A detection method for an electrical automation feeding device

Abstract

The present application relates to the technical field of image data processing, and particularly relates to a detection method for an electrical automation feeding device. The method includes: dividing a conveyor belt into multiple initial conveyor belt image blocks, merging them according to the gray-scale distribution of adjacent pixel points between the initial conveyor belt image blocks, further obtaining corresponding contrast feature values according to the gray-scale gradient distribution feature and gray-scale value distribution feature of the crack defect area in the surface defect merged image block, obtaining a contrast weight according to the change of the gray-scale value difference of the crack defect in the surface defect merged image block under different illumination intensity levels, obtaining the crack feature suppression degree representing the insignificance of the crack area according to the contrast feature value and the contrast weight, and improving the edge detection algorithm according to the crack feature suppression degree to complete the detection of the conveyor belt defect. The present application has higher accuracy in detecting surface crack defects of the conveyor belt.

Description

Technical Field

[0001] This application relates to the technical field of image data processing, and specifically relates to a detection method for an electrical automation feeding device. Background Art

[0002] An electrical automation feeding device uses a conveyor belt as the core conveying component, and selects appropriate connection components and control components according to material characteristics and production requirements. Feeding devices capable of automated operations, such as sensors and robotic arms. Commonly used electrical automation feeding devices include servo roller feeders, buffer feeders, feeders integrated with stirring and screening functions, etc. Different types of electrical automation feeding devices can transport different materials, such as packages, plastic products, hardware parts, etc.

[0003] During the operation of the automated feeding device, the high-intensity operation makes the safety detection of components extremely important. Especially the damage to the conveyor belt during the material transmission process. The conveyor belt has the characteristics of easy installation, large carrying capacity, high speed, and stability, and is widely used in the field of material transmission. However, during the high-intensity operation of the conveyor belt, cracks may appear on its surface. If the cracks are not processed in time, the conveyor belt may be torn, affecting the stability of the conveyor belt transmission, and even potential safety hazards may occur. The area corresponding to the crack defect is small and not easily observable by the naked eye. If the detection is stopped, it will affect the overall transportation equipment. Therefore, it is necessary to use computer vision to perform real-time detection without stopping the machine.

[0004] However, since there may also be dirty areas on the surface of the conveyor belt, and the dirty areas do not affect the operation of the conveyor belt, it is necessary to distinguish between the dirty areas and crack defects when detecting the cracks on the conveyor belt. The existing technology usually distinguishes them according to the size of the gray value in the collected surface image, and uses the characteristic that the gray value inside the dirty area changes randomly to distinguish between the dirty area and the crack defect. However, considering the working environment of the conveyor belt, the gray values of the dirty patches and crack areas on the conveyor belt may both be evenly distributed, resulting in a low accuracy in detecting the crack defects on the surface of the conveyor belt. Summary of the Invention

[0005] In order to solve the technical problem of the low accuracy of detecting crack defects on the surface of the conveyor belt in the existing technology, the purpose of this application is to provide a detection method for an electrical automation feeding device, and the specific technical solution adopted is as follows:

[0006] This application proposes a detection method for an electrical automation feeding device, and the method includes:

[0007] Obtain the surface images of the conveyor belt at each light intensity level, and divide the surface images of the conveyor belt into a preset number of initial conveyor belt image blocks;

[0008] At each light intensity level, according to the gray value distribution of the adjacent pixel points between adjacent initial conveyor belt image blocks, merge all the initial conveyor belt image blocks to obtain at least one surface defect merged image block; according to the change of the gray value of the pixel points in each surface defect merged image block along the running direction of the conveyor belt, and the overall gray value difference distribution, obtain the contrast feature value of each surface defect merged image block at each light intensity level;

[0009] According to the distribution of the change of the gray value difference of the pixel points in the surface defect merged image block at different light intensity levels, obtain the contrast weight of each surface defect merged image block at each light intensity level; according to the contrast feature value and the contrast weight, obtain the crack feature suppression degree of each surface defect merged image block;

[0010] Perform edge detection on each surface defect merged image block through the crack feature suppression degree, obtain the improved edge detection result of each surface defect merged image block, and perform conveyor belt defect detection according to the improved edge detection result.

