A method and system for visual inspection of leather wear quality

Abstract

The present application relates to the technical field of industrial image processing, in particular to a kind of leather wear quality visual inspection method and system.The method comprises: based on multiple smooth scales, leather image is smoothed, obtains multiple smooth images, the elimination degree of leather texture in the multiple smooth images is different;Determine the wear degree of each smooth image;Based on the wear degree of each smooth image and the smooth scale of each smooth image, determine the real wear degree of leather;Determine the quality of leather based on the real wear degree.The method can accurately determine the quality of leather.

Description

Technical Field

[0001] This invention relates to the field of industrial image processing technology, specifically to a visual inspection method and system for leather wear quality. Background Technology

[0002] In the production of leather goods, worn areas can affect the product's appearance and lifespan. By inspecting worn areas, quality problems can be detected in a timely manner, ensuring that the product meets quality standards.

[0003] In related technologies, the main method for detecting leather surface wear is manual visual inspection. Professionals rely on the naked eye to observe the leather surface and judge the wear condition of the leather, such as whether there are scratches, cracks or discoloration, and thus determine the quality of the leather.

[0004] However, manual visual inspection is easily affected by subjective factors such as fatigue and experience differences of the inspectors, which may lead to inaccurate wear judgment and lower accuracy in leather quality inspection. Summary of the Invention

[0005] To address the technical problem of low accuracy in leather quality inspection due to manual visual inspection, the present invention aims to provide a method and system for visual inspection of leather wear quality. The specific technical solution adopted is as follows:

[0006] This application provides a visual detection method for leather wear quality, comprising: smoothing a leather image based on multiple smoothing scales to obtain multiple smoothed images, wherein the degree of leather texture elimination in the multiple smoothed images is different; determining the wear degree of each smoothed image; determining the true wear degree of the leather based on the wear degree of each smoothed image and the smoothing scale of each smoothed image; and determining the leather quality based on the true wear degree.

[0007] Optionally, determining the wear area of ​​each smoothed image specifically includes: determining the best smoothed image from the plurality of smoothed images, the best smoothed image being a smoothed image capable of eliminating the texture of the leather itself; performing threshold segmentation on the best smoothed image to obtain an initial wear area; and determining the wear areas of other smoothed images based on the position of the initial wear area on the best smoothed image.

[0008] Optionally, determining the best smoothed image from the plurality of smoothed images specifically includes: determining the texture representation degree of each smoothed image in ascending order of smoothing scale, until the texture representation degree of the first smoothed image is less than the texture representation degree threshold, wherein the first smoothed image is any one of the plurality of smoothed images, and the texture representation degree is used to characterize the degree of display of leather texture in a smoothed image; and determining the first smoothed image as the best smoothed image.

[0009] Optionally, determining the texture representation of the first smoothed image includes: performing threshold segmentation on the first smoothed image to obtain worn pixels and non-worn pixels in the first smoothed image; and determining the texture representation of the first smoothed image based on the number of worn pixels and non-worn pixels.

[0010] Optionally, the initial wear region includes at least one first wear connected region. Based on the position of the initial wear region on the optimal smooth image, determining the wear region corresponding to the second smooth image includes: determining the geometric center of the at least one first wear connected region, wherein the second smooth image is a smooth image corresponding to a second smooth scale, the second smooth scale is an adjacent smooth scale smaller than the maximum smooth scale, and the maximum smooth scale is the smooth scale corresponding to the optimal smooth image; determining the geometric center of the at least one first wear connected region as a cluster center, performing a clustering operation on the second smooth image to obtain at least one cluster; and determining the region formed by the outer contours of the at least one cluster as the wear region of the second smooth image.

[0011] Optionally, the wear degree of the second smoothed image is determined based on the area and number of connected components of the worn region of the second smoothed image. Specifically, this includes: determining the number of worn connected components corresponding to each cluster; determining the ratio between the number of worn connected components corresponding to the cluster and the number of worn connected components of the largest cluster as the weight of each cluster, wherein the largest cluster is the cluster that includes the most worn connected components; and determining the wear degree of the second smoothed image based on the weight of each cluster and the area of ​​each cluster.

[0012] Optionally, the visual inspection method for leather wear quality further includes: performing morphological erosion on the wear area of ​​each smooth image to obtain the number of connected regions of the wear area of ​​each smooth image.

[0013] Optionally, the visual inspection method for leather wear quality further includes: acquiring an initial leather image; performing grayscale processing on the initial leather image to obtain an initial grayscale image; and identifying the initial grayscale image as the leather image.

