Corrugated Carton Quality Inspection System Based on Machine Vision

Abstract

The present invention relates to the technical field of image data processing. More specifically, the present invention relates to a corrugated cardboard box quality detection system based on machine vision. The system includes: an image acquisition module that acquires a surface image of the corrugated cardboard box, where the surface image is a grayscale image; a noise processing module that includes: a parameter unit that constructs a Gaussian filter kernel; an analysis unit that calculates a standard deviation adjustment coefficient and a Gaussian weight adjustment coefficient; a processing unit that performs adaptive Gaussian filtering on the image; a quality detection module that uses an adaptive double-threshold segmentation algorithm to identify corrugation defect regions in the processed image, so as to realize the quality detection of corrugated cardboard boxes based on machine vision. The present invention effectively removes noise and retains image details through an adaptive Gaussian filtering method, avoids the situation where the image enhancement effect in traditional filtering methods is too small or useless, improves the image quality, and enhances the accuracy of subsequent corrugated cardboard box quality detection.

Description

Technical Field

[0001] The present invention relates to the technical field of image data processing. More specifically, the present invention relates to a corrugated cardboard box quality detection system based on machine vision. Background Art

[0002] Corrugated cardboard boxes, due to their advantages such as being lightweight, sturdy, and environmentally friendly, can be applied to the transportation and storage of various commodities. However, during transportation and warehousing, cardboard boxes may face problems such as crushing damage, humidity changes, tearing, etc. The traditional method of manually inspecting the quality of cardboard boxes has drawbacks such as low efficiency, large errors, and high labor intensity, and it is difficult to meet the requirements of modern production and quality control. In the prior art, the surface image of a corrugated cardboard box can be segmented and recognized through an adaptive double-threshold segmentation algorithm. However, due to the complex gray-scale distribution of the surface image of the corrugated cardboard box, it will interfere with the segmentation. Therefore, Gaussian filtering can be used to filter the image.

[0003] The patent document with the publication number CN105096276B discloses an image Gaussian filtering method and device. Among them, the method includes: traversing the pixels included in the image to be filtered in a preset order, and obtaining the connected gray-scale matrix of the currently traversed pixel; determining, from the pixel units included in the connected gray-scale matrix, the first pixel units for which the convolution results have been cached in the buffer and the second pixel units for which the convolution results have not been cached, where the pixel unit is a pixel row or a pixel column in the connected gray-scale matrix; performing Gaussian filtering on the current pixel according to the determined first pixel units and second pixel units. Caching the convolution results of the pixel units in the buffer occupies very little memory resources, and since there are overlapping pixel units in the connected gray-scale matrices of adjacent pixels, the cache hit rate is very high, and the convolution results of the already cached pixel units can be directly used to perform Gaussian filtering on the current pixel, reducing the spatial complexity and time complexity of Gaussian filtering.

[0004] However, the above patent document does not solve the problem that during the process of filtering the surface image of a corrugated cardboard box through Gaussian filtering, since conventional Gaussian filtering directly performs weighted filtering on the pixel points in the image according to the distance between pixel points, without specifically considering improving the accuracy of detecting the quality of corrugated lines, resulting in a small image enhancement effect, reducing the accuracy of subsequent quality detection of corrugated cardboard boxes, or performing useless image enhancement operations, thereby increasing the time and resource consumption of data processing. Summary of the Invention

[0005] In order to solve the problem that due to conventional Gaussian filtering directly performing weighted filtering on the pixel points in the image according to the distance between pixel points, resulting in a small image enhancement effect or performing useless image enhancement operations, the present invention proposes a corrugated cardboard box quality detection system based on machine vision.