[0011] Further, the merging of all the initial conveyor belt image blocks to obtain at least one surface defect merged image block includes:

[0012] For any two adjacent initial conveyor belt image blocks:

[0013] Take one of the initial conveyor belt image blocks as the first initial conveyor belt image block, and take the other initial conveyor belt image block as the second initial conveyor belt image block; take the pixel points adjacent to the block boundary in the first initial conveyor belt image block as the first boundary pixel points; take the pixel points adjacent to the block boundary in the second initial conveyor belt image block as the second boundary pixel points; match the first boundary pixel points with the second boundary pixel points to obtain at least two matching pairs, and the first boundary pixel point and the second boundary pixel point in the matching pair are adjacent; take the matching pair in which both the first boundary pixel point and the second boundary pixel point are abnormal pixel points as the abnormal matching pair; count the number of abnormal matching pairs corresponding to the first initial conveyor belt image block and the second initial conveyor belt image block;

[0014] When the block boundary is parallel to the running direction of the conveyor belt, merge the first initial conveyor belt image block and the second initial conveyor belt image block with the number of abnormal matching pairs greater than the preset first threshold to obtain the corresponding surface defect merged image block;

[0015] When the block boundary is not parallel to the running direction of the conveyor belt, the first initial conveyor belt image block and the second initial conveyor belt image block with the number of abnormal matching binary groups greater than a preset second threshold are merged to obtain a corresponding surface defect merged image block.

[0016] Further, the method for obtaining the contrast feature value includes:

[0017] Select an arbitrary surface defect merged image block as the target surface defect merged image block at any illumination intensity level; use the initial conveyor belt image block included in the target surface defect merged image block as the reference initial conveyor belt image block; calculate the mean gray scale gradient of the pixel points in each reference initial conveyor belt image block along the running direction of the conveyor belt; calculate the variance of the gray scale values corresponding to all pixel points in each reference initial conveyor belt image block;

[0018] Multiply the maximized value of the variance of the gray scale values corresponding to each reference initial conveyor belt image block by the corresponding mean gray scale gradient to obtain the average gradient amplitude corresponding to each reference initial conveyor belt image block; take the mean of all the average gradient amplitudes of the target surface defect merged image block as the contrast feature value of the target surface defect merged image block.

[0019] Further, the method for obtaining the contrast weight includes:

[0020] In each surface defect merged image block at each illumination intensity level, take the mean gray scale value of all abnormal pixel points as the first gray scale mean, take the mean gray scale value of all other pixel points outside the abnormal pixel points as the second gray scale mean, and take the difference between the second gray scale mean and the first gray scale mean as the corresponding gray scale between-class difference;

[0021] Maximize the gray scale between-class differences corresponding to each surface defect merged image block at each illumination intensity level to obtain the contrast weight of each surface defect merged image block at each illumination intensity level.

[0022] Further, the method for obtaining the crack feature suppression degree includes:

[0023] Take the product of the contrast feature value and the contrast weight as the gray scale feature intensity of each surface defect merged image block at each illumination intensity level; take the sum of all the gray scale feature intensities corresponding to each abnormal merged image as the crack feature suppression degree of each surface defect merged image block.

[0024] Further, the method for obtaining the improved edge detection result includes:

[0025] During the process of the edge detection algorithm processing each surface defect merged image patch, the product of the original gray gradient value of each pixel point after non-maximum suppression and the normalized value of the negative correlation mapping value of the crack feature suppression degree of the surface defect merged image patch where it is located is used as the improved gray gradient value of each pixel point. According to the improved gray gradient value and the corresponding gradient direction, through double-threshold detection, the improved edge detection result corresponding to each surface defect merged image patch is obtained.

[0026] Further, the edge detection algorithm is the canny edge detection algorithm.

[0027] Further, the conveyor belt defect detection based on the improved edge detection result includes:

[0028] When continuous edges appear in the improved edge detection result, the area corresponding to the continuous edges is used as the crack defect area on the conveyor belt; when no continuous edges appear in the improved edge detection result, there are no crack defects on the corresponding conveyor belt.

[0029] Further, the method for obtaining abnormal pixel points includes:

[0030] For the gray values of pixel points in the conveyor belt surface image under each light intensity level, the corresponding segmentation threshold is obtained by the maximum inter-class variance method; the pixel points with gray values less than the corresponding segmentation threshold in the conveyor