[0014] This application also provides a visual inspection system for leather wear quality, including a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, it implements any of the steps of the visual inspection method for leather wear quality described above.

[0015] The present invention has the following beneficial effects:

[0016] The leather wear quality visual inspection method provided in this application can smooth a leather image based on multiple smoothing scales to obtain multiple smoothed images, and then determine the wear degree of each smoothed image; then, based on the wear degree of each smoothed image and the smoothing scale of each smoothed image, determine the true wear degree of the leather; finally, determine the leather quality based on the true wear degree. Since the degree of leather texture elimination in the multiple smoothed images obtained based on multiple smoothing scales is different, the degree of wear texture elimination obtained is also different. Therefore, the wear degree under multiple smoothing scales can be obtained. The true wear degree of the leather obtained based on the wear degree under multiple smoothing scales is more realistic and reliable, and the leather quality can be accurately determined based on this. Attached Figure Description

[0017] To more clearly illustrate the technical solutions and advantages 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.

[0018] Figure 1 A flowchart illustrating a visual inspection method for leather wear quality according to an embodiment of the present invention;

[0019] Figure 2 This is a flowchart of another visual inspection method for leather wear quality provided in one embodiment of the present invention. Detailed Implementation

[0020] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of a visual inspection method and system for leather wear quality proposed according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0021] 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 invention pertains.

[0022] In the production of leather goods, worn areas can affect the product's appearance and lifespan. By inspecting worn areas, quality problems can be detected in a timely manner, ensuring that the product meets quality standards.

[0023] Currently, leather surface wear detection mainly relies on manual visual inspection or traditional image processing techniques. Manual visual inspection depends on professionals who observe the leather surface with the naked eye to determine the wear condition, such as scratches, cracks, or discoloration. This method is intuitive and does not require complex equipment, but it is inefficient and easily affected by subjective factors such as inspector fatigue and experience differences, leading to a high misjudgment rate.

[0024] Traditional image processing techniques analyze leather images using computer vision algorithms (such as edge detection, thresholding, or morphological manipulation) to automatically identify worn areas. This method is highly automated, objective, and can quickly process large numbers of samples and generate quantitative data, greatly improving detection efficiency and consistency.

[0025] However, existing detection technologies still have the following drawbacks: Because leather surfaces possess their own texture, and because this natural texture exhibits localized grayscale abrupt changes in images similar to wear, traditional thresholding and edge detection methods often mistakenly classify the inherent texture of the leather surface as wear areas when directly detecting leather wear. This affects the accuracy and reliability of the detection.

[0026] The following description, in conjunction with the accompanying drawings, details the specific scheme of a visual inspection method and system for leather wear quality provided by the present invention.

[0027] Please see Figure 1 The diagram illustrates a flowchart of a visual inspection method for leather wear quality according to an embodiment of the present invention.

[0028] S101. Smooth the leather image based on multiple smoothing scales to obtain multiple smoothed images.

[0029] The degree of leather texture removal varies across the multiple smoothed images.

[0030] In this embodiment, the smoothing scale is a Gaussian filter smoothing scale. The larger the smoothing scale, the greater the degree of elimination of leather texture.

[0031] It is understandable that the greater the degree of removal of leather texture, the greater the degree of removal of texture in the worn area.

[0032] Alternatively, the leather image can be smoothed using high and low filters.

[0033] It should be understood that the degree of leather texture elimination varies in different smoothed images, and the smoothness also varies.

[0034] In one alternative implementation, the leather image can be a leather image captured by an industrial-grade charge-coupled device (CCD) camera.

[0035] Optionally, the camera should have a resolution of 12 megapixels or higher, a fixed-focus macro lens of 60mm, an aperture of f / 8, a shooting distance of 50cm, and an image size of 512×512 pixels.

[0036] In one alternative implementation, the leather image can be a grayscale image obtained by processing the leather image captured by the industrial-grade CCD camera described above.

[0037] Specifically, an initial leather image is acquired using the method described above, and then the initial leather image is converted to grayscale to obtain an initial grayscale image; this initial grayscale image is then identified as the leather image.

[0038] Optionally, grayscale processing can be performed using histograms, and histogram equalization can enhance the overall contrast and eliminate the impact of the background and foreground being too bright or too dark on the image.

[0039] Understandably, performing subsequent image processing based on grayscale images of leather can simplify image processing and make it easier to perform threshold segmentation of leather images based on their grayscale values.

[0040] S102. Determine the degree of wear for each smoothed image.