[0006] The system includes the following modules: an image acquisition module that acquires a surface image of a corrugated cardboard box, where the surface image is a grayscale image; a noise processing module that includes: a parameter unit that obtains an initial standard deviation according to the grayscale histogram of the grayscale image and constructs a Gaussian filter kernel based on the initial standard deviation and a preset filter kernel size; an analysis unit that designates any pixel point in the grayscale image as a target pixel point, designates the remaining pixel points within the Gaussian filter kernel with the target pixel point as the center of the Gaussian filter kernel as neighborhood pixel points; calculates a standard deviation adjustment coefficient of the target pixel point based on the grayscale deviation degree of the target pixel point and the grayscale difference between the target pixel point and all neighborhood pixel points; designates any neighborhood pixel point as a target neighborhood pixel point and calculates a Gaussian weight adjustment coefficient of the target neighborhood pixel point based on the grayscale deviation degree of the target neighborhood pixel point, the deviation of the standard deviation adjustment coefficient relative to the average standard deviation adjustment coefficient of all pixel points, and the deviation of the grayscale value of the target neighborhood pixel point relative to the average grayscale value of all neighborhood pixel points; a processing unit that adaptively filters the grayscale image using Gaussian filtering according to the Gaussian weight adjustment coefficient to obtain a processed image; and a quality detection module that uses an adaptive double-threshold segmentation algorithm to identify the corrugation defect area in the processed image to achieve the quality detection of the corrugated cardboard box based on machine vision.

[0007] Based on the grayscale difference and the standard deviation adjustment coefficient of the target pixel point, more accurate noise removal is achieved, improving the image processing effect; through the adaptive Gaussian filtering method, noise is effectively removed and image details are retained, avoiding the situation where the image enhancement effect is too small or useless in traditional filtering methods; the adaptive filtering technology improves the image quality, ensures more accurate subsequent defect recognition, and reduces missed detections and false detections; the system realizes automated detection, reduces manual intervention, provides real-time feedback of results, and improves the detection efficiency and product quality consistency of the production line.

[0008] Further, the image acquisition module further includes: a pre-segmentation unit that segments the surface image of the corrugated cardboard box in the acquired original image through semantic segmentation to obtain the surface image of the corrugated cardboard box.

[0009] Further, the initial standard deviation is obtained by: using the 3-sigma rule according to the grayscale histogram of the grayscale image to obtain the initial standard deviation.

[0010] Further, the standard deviation adjustment coefficient satisfies the following relational expression:

[0011] ; where is the standard deviation adjustment coefficient of the th pixel point, is the grayscale deviation degree of the th pixel point, is the The gray value of a pixel is the average gray value of all pixels in the grayscale image is the standard deviation of Gaussian filtering is the number of neighboring pixels of the th pixel The th pixel and the th neighboring pixel is the Euclidean distance The th neighboring pixel of the th pixel is the gray value

[0012] Dynamically adjust the standard deviation according to the gray difference of each pixel and the spatial position of neighboring pixels to ensure more delicate noise processing in different regions of the image and avoid over-smoothing or information loss; by reasonably adjusting the influence weight of neighboring pixels, especially giving greater weight to pixels with large gray differences in the neighborhood, it helps to retain details in the image and reduce the blurring effect

[0013] Furthermore, the Gaussian weight adjustment coefficient satisfies the following relational expression

[0014] ; where is the Gaussian weight adjustment coefficient of the th neighboring pixel of the th pixel is the gray deviation degree of the th neighboring pixel of the th pixel is the gray value of the th neighboring pixel of the th pixel is the average gray value of all pixels in the grayscale image is the initial standard deviation of Gaussian filtering is the standard deviation adjustment coefficient of the th pixel is the average standard deviation adjustment coefficient of all pixels in the grayscale image is the range of the standard deviation adjustment coefficients of all pixels in the grayscale image is the average gray value of all neighboring pixels of the th pixel is the th pixel and the range of gray values of all neighboring pixels

[0015] By dynamically adjusting according to the gray-scale characteristics, standard deviation, and gray-scale range of each pixel point and its neighboring pixel points, the filter can adapt to the image features of different regions, such as details, edges, noise, etc., improving the accuracy of image processing; by adjusting the influence intensity of the Gaussian filter kernel in each pixel region, noise can be effectively suppressed, especially in the high-noise region of the image, while avoiding over-smoothing in the detail region and maintaining important image information.

