidate point and repeat the above candidate target point steps until a candidate target point is detected ( Figure 3If there are reference points on both sides of the horizontal plane of the reference point 3 in the figure, then continue to detect whether the first distance difference between the reference point 3 and its left and right adjacent reference points and the first standard distance are both within the preset distance difference range. The method for obtaining the preset distance difference range in this scheme is: according to the reference point closest to the illuminated edge point in the horizontal direction, obtain the four adjacent reference points in the horizontal direction of the reference point, calculate the horizontal distance between any two adjacent reference points respectively, obtain the mean of the horizontal distance, calculate the first standard difference d1 between the mean of the horizontal distance and the first standard distance, and use [0, d1] as the preset distance difference range.

[0092] If the first distance difference is within the range of [0, d1], the candidate target point is taken as the target point ( Figure 3 After determining the target point, obtain the reference point on the same side of the illuminated edge point in the horizontal direction of reference point 3 as the base point ( Figure 3 4 in the image), obtain a pixel point with the same distance from the second standard in the vertical direction above the reference point 4 as the reference base point (that is, the reference point 2 is lost and h1=h2), use the Sobel operator to obtain the gradient value of the base point and the reference base point, and then use the arctan function to calculate the angle difference of the gradient direction angle between the base point and the reference base point, and detect whether the angle difference is within the preset angle difference threshold range. The method for obtaining the preset angle difference threshold range in this scheme is: according to the reference point closest to the illuminated edge point in the horizontal direction, obtain the three adjacent reference points in the vertical direction of the reference point, respectively calculate the angle difference of the gradient direction angle between any two adjacent reference points, and obtain the maximum value g between all angle differences. max and the minimum value g min , get the preset angle difference threshold range [g min , g max ].

[0093] If the angle difference is in [g min , g max ] range, and the reference base point belongs to a pixel point in the illuminated area of ​​the first sub-area, then the reference base point is marked as a lost reference point ( Figure 3 The lost reference point 2 in the image is then used as the base point, and the steps of the base point are repeated to obtain all the lost reference points corresponding to the target point. The sobel operator and the arctan function belong to the prior art and will not be described here.

[0094] After obtaining the traversal of the lost reference points with reference point 3 as the target point, the reference point on the same side of the illuminated edge point in the horizontal direction of reference point 3 is taken as a candidate target point ( Figure 34 in the figure), calculate the distance S2 between reference point 4 and reference point 3 and the distance S3 between reference point 4 and reference point 5. If the distance difference |S3-S2| between distance S2 and distance S3 is in the range [0, d1], then mark reference point 4 as the target point. Similarly, according to the above method of traversing lost reference points with reference point 3 as the target point, obtain all lost reference points corresponding to reference point 4 as the target point. By analogy, take the reference point on the same side of the illuminated edge point in the horizontal direction of reference point 4 as the candidate target point ( Figure 3 5), repeat the above steps of candidate target points to obtain all lost reference points in the first sub-region, and count the number of lost reference points in the first sub-region.

[0095] It should be noted that in the embodiment of the present invention, there is no limitation on the above-mentioned preset distance difference range and angle difference threshold range, which can be set according to the implementation scenario.

[0096] 3) obtaining the local matching degree of the local sub-region pair, calculating the first quantity ratio between the reference point quantity difference and the quantity of the lost reference points, obtaining the first grayscale mean of all pixels in the illuminated area and the second grayscale mean of all pixels in the non-illuminated area, calculating the first grayscale mean difference absolute value between the first grayscale mean and the second grayscale mean, respectively obtaining the grayscale mean of each local sub-region in the local sub-region pair, obtaining the second grayscale mean difference absolute value, and obtaining the difference absolute value ratio between the first grayscale mean difference absolute value and the second grayscale mean difference absolute value;

[0097] The local matching degree of the local sub-region pair, the average value of the first quantity ratio and the difference absolute value ratio are calculated as the final matching degree of the local sub-region pair.