[0041] It should be understood that smoothing leather images can not only eliminate the texture of the leather itself, but also the wear texture of worn areas on the leather.

[0042] It is understandable that, since different smoothed images have different degrees of smoothness, the wear characteristics of the worn parts on different smoothed images are also different. Therefore, the degree of wear of each smoothed image can be determined.

[0043] Alternatively, the degree of wear on each smoothed image can be determined by the size of the wear area on each smoothed image.

[0044] S103. Determine the true wear level of the leather based on the wear level of each smoothed image and the smoothing scale of each smoothed image.

[0045] It is understandable that the wear level on a single smoothed image cannot represent the true wear level of the leather. Therefore, the true wear level of the leather can be determined by combining the wear levels of multiple smoothed images.

[0046] It should be understood that different smoothing scales produce varying degrees of smoothing on worn and unworn areas of the leather. The contrast between the two areas can, to some extent, reflect the wear depth of the worn area; a higher contrast indicates a deeper wear thickness and a more severe degree of wear. Therefore, different smoothing scales can be assigned corresponding weights to the smoothed images, with a higher smoothing scale resulting in a higher wear weight. Finally, the leather wear degrees from different smoothing scales are weighted and fused to obtain the final leather surface wear degree.

[0047] Optionally, the actual wear and tear of the leather image satisfies the following formula:

[0048] ;

[0049] in, Indicates the true degree of wear and tear on the leather. Indicates the number of multiple smoothed images. Indicates the first The weight of the degree of wear in a smooth image. Indicates the first The degree of wear and tear on a smooth image.

[0050] It should be understood that since the degree of wear is more severe in smoothed images with a larger smoothing scale, the smoothing scale can be used to determine the weight. Furthermore, since the influence coefficient of the smoothing scale on the degree of wear should not be too large, the smoothing scale can be square-rooted to make the weight changes more gradual.

[0051] Optionally, the weights of a smoothed image satisfy the following formula:

[0052] ;

[0053] in, Indicates the first The weight of the degree of wear in a smooth image. Indicates the first Smoothing scale of a smooth image.

[0054] S104. Determine leather quality based on actual wear and tear.

[0055] It should be understood that the greater the actual wear and tear, the worse the quality of the leather; the less the actual wear and tear, the better the quality of the leather.

[0056] In one alternative implementation, a quality range can be defined, and the quality of the leather can be determined based on the range to which the actual wear level belongs.

[0057] In one alternative implementation, the wear level of each smoothed image can be normalized so that the actual wear level is a constant less than or equal to 1, with the quality range set between 0 and 1.

[0058] For example, when the actual wear level is greater than or equal to 0.7, the leather is determined to be severely worn; when the actual wear level is less than 0.7 but greater than or equal to 0.3, the leather is determined to be slightly worn; and when the actual wear level is less than 0.3, the leather is determined to be normal leather.

[0059] In this embodiment, since the degree of leather texture elimination in the multiple smooth images obtained based on multiple smoothing scales is different, the degree of wear texture elimination is also different. Therefore, the wear degree under multiple smoothing scales can be obtained. The true wear degree of leather obtained based on the wear degree under multiple smoothing scales is more realistic and reliable. Based on this, the leather quality can be accurately determined.

[0060] Combination Figure 1 ,like Figure 2 As shown, the above S102 is mainly implemented through S201-S202.

[0061] S201. Determine the wear area for each smoothed image.

[0062] In this context, a wear region consists of at least one wear-connected domain.

[0063] It should be understood that wear on the leather surface is not necessarily a single piece, but may consist of multiple small wear areas. In the embodiments of this application, a wear connectivity region represents a small wear area.

[0064] Understandably, because the sharpness of each smoothed image is different, the number of worn connected regions displayed in the worn area of ​​each smoothed image is also different.

[0065] It should be understood that morphological erosion operations can highlight the main structure and eliminate slight adhesion. In one implementation of this application, morphological erosion can be performed on the worn area of ​​each smooth image to obtain the number of connected regions of the worn area of ​​each smooth image.

[0066] In one alternative implementation, each smoothed image can be thresholded to obtain worn and unworn regions.

[0067] In one implementation of this application, the best smoothed image can be determined from multiple smoothed images first, and the best smoothed image can be thresholded to obtain the initial wear region; then, based on the position of the initial wear region on the best smoothed image, the wear regions of other smoothed images can be determined.

[0068] The best smoothed image is one that can eliminate the texture of the leather itself.