[0016] Further, the adaptive filtering of the grayscale image to obtain the processed image includes: performing adaptive filtering on each pixel point in the grayscale image, where the adaptive filtering is to weight the Gaussian weights of the target neighboring pixel points with the Gaussian weight adjustment coefficient to obtain the adjusted Gaussian weights of the target neighboring pixel points, taking the ratio of the adjusted Gaussian weights of the target neighboring pixel points to the sum of the adjusted Gaussian weights of all neighboring pixel points as the actual weights of the target neighboring pixel points, and filtering the target pixel point based on the actual weights of all neighboring pixel points to obtain the filtered value of the target pixel point; obtaining the filtered values of all pixel points, and further obtaining the processed image.

[0017] By dynamically adjusting the Gaussian weights, details and edges in the image are effectively retained, especially in regions with high noise, avoiding the loss of details; the filtering process is adjusted according to the neighborhood characteristics of each pixel point, can adapt to changes in different regions, and provides a more accurate processing effect; the adaptive filtering can remove noise while maintaining the important structure and details of the image, avoiding the blurring problem caused by traditional filtering.

[0018] Further, the adaptive dual-threshold segmentation algorithm adopts an adaptive dual-threshold segmentation algorithm based on a Gaussian mixture model.

[0019] Further, the implementation of the corrugated cardboard box quality inspection based on machine vision includes: in response to the ratio of the number of all pixel points in the corrugated defect region to the number of all pixel points in the processed image being greater than a preset defect threshold, determining that the quality inspection of the corrugated cardboard box is unqualified and issuing a warning prompt to complete the quality inspection of the corrugated cardboard box based on machine vision.

[0020] No manual intervention is required, reducing human errors and improving efficiency and accuracy; accurately judging the defect region through the ratio of the number of pixel points to ensure reliable detection results; when the defect exceeds the set threshold, the system immediately issues a warning for quick processing; the threshold can be adjusted according to requirements to adapt to different production and quality requirements; suitable for large-scale production, improving production efficiency and consistency.

[0021] The present invention has the following technical effects:

[0022] The standard deviation adjustment coefficient is calculated by considering the gray deviation degree of the target pixel point and the gray difference from the neighboring pixel points. At the same time, the Gaussian weight adjustment coefficient of the target neighboring pixel points is calculated based on multiple parameters, which changes the conventional Gaussian filtering method that simply weights according to the pixel point distance. It can filter the image more accurately, reduce useless image enhancement operations, and the enhancement effect is more significant and targeted. After adaptively filtering the grayscale image, the threshold segmentation is used to identify the corrugated defect area, which improves the image quality, enables the subsequent quality detection module to more accurately identify the defects of the corrugated cardboard box, improves the accuracy and efficiency of the quality detection of the corrugated cardboard box, helps to detect quality problems in time, reduces the outflow of defective products, improves the overall quality level of the product, and better meets the needs of modern production and quality control. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] By referring to the following detailed description with reference to the accompanying drawings, the above and other objects, features and advantages of the exemplary embodiments of the present invention will become readily understood. In the drawings, several embodiments of the present invention are shown by way of example and not limitation, and the same or corresponding reference numerals are for the same or corresponding parts, wherein:

[0024] Figure 1 FIG. is a system block diagram of a corrugated cardboard box quality detection system based on machine vision according to an embodiment of the present invention.

[0025] Figure 2 FIG. is a grayscale image of a corrugated cardboard box in a corrugated cardboard box quality detection system based on machine vision according to an embodiment of the present invention.