[0098] In one embodiment, taking the obtained zth local sub-region pair as an example, firstly, the local matching degree of the zth local sub-region pair is obtained according to the above method for obtaining the local matching degree of the sub-region pair, the number of reference points of each local sub-region in the zth local sub-region pair is calculated to obtain the reference point number difference, and according to the number of lost reference points and the reference point number difference in the local sub-region (that is, the first sub-region) belonging to the suspected scratch region in the zth local sub-region pair, a first quantity ratio between the number of lost reference points and the reference point number difference is calculated, and according to the illuminated region and the non-illuminated region in the first sub-region, the first quantity ratio of all pixel points in the illuminated region is calculated. The first grayscale mean and the second grayscale mean of all pixels in the unilluminated area are calculated, the first grayscale mean difference absolute value between the first grayscale mean and the second grayscale mean is calculated, the grayscale mean of each local sub-area in the z-th local sub-area pair is obtained, the second grayscale mean difference absolute value between the grayscale means is obtained, the difference absolute value ratio between the first grayscale mean difference absolute value and the second grayscale mean difference absolute value is obtained, and the local matching degree of the z-th local sub-area pair, the first quantity ratio and the difference absolute value ratio are calculated as the final matching degree of the z-th local sub-area pair. Then, the final matching degree of the z-th local sub-area pair is expressed as:

[0099]

[0100] Among them, S z ' represents the final matching degree of the zth local sub-region pair, T z represents the local matching degree of the zth local sub-region pair, d z represents the number of lost reference points in the first sub-region of the z-th local sub-region pair, F zt represents the number of reference points in the local sub-region (that is, the second sub-region) belonging to the scratch-free template region in the z-th local sub-region pair, Q zt represents the number of reference points in the first sub-region, F zh represents the mean grayscale value of the pixels in the second sub-region of the z-th local sub-region pair, Q zh represents the mean grayscale value of the pixels in the first sub-region of the z-th local sub-region pair, represents the mean grayscale value of the pixels in the illuminated area in the first sub-area, h zi represents the gray value of the i-th pixel in the first sub-region belonging to the illuminated area, represents the mean grayscale value of the pixels in the first sub-region that are in the unilluminated area, h zj represents the gray value of the jth pixel in the first sub-region that belongs to the unilluminated area, || represents the absolute value sign, N i Indicates the number of pixels in the first sub-area that belong to the illuminated area, N jIndicates the number of pixels in the first sub-region that are in the dark area.

[0101] It should be noted that reflects the existence rate of missing reference points, The smaller the value is, the less likely it is that the reference point is lost due to illumination, and the more likely it is that the local sub-region in the z-th local sub-region pair that belongs to the suspected scratch region is the scratch sub-region. It reflects the difference in grayscale value changes between the local sub-region belonging to the suspected scratch area in the z-th local sub-region pair and the local sub-region belonging to the scratch-free template area in the z-th local sub-region pair. The smaller the value of , the smaller the grayscale value change in the local subregion belonging to the suspected scratch region in the zth local subregion pair is, and the greater the possibility that the local subregion belonging to the suspected scratch region in the zth local subregion pair is the scratch subregion.

[0102] Similarly, repeat the above steps of obtaining the scratch subregion in the zth local subregion pair to obtain all the scratch subregions in each local subregion pair in the ith target subregion pair, obtain all the scratch subregions in each local subregion pair in the ith target subregion pair respectively, and use all the scratch subregions in each local subregion pair in all target subregion pairs as all the scratch subregions in the mth suspected scratch region. Finally, based on all the scratch subregions in each suspected scratch region, obtain the scratch subregions in all suspected scratch regions.

[0103] At this point, all scratch sub-regions in each suspected scratch region are obtained.

[0104] The model building module 14 is used to mark the grayscale image according to all scratch sub-areas to obtain a marked image, use the marked image to train the quality detection model to obtain a trained quality detection model, and use the trained quality detection model to perform real-time quality detection on the flange surface in the electronically controlled silicone oil clutch.