[0069] It is understandable that the texture of leather in different smoothed images varies depending on the smoothing scale. Therefore, for different smoothed images, there exists a smoothing scale that can make the texture of leather just smooth. This smoothed image can be determined as the best smoothed image, and the smoothing scale can be determined as the best smoothing scale.

[0070] In one alternative implementation, the texture representation of each smoothed image can be determined sequentially in ascending order of smoothing scale until a smoothed image with a texture representation less than a texture representation threshold (e.g., the first smoothed image) is found, and then the first smoothed image is determined as the best smoothed image.

[0071] It should be understood that texture representation is used to characterize the degree to which the leather texture of a smooth image is displayed.

[0072] Understandably, the larger the smoothing scale, the smaller the texture representation. When a smaller smoothing scale can just eliminate the leather texture, it is not very meaningful to filter with a larger smoothing scale, and it will reduce the texture representation of the worn area. Therefore, determining the optimal smoothing scale in ascending order can result in a smooth image with the highest clarity that can eliminate the leather texture.

[0073] It should be understood that the first smoothed image is any one of the plurality of smoothed images.

[0074] Optionally, a Gaussian filter intensity range can be selected, and then multiple smoothing scales can be obtained starting from the smallest smoothing scale and following a preset step size.

[0075] In one alternative implementation, any smooth image (e.g., a first smooth image) can be thresholded to obtain worn and unworn pixels in the first smooth image; then, the texture representation of the first smooth image is determined based on the number of worn and unworn pixels.

[0076] Alternatively, the first smoothed image can be segmented using Otsu thresholding.

[0077] Alternatively, the first smoothed image can be segmented based on the Otsu threshold segmentation using the grayscale values ​​of the smoothed image.

[0078] It should be understood that after segmenting the first smooth image, the foreground pixels and background pixels of the first smooth image can be obtained, where the foreground pixels are worn pixels and the background pixels are unworn pixels.

[0079] It is understandable that the number of worn pixels can characterize the texture representation of the worn area, and the number of non-worn pixels can characterize the texture representation of the leather texture. Therefore, the texture representation of the first smooth image can be determined based on the number of worn pixels and the number of non-worn pixels.

[0080] It should be understood that the representation of leather texture needs to be determined based on the pixels of the leather area. Therefore, the leather area needs to be determined from the first smoothed image.

[0081] Specifically, the smoothed image can first be divided into several image blocks. Then, the ratio of foreground pixels to background pixels within each image block is determined. Image blocks with a ratio less than a threshold are identified as leather region image blocks. The average ratio within the leather region image blocks is then used to determine the texture representation of the first smoothed image.

[0082] For example, the size of the image block can be 8*8 pixels, and the percentage threshold can be 0.7.

[0083] Optionally, the texture representation of a smooth image satisfies the following formula:

[0084] ;

[0085] in, Characterizes texture expressiveness. Characterizing the first The proportion of each image patch. This indicates the number of image patches with a percentage less than 0.7.

[0086] It should be understood that, based on the above formula, the leather area can be screened out first, and then the texture representation can be accurately characterized by the ratio of non-worn pixels to worn pixels in the leather area.

[0087] Optionally, when the texture representation is less than the texture representation threshold, the smoothed image can be determined as the best smoothed image.

[0088] It is understandable that since texture representation can characterize the degree of display of leather texture, when the texture representation is less than the texture representation threshold, it means that the leather texture in the smoothed image has been basically eliminated. At this time, the smoothed image is determined as the best smoothed image, and the obtained best smoothed image is more reliable.

[0089] In this embodiment of the application, after obtaining the optimal smooth image, the initial wear area of ​​the optimal smooth image can be obtained based on the above threshold segmentation results.

[0090] Optionally, image blocks with a ratio greater than or equal to 0.7 can be identified as initial wear areas.

[0091] In one alternative implementation, since the multiple smoothed images are obtained from a single leather image, the approximate locations of the wear regions on the multiple smoothed images are similar. After obtaining the initial wear region of the optimal smoothed image, the wear regions of the smoothed images corresponding to adjacent smoothing scales can be traced based on the initial wear region.

[0092] Specifically, assuming the adjacent smoothing scale is the second smoothing scale, and the smoothed image corresponding to the second smoothing scale is the second smoothed image, the steps to determine the wear area of ​​the second smoothed image can be divided into steps one to three.

[0093] Step 1: Determine the geometric center of at least one first worn connected domain.

[0094] Wherein, the at least one first wear connectivity region is a wear connectivity region included in the initial wear region.