[0026] Figure 3 FIG. is a gradient image of a corrugated cardboard box in a corrugated cardboard box quality detection system based on machine vision according to an embodiment of the present invention.

[0027] Figure 4 FIG. is a grayscale histogram of a grayscale image of a corrugated cardboard box in a corrugated cardboard box quality detection system based on machine vision according to an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0028] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0029] It should be understood that when terms such as "first" and "second" are used in the claims, the specification and the drawings of the present invention, they are only used to distinguish different objects, rather than to describe a specific order. The terms "comprising" and "including" used in the specification and claims of the present invention indicate the presence of the described features, wholes, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components and / or their combinations.

[0030] The present invention provides a corrugated cardboard box quality detection system based on machine vision. As Figure 1 shown, the corrugated cardboard box quality detection system based on machine vision includes an image acquisition module 110, a noise processing module 120 and a quality detection module 130. Among them, the noise processing module 120 includes a parameter unit 121, an analysis unit 122 and a processing unit 123, which will be specifically described below.

[0031] The image acquisition module 110 acquires the surface image of the corrugated cardboard box, performs grayscale processing on the surface image to obtain a grayscale image, see Figure 2 , and then stores the grayscale image in the cloud.

[0032] Specifically, the image acquisition module further includes:

[0033] A pre-segmentation unit that segments the surface image of the corrugated cardboard box in the original image collected by the camera through semantic segmentation to obtain the surface image of the corrugated cardboard box.

[0034] It should be noted that since the grayscale values of the pixel points in the grayscale image are relatively complex, it will affect the detection of the quality of the corrugated cardboard box. Therefore, it is necessary to perform filtering processing on the grayscale image. However, the conventional Gaussian filtering only performs weighted average filtering based on the distance between pixel points in the image, and cannot reduce the influence of complex grayscale values on the quality detection of the corrugated cardboard box, so it cannot perform targeted filtering. Therefore, it is processed through the noise processing module 120.

[0035] The noise processing module 120 reads and processes the grayscale image in the cloud, and then sends the processed image to the quality detection module, including:

[0036] It should be noted that for the grayscale image of the corrugated cardboard box, its gradient image is as Figure 3 shown. It can be seen from Figure 3 that the gradient changes erratically and varies greatly, indicating that the change of grayscale values in the grayscale image is relatively complex, which will affect the subsequent threshold segmentation. Therefore, it is necessary to perform filtering processing on the grayscale image to eliminate the influence on the subsequent quality detection due to poor image quality.

[0037] The parameter unit 121 obtains an initial standard deviation according to the gray-level histogram of the gray-scale image, and constructs a Gaussian filter kernel based on the initial standard deviation and a preset filter kernel size.

[0038] Implementers can set the filter kernel size according to the specific implementation situation. For example, .

[0039] Specifically, the way to obtain the initial standard deviation is as follows:

[0040] Obtain the gray-level histogram of the gray-scale image according to the gray-scale image. Refer to Figure 4 , from Figure 4 it can be seen that the gray-level histogram also satisfies an approximately Gaussian distribution state. Therefore, based on the gray-level histogram of the gray-scale image, using the 3-sigma rule, the initial standard deviation is obtained.

[0041] It should be noted that for the pixel points in the gray-scale image, the greater the difference between the gray value and the average gray value, the greater the deviation degree of such pixel points from the normal gray value, and such pixel points are more likely to affect the quality inspection. Then, a larger Gaussian filter standard deviation is required to increase the filtering intensity of such pixel points; similarly, for the neighboring pixel points of each pixel point (the remaining pixel points within the Gaussian filter kernel), the greater the deviation degree from the normal gray value, the more it indicates that it is an abnormal pixel point. When taking it as a neighboring pixel point, the corresponding weight should be appropriately reduced to prevent the pixel point from being readjusted to an abnormal gray value pixel point during the filtering process according to the gray value of the relatively abnormal neighboring pixel points.