[0105] According to the above steps of obtaining the scratch sub-area, the obtained grayscale image of the flange surface of the electronically controlled silicone oil clutch is marked according to all the scratch sub-areas to obtain the marked image, and all the marked images are divided into a training set and a test set. The training set is used to train the quality detection model, and the test set is used to evaluate the performance of the quality detection model to obtain the final trained quality detection model, wherein the neural network model used in the training quality detection model mainly includes a convolutional neural network model. It should be noted that since the training convolutional neural network model is a well-known technology, it will not be described here. The trained quality detection model is used to perform real-time quality detection on the flange surface of the electronically controlled silicone oil clutch.

[0106] At this point, the quality inspection model is used to perform real-time quality inspection on the flange surface of the electronically controlled silicone oil clutch.

[0107] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included in the protection scope of the present invention.

Claims

1. An electronically controlled silicone oil clutch image analysis system based on big data, characterized in that: The system comprises: An image processing module is used to obtain a surface image of a flange surface in an electronically controlled silicone oil clutch, and grayscale the surface image to obtain a corresponding grayscale image; A template acquisition module, used for performing partition processing on the grayscale image according to the hole positions on the flange surface to obtain at least one target area containing a single hole, and for any target area, obtaining the gradient values ​​of all pixels in the target area, and obtaining the texture distribution feature value of the target area according to the gradient value difference of each pixel in the target area, and obtaining a scratch-free template area according to the texture distribution feature value of each target area; A region screening module, used to respectively obtain the overall matching degree between the scratch-free template region and each of the target regions, screen all target regions to obtain suspected scratch regions according to the overall matching degree, and for any suspected scratch region, perform local sub-region matching between the suspected scratch region and the scratch-free template region to obtain a scratch sub-region in the suspected scratch region, and obtain a scratch sub-region in each of the suspected scratch regions; The model building module is used to mark the grayscale image according to all scratch sub-areas to obtain a marked image, use the marked image to train the quality detection model to obtain a trained quality detection model, and use the trained quality detection model to perform real-time quality detection on the flange surface in the electronically controlled silicone oil clutch.

2. The electronically controlled silicone oil clutch image analysis system based on big data according to claim 1 is characterized in that: The template acquisition module performs partition processing on the grayscale image according to the hole positions on the flange surface to obtain at least one target area containing a single hole, including: The inner edge line of the flange surface and the outer edge line of the flange surface in the grayscale image are obtained to obtain a circular area formed by the inner edge line of the flange surface and the outer edge line of the flange surface, the number of holes in the grayscale image is counted, and the circular area is evenly divided into N target areas equal to the number of the holes, and one of the target areas contains one hole.

3. The electronically controlled silicone oil clutch image analysis system based on big data according to claim 2 is characterized in that: The template acquisition module acquires the texture distribution feature value of the target area according to the gradient value difference of each pixel point in the target area, including: If the gradient value of any pixel point in the target area is greater than a preset gradient threshold, the pixel point is marked as a reference point to obtain all reference points in the target area; All reference points are clustered to obtain at least one cluster, the total number of clusters and the number of reference points in each cluster are counted, the vertical distance between the cluster center point of each cluster and the inner edge line of the flange surface in the target area is calculated respectively, and all clusters are arranged in descending order of the vertical distance to obtain a cluster sequence. For any reference point in the target area, all reference points contained in the neighborhood of the reference point and the reference point form a reference point set of the reference point, and according to the reference point set of each reference point in the target area, the total number of reference point sets and the number of reference points in each reference point set are counted, and all reference point sets are sorted according to the traversal order of the reference points in the target area to obtain a reference point set sequence; For an r-th reference point set in the reference point set sequence, calculate an absolute value of a difference between the number of reference points in the r-th reference point set and the number of reference points in the r-th cluster in the cluster sequence, calculate an average of all absolute values ​​of the differences according to the total number of reference point sets, and obtain a first addition result between a constant 1 and the average; Calculating a first absolute value of a difference between the total number of the reference point set and the total number of clusters, and obtaining a second addition result between a constant 1 and the first absolute value of the difference; The product of the first addition result and the second addition result is used as the texture distribution feature value of the target area.