[0095] Based on the description of the above embodiments, it should be understood that a wear region includes at least one wear connected domain. In the embodiments of this application, the wear connected domain included in the initial wear region is determined as the first wear connected domain.

[0096] It should be understood that since the initial wear area is the wear area of ​​the best smoothed image, and the smoothing scale of the best smoothed image is the maximum smoothing scale, the second smoothing scale is an adjacent smoothing scale that is smaller than the maximum smoothing scale.

[0097] It is understood that the second smoothing scale is adjacent to the maximum smoothness. Therefore, the sharpness of the second smoothed image is not much different from that of the best smoothed image, and the difference between the wear area on the second smoothed image and the initial wear area is also small. Therefore, the wear area of ​​the second smoothed image can be determined based on the initial wear area.

[0098] Optionally, a morphological closing operation can be performed on the initial wear region to obtain at least one first wear connected region included in the initial wear region.

[0099] In one alternative implementation, the position of the component on the second smoothed image can be determined based on the pixel position of the geometric center of at least one first worn connected domain.

[0100] Step 2: Determine the geometric center of at least one first worn connected domain as the cluster center, and perform a clustering operation on the second smooth image to obtain at least one cluster.

[0101] Specifically, clustering is performed on the second smoothed image using the geometric center of at least one first wear connected region as the cluster center. Regions with close proximity and high pixel density are found in the second smoothed image to obtain at least one cluster. This enables the updating and matching of connected regions in the second smoothed image, and obtains the clusters corresponding to each first wear connected region in the adjacent and smaller smoothed images of the initial wear region.

[0102] Alternatively, clustering can be performed using a clustering algorithm (density-based spatial clustering of applications with noise, DBSCAN).

[0103] Step 3: Determine the area formed by the outer contour of at least one cluster as the wear area of ​​the second smooth image.

[0104] Optionally, the wear region of the smoothed image corresponding to a smaller smoothing scale can be determined based on the geometric center of the wear connected region included in the wear region of the second smoothed image. By repeating steps one to three above, the wear region of each smoothed image can be obtained sequentially.

[0105] It should be understood that, based on steps one to three above, the wear regions of the smoothed images corresponding to adjacent smoothing scales can be obtained sequentially. Since the difference in sharpness between the smoothed images of adjacent smoothing scales is small, the difference in wear regions is also small. Therefore, based on this method, the wear region of each smoothed image can be accurately obtained, and the difference in wear regions of different smoothed images can be clearly shown.

[0106] S202. Based on the area of ​​the worn region and the number of connected components of each smoothed image, determine the degree of wear corresponding to each smoothed image.

[0107] It should be understood that this number of connected components refers to the number of worn connected components in the worn region of each smoothed image.

[0108] It is understandable that, for smoothed images at different smoothing scales, the larger the area of ​​the worn region, the greater the degree of leather wear; and the more connected regions of wear, the more dispersed the leather wear area, and the smaller the degree of wear.

[0109] The following uses the second smoothed image as an example to illustrate the method for determining the degree of wear on a smoothed image.

[0110] Specifically, the number of worn connected components corresponding to each cluster is determined; then the ratio between the number of worn connected components corresponding to the cluster and the number of worn connected components of the largest cluster is determined as the weight of each cluster; finally, based on the weight of each cluster and the area of ​​each cluster, the degree of wear of the second smooth image is determined.

[0111] Among them, the largest cluster is the cluster that includes the most worn connected domains.

[0112] Optionally, the number of worn connected components corresponding to each cluster can be obtained based on the above morphological erosion operation.

[0113] Optionally, the number of worn connected components in each cluster can be normalized to obtain the weight of that cluster.

[0114] Optionally, the weight of a cluster satisfies the following formula:

[0115] ;

[0116] in, Indicates the first The weights of each cluster, Indicates the first The number of worn connected domains included in each cluster. This indicates the number of worn connected components included in the largest cluster.

[0117] It should be understood that determining the weight of a cluster based on the above formula can avoid a weight greater than 1, and can also reflect the size of the number of connected components included in the cluster.

[0118] Alternatively, the degree of wear on a smooth image satisfies the following formula:

[0119] ;

[0120] in, Indicates the degree of wear and tear on a smooth image. This indicates the number of clusters in the smoothed image. Indicates the first The weights of each cluster, Indicates the first The area of ​​each cluster.

[0121] Based on this formula, the area of ​​each cluster can be weighted and summed to obtain the wear level of the smoothed image.