[0042] The analysis unit 122 designates any pixel point in the gray-scale image as the target pixel point, designates the remaining pixel points within the Gaussian filter kernel as neighboring pixel points with the target pixel point as the center of the Gaussian filter kernel. When the target pixel point is located at the image edge, the target pixel point only needs to be within the Gaussian filter kernel.

[0043] Based on the gray deviation degree of the target pixel point and the gray difference between the target pixel point and all neighboring pixel points, calculate the standard deviation adjustment coefficient of the target pixel point.

[0044] Specifically, the standard deviation adjustment coefficient satisfies the following relational expression:

[0045] ;

[0046] In the formula, is the standard deviation adjustment coefficient of the th pixel point, is the gray deviation degree of the th pixel point, is the gray value of the th pixel point, is the average gray value of all pixel points in the grayscale image, is the standard deviation of Gaussian filtering, is the number of neighboring pixel points of the th pixel point, is the Euclidean distance between the th pixel point and the th neighboring pixel point, is the gray value of the th neighboring pixel point of the th pixel point,

[0047] Among them, represents the degree of gray deviation of the th pixel point, which is obtained by calculating the square of the difference between the gray value of this pixel point and the average gray value of all pixel points. The larger the difference, the more likely the gray value of this pixel point is an abnormal gray value. The square of this difference is compared with the square of the standard deviation, and the difference is normalized by using the standard deviation as a standard; represents the gray difference between the th pixel point and all neighboring pixel points, which is weighted by the distance between pixel points, so that the influence of the gray difference of neighboring pixel points farther away from the th pixel point is smaller.

[0048] It should be noted that for any pixel point, during the Gaussian filtering process, it is necessary to perform weighted averaging according to the gray values of neighboring pixel points to obtain the gray value of the filtered pixel point. However, during the weighted averaging process, if there are also neighboring pixel points with a large deviation from the average gray value among all neighboring pixel points of this pixel point, such neighboring pixel points will affect the filtering of this pixel point, making the filtered gray value may still deviate. Therefore, the gray values of these neighboring pixel points should be adjusted to reduce the influence of neighboring pixel points with a large deviation on the filtering.

[0049] Denote any neighboring pixel point as the target neighboring pixel point, and calculate the Gaussian weight adjustment coefficient of the target neighboring pixel point based on the gray deviation degree of the target neighboring pixel point, the normalized deviation of the standard deviation adjustment coefficient with respect to the average standard deviation adjustment coefficient of all pixel points, and the normalized deviation of the gray value of the target neighboring pixel point with respect to the average gray value of all neighboring pixel points.

[0050] Specifically, the Gaussian weight adjustment coefficient satisfies the following relational expression:

[0051] ;

[0052] In the formula, is the The Gaussian weight adjustment coefficient of the neighboring pixel points of the th pixel point, is the degree of gray deviation of the neighboring pixel points of the th pixel point, is the gray value of the neighboring pixel points of the th pixel point, is the average gray value of all pixel points in the grayscale image, is the initial standard deviation of Gaussian filtering, is the standard deviation adjustment coefficient of the th pixel point, is the range of the standard deviation adjustment coefficients of all pixel points in the grayscale image, is the average gray value of all neighboring pixel points of the th pixel point, is the range of the gray values of all neighboring pixel points of the th pixel point.