4. The electronically controlled silicone oil clutch image analysis system based on big data according to claim 1, characterized in that: The template acquisition module obtains a scratch-free template area according to the texture distribution feature value of each target area, including: According to the texture distribution characteristic value of each target area, the target area corresponding to the minimum value of the texture distribution characteristic value is used as the scratch-free template area.

5. The electronically controlled silicone oil clutch image analysis system based on big data according to claim 3 is characterized in that: The region screening module obtains the overall matching degree between the scratch-free template region and each of the target regions respectively, and screens suspected scratch regions in all target regions according to the overall matching degree, including: For any target area, the target area and the scratch-free template area are respectively divided into at least one sub-area of ​​a preset size, and the sub-areas of the target area and the scratch-free template area are matched at the same position to obtain a matching sub-area pair, and the local matching degree of each sub-area pair is obtained according to the difference in the grayscale value and the number of reference points in the sub-area pair; The average of all local matching degrees is calculated as the overall matching degree between the target area and the scratch-free template area. According to the overall matching degree between each target area and the scratch-free template area, if any overall matching degree is less than a preset matching degree threshold, the target area corresponding to the overall matching degree is marked as a suspected scratch area, and all suspected scratch areas are obtained.

6. The electronically controlled silicone oil clutch image analysis system based on big data according to claim 5 is characterized in that: The region screening module obtains the local matching degree of each sub-region pair according to the difference in the region grayscale value and the number of reference points in the sub-region pair, including: For any pair of sub-regions, respectively obtain the number ratio between the number of pixel points and the number of reference points in each sub-region of the pair of sub-regions, obtain the subtraction absolute value of the number ratios, and obtain the subtraction result between a constant 1 and the subtraction absolute value; Using a structural similarity index to obtain the local similarity of the sub-region pair; A weighted sum is performed on the local similarity and the subtraction result of the sub-region pair to obtain a corresponding weighted sum result as the local matching degree of the sub-region pair.

7. The electronically controlled silicone oil clutch image analysis system based on big data according to claim 5 is characterized in that: The region screening module performs local sub-region matching between the suspected scratch region and the scratch-free template region to obtain a scratch sub-region in the suspected scratch region, including: According to the local matching degree of each sub-region pair between the suspected scratch region and the scratch-free template region, the sub-region pairs corresponding to the local matching degree less than the matching degree threshold are marked as target sub-region pairs, and all target sub-region pairs are obtained; For any target sub-region pair, each sub-region in the target sub-region pair is divided into at least one local sub-region of a preset size, and each sub-region in the target sub-region pair is matched with local sub-regions at the same position to obtain a matching local sub-region pair; For any local sub-region pair, a reference point quantity difference is obtained according to the number of reference points of each local sub-region in the local sub-region pair, and if the reference point quantity difference is less than or equal to a preset value, the local sub-region in the local sub-region pair belonging to the suspected scratch region is marked as a scratch sub-region; If the reference point quantity difference is greater than a preset value, a final matching degree of the local sub-region pair is obtained, and when the final matching degree is less than the matching degree threshold, a local sub-region in the local sub-region pair belonging to the suspected scratch region is marked as a scratch sub-region; The scratch subregion in each of the local subregion pairs is obtained to obtain all the scratch subregions in the target subregion pairs; and all the scratch subregions in each of the target subregion pairs are obtained to obtain the scratch subregion in the suspected scratch region.