[0122] In this embodiment, since the larger the area of ​​the wear region, the greater the degree of leather wear, and the more connected components of the wear region, the more dispersed the wear region, and the less the degree of leather wear. Therefore, after determining the wear region of each smooth image, the degree of wear of each smooth image can be accurately characterized based on the area of ​​the wear region and the number of connected components of the wear region.

[0123] This application also proposes a visual inspection system for leather wear quality, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the steps of any of the above-mentioned visual inspection methods for leather wear quality.

[0124] It should be noted that the order of the above embodiments of the present invention is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0125] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

Claims

1. A visual inspection method for leather wear quality, characterized in that, The method includes: The leather image is smoothed based on multiple smoothing scales to obtain multiple smoothed images, and the degree of leather texture elimination in the multiple smoothed images is different; Determine the degree of wear for each smoothed image; The smoothing scale of each smoothed image is transformed by the square root to obtain the weight of each smoothed image. The smoothing scale is positively correlated with the weight. The wear level of each smoothed image is weighted and fused based on the weight of each smoothed image to obtain the true wear level of the leather; The quality of the leather is determined based on the actual degree of wear.

2. The method for visually inspecting leather wear quality according to claim 1, characterized in that, Determining the degree of wear for each smoothed image includes: Determine the wear region for each smoothed image, wherein a wear region consists of at least one wear connected region; The degree of wear of each smoothed image is determined based on the area of ​​the worn region and the number of connected components in each smoothed image.

3. The method for visually inspecting leather wear quality according to claim 2, characterized in that, Determining the wear area of ​​each smoothed image includes: The optimal smoothing image is determined from the plurality of smoothing images, wherein the optimal smoothing image is a smoothing image that can eliminate the texture of the leather itself; The optimal smooth image is segmented using a threshold to obtain the initial wear region; Based on the location of the initial wear area on the best smoothed image, the wear areas of other smoothed images are determined.

4. The method for visually inspecting leather wear quality according to claim 3, characterized in that, Determining the optimal smoothed image from the plurality of smoothed images includes: The texture representation degree of each smoothed image is determined sequentially in order of increasing smoothness scale, until the texture representation degree of the first smoothed image is less than the texture representation degree threshold. The first smoothed image is any one of the plurality of smoothed images. The texture representation degree is used to characterize the degree of display of leather texture in a smoothed image. The first smoothed image is determined as the optimal smoothed image.

5. The method for visually inspecting leather wear quality according to claim 4, characterized in that, Determine the texture representation of the first smoothed image, including: Threshold segmentation is performed on the first smoothed image to obtain worn and unworn pixels in the first smoothed image; The texture representation of the first smoothed image is determined based on the number of worn pixels and the number of unworn pixels.

6. The method for visually inspecting leather wear quality according to claim 3, characterized in that, The initial wear region includes at least one first wear connected region. Based on the position of the initial wear region on the optimal smoothed image, the wear region corresponding to the second smoothed image is determined, including: Determine the geometric center of the at least one first worn connected region, the second smoothed image is the smoothed image corresponding to the second smoothed scale, the second smoothed scale is the adjacent smoothed scale smaller than the maximum smoothed scale, and the maximum smoothed scale is the smoothed scale corresponding to the best smoothed image; The geometric center of the at least one first worn connected domain is determined as the cluster center, and a clustering operation is performed on the second smooth image to obtain at least one cluster. The region formed by the outer contours of the at least one cluster is defined as the wear region of the second smoothed image.

7. The method for visually inspecting leather wear quality according to claim 6, characterized in that, The degree of wear in the second smoothed image is determined based on the area of ​​the worn region and the number of connected components, including: Determine the number of worn connected components corresponding to each cluster; The ratio between the number of worn connected components corresponding to the cluster and the number of worn connected components of the largest cluster is determined as the weight of each cluster, where the largest cluster is the cluster with the most worn connected components. The degree of wear on the second smoothed image is determined based on the weight and area of ​​each cluster.

8. The method for visually inspecting leather wear quality according to claim 2, characterized in that, The method further includes: Morphological erosion is performed on the worn region of each smoothed image to obtain the number of connected regions in the worn region of each smoothed image.

9. The method for visually inspecting leather wear quality according to claim 1, characterized in that, The method further includes: Acquire initial leather images; The initial leather image is converted to grayscale to obtain an initial grayscale image; The initial grayscale image is identified as the leather image.

10. A visual inspection system for leather wear quality, characterized in that, The method includes a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor executes the computer program to implement the steps of the visual inspection method for leather wear quality as described in any one of claims 1-9.

Images