[0053] Among them, represents the degree of gray deviation of the neighboring pixel points of the th pixel point, reflecting the deviation degree of the gray value of the th neighboring point from the average gray value. The greater the deviation of the gray value, the larger this value is, and the greater the impact on the Gaussian weight adjustment coefficient; represents the normalized deviation of the standard deviation adjustment coefficient of the th pixel point relative to the average standard deviation adjustment coefficient of all pixel points, which is the product of the normalized deviation of the gray value of the neighboring pixel points relative to the average gray value of all neighboring pixel points. By subtracting the average value and then dividing by the range of the data items, the standard deviation adjustment coefficient and the gray value are adjusted to the range, and then the two are multiplied. When the above two deviations are both relatively low or both relatively high relative to their average values, the Gaussian weight adjustment coefficient is higher; for the th pixel point in the grayscale image, the higher its standard deviation adjustment coefficient indicates that the gray value of this pixel point requires a greater degree of filtering, so higher weights need to be set for the neighboring pixel points with gray values that differ greatly from its own; conversely, if the standard deviation adjustment coefficient of this pixel point is lower, then higher weights need to be set for the neighboring pixel points with gray values that differ less from its own. The above effect on the Gaussian weight adjustment coefficient is achieved through the method of multiplying after standardization.

[0054] The processing unit 123 adaptively filters the grayscale image using Gaussian filtering according to the Gaussian weight adjustment coefficient to obtain a processed image.

[0055] It should be noted that for any pixel point in the grayscale image, the Gaussian filter kernel at the position of this pixel point can be preliminarily determined according to its Gaussian filtering standard deviation. Then, the weights of the neighboring pixel points are adjusted according to the Gaussian weight adjustment coefficients of each neighboring pixel point, and thus the pixel point can be filtered more pertinently.

[0056] Specifically, the step of adaptively filtering the grayscale image to obtain a processed image includes:

[0057] Performing adaptive filtering on each pixel point in the grayscale image. The adaptive filtering is to weight the Gaussian weights of the target neighboring pixel points with the Gaussian weight adjustment coefficient as the weight to obtain the adjusted Gaussian weights of the target neighboring pixel points. The ratio of the adjusted Gaussian weight of the target neighboring pixel point to the sum of the adjusted Gaussian weights of all neighboring pixel points is used as the actual weight of the target neighboring pixel point. Filtering the target pixel point according to the actual weights of all neighboring pixel points to obtain the filtered value of the target pixel point; obtaining the filtered values of all pixel points, and further obtaining the processed image.

[0058] The quality detection module 130 uses an adaptive double-threshold segmentation algorithm to identify the corrugated defect area in the processed image. The corrugated defect area is the area where the pixel points are located between the two thresholds in the segmentation result, so as to realize the quality detection of corrugated cardboard boxes based on machine vision.

[0059] Specifically, the adaptive double-threshold segmentation algorithm uses an adaptive double-threshold segmentation algorithm based on a Gaussian mixture model.

[0060] Specifically, the step of realizing the quality detection of corrugated cardboard boxes based on machine vision includes:

[0061] In response to the ratio of the number of all pixel points in the corrugated defect area to the number of all pixel points in the processed image being greater than a preset defect threshold, it is determined that the quality detection of the corrugated cardboard box is unqualified, and a warning prompt is issued to complete the quality detection of the corrugated cardboard box based on machine vision.

[0062] The implementer can set the defect threshold according to the actual implementation situation. For example, 5%.

[0063] Although this specification has shown and described multiple embodiments of the present invention, it is obvious to those skilled in the art that such embodiments are provided only by way of example. Those skilled in the art will think of many changes, alterations, and alternative ways without departing from the spirit and concept of the present invention. It should be understood that various alternative solutions of the embodiments of the present invention described herein can be adopted in the process of practicing the present invention.

[0064] The above are all preferred embodiments of the present invention, and do not limit the protection scope of the present invention accordingly. Therefore, all equivalent changes made according to the structure, shape and principle of the present invention shall be covered within the protection scope of the present invention.