8. The electronically controlled silicone oil clutch image analysis system based on big data according to claim 7 is characterized in that: The region screening module obtains the final matching degree of the local sub-region pair, including: Obtaining a local matching degree of the local sub-region pair, marking a local sub-region in the local sub-region pair belonging to the suspected scratch region as a first sub-region, obtaining an illuminated edge point in the first sub-region, dividing the first sub-region into an illuminated region and a non-illuminated region according to the illuminated edge point, obtaining the number of lost reference points in the first sub-region, and calculating a first quantity ratio between the reference point quantity difference and the number of lost reference points; Obtain a first grayscale mean of all pixels in the illuminated area and a second grayscale mean of all pixels in the non-illuminated area, calculate a first grayscale mean difference absolute value between the first grayscale mean and the second grayscale mean, respectively obtain a grayscale mean of each local sub-area in the local sub-area pair, obtain a second grayscale mean difference absolute value, and obtain a difference absolute value ratio between the first grayscale mean difference absolute value and the second grayscale mean difference absolute value; The local matching degree of the local sub-region pair, the average value of the first quantity ratio and the difference absolute value ratio are calculated as the final matching degree of the local sub-region pair.

9. The electronically controlled silicone oil clutch image analysis system based on big data according to claim 8, characterized in that: The region screening module obtains the illumination edge points in the first sub-region, and divides the first sub-region into an illumination region and a non-illumination region according to the illumination edge points, including: Obtain a pixel point corresponding to the maximum gradient value in the first sub-region as an illuminated edge point, and according to a reference point position in a local sub-region belonging to the scratch-free template region in the local sub-region pair, obtain a pixel point at the same position as the reference point in the first sub-region as a first pixel point, mark a first pixel point that does not belong to the reference point in the first pixel points as a second pixel point, and obtain a grayscale value range corresponding to each second pixel point according to the grayscale value of each pixel point within a local neighborhood of each second pixel point, and perform a union operation on all grayscale value ranges and the grayscale value of the illuminated edge point to obtain a final grayscale value range; For any second pixel point, obtain a non-reference point within a local neighborhood of the second pixel point, and if the grayscale value of any non-reference point is within the final grayscale value range, mark the non-reference point as an illuminated pixel point, take any illuminated pixel point as the second pixel point, repeat the steps for the second pixel point, and obtain all illuminated pixels points; All illuminated pixel points corresponding to each of the second pixel points are obtained to form an illuminated area of ​​the first sub-area, and the remaining area except the illuminated area is used as a non-illuminated area of ​​the first sub-area.

10. The electronically controlled silicone oil clutch image analysis system based on big data according to claim 8, characterized in that: The region screening module obtains the number of lost reference points in the first sub-region, including: Acquire a reference point closest to the illuminated edge point in the horizontal direction within the first sub-region as a candidate point, calculate the distance between the candidate point and the nearest non-illuminated edge point to the candidate point in the horizontal direction as a first standard distance, and calculate the distance between the candidate point and the nearest reference point to the candidate point in the lower vertical direction as a second standard distance; The reference point closest to the candidate point in the lower vertical direction is taken as the candidate target point, and whether there are reference points on both horizontal sides of the candidate target point is detected. If not, the candidate target point is taken as the candidate point, and the steps of the candidate target point are repeated until it is detected that there are reference points on both horizontal sides of the candidate target point. Then, the candidate target point is taken as the target point, and any reference point on both horizontal sides of the target point is taken as the candidate target point, and the steps of the candidate target point are repeated to obtain all target points; For any target point, detect whether the distance difference between the target point and its left and right adjacent reference points and the first standard distance is within a preset distance difference range; if both are within the preset distance difference range, select a reference point on the same side as the illuminated edge point from the left adjacent reference point and the right adjacent reference point of the target point as a base point, obtain a pixel point in the upper vertical direction of the base point that is equal to the second standard distance as a reference base point, calculate the angle difference of the gradient direction angle between the base point and the reference base point, and if the angle difference is within a preset angle difference threshold range and the reference base point belongs to a pixel point in the illuminated area, mark the reference base point as a lost reference point, take the lost reference point as the base point, repeat the base point step, and obtain all lost reference points corresponding to the target point; All lost reference points corresponding to each of the target points are acquired as lost reference points in the first sub-area, and the number of lost reference points in the first sub-area is counted.

Images