Claims

1. A corrugated cardboard box quality inspection system based on machine vision, characterized in that, It includes the following modules: An image acquisition module that acquires the surface image of the corrugated cardboard box, and the surface image is a grayscale image; A noise processing module, including: a parameter unit that obtains an initial standard deviation according to the grayscale histogram of the grayscale image, and constructs a Gaussian filter kernel based on the initial standard deviation and a preset filter kernel size; an analysis unit that records any pixel point in the grayscale image as a target pixel point, and takes the target pixel point as the center of the Gaussian filter kernel and records the remaining pixel points in the Gaussian filter kernel as neighborhood pixel points; Calculate the standard deviation adjustment coefficient of the target pixel based on the gray deviation degree of the target pixel and the gray difference between the target pixel and all neighboring pixels. The formula is as follows: , is the standard deviation adjustment coefficient of the th pixel, is the gray deviation degree of the th pixel, is the gray value of the th pixel, is the average gray value of all pixels in the grayscale image, is the initial standard deviation of Gaussian filtering, is the number of neighboring pixels of the th pixel, is the th pixel and the th neighboring pixel Euclidean distance, is the th pixel of the th neighboring pixel gray value, is the absolute value symbol; Denote any neighborhood pixel as the target neighborhood pixel; calculate the Gaussian weight adjustment coefficient of the target neighborhood pixel based on the gray deviation degree of the target neighborhood pixel, the deviation of the standard deviation adjustment coefficient of the target neighborhood pixel relative to the average standard deviation adjustment coefficient of all pixels, and the deviation of the gray value of the target neighborhood pixel relative to the average gray value of all neighborhood pixels. The formula is: , is the Gaussian weight adjustment coefficient of the th neighborhood pixel of the th pixel point, is the gray deviation degree of the th neighborhood pixel of the th pixel point, , are the range and average standard deviation adjustment coefficient of the standard deviation adjustment coefficients of all pixel points in the gray image respectively, , are the range and average gray value of the gray values of all neighborhood pixels of the th pixel point respectively; A processing unit that adaptively filters the grayscale image using Gaussian filtering according to the Gaussian weight adjustment coefficient to obtain a processed image; A quality detection module that uses an adaptive double-threshold segmentation algorithm to identify the corrugation defect area in the processed image to achieve the quality detection of the corrugated cardboard box based on machine vision.

2. The corrugated cardboard box quality inspection system based on machine vision according to claim 1, characterized in that, The image acquisition module further includes: A pre-segmentation unit that segments the surface image of the corrugated cardboard box in the acquired original image through semantic segmentation to obtain the surface image of the corrugated cardboard box.

3. The corrugated cardboard box quality inspection system based on machine vision according to claim 1, characterized in that, The way to obtain the initial standard deviation is: According to the grayscale histogram of the grayscale image, the initial standard deviation is obtained using the 3σ criterion.

4. The corrugated carton quality inspection system based on machine vision according to claim 1, characterized in that, The adaptive filtering of the grayscale image to obtain a processed image includes: Performing adaptive filtering on each pixel point in the grayscale image. The adaptive filtering is to weight the Gaussian weights of the target neighborhood pixel points with the Gaussian weight adjustment coefficient as the weight to obtain the adjusted Gaussian weights of the target neighborhood pixel points, and take the ratio of the adjusted Gaussian weights of the target neighborhood pixel points to the sum of the adjusted Gaussian weights of all neighborhood pixel points as the actual weights of the target neighborhood pixel points, and filter the target pixel point according to the actual weights of all neighborhood pixel points to obtain the filtered value of the target pixel point; obtaining the filtered values of all pixel points, and further obtaining the processed image.

5. The corrugated carton quality detection system based on machine vision according to claim 1, wherein The adaptive double-threshold segmentation algorithm uses an adaptive double-threshold segmentation algorithm based on the Gaussian mixture model.

6. The corrugated carton quality inspection system based on machine vision according to claim 1, wherein, The implementation of the quality detection of the corrugated cardboard box based on machine vision includes: In response to the ratio of the number of all pixel points in the corrugation defect area to the number of all pixel points in the processed image being greater than a preset defect threshold, it is determined that the quality detection of the corrugated cardboard box is unqualified, and a warning prompt is issued to complete the quality detection of the corrugated cardboard box based on machine vision.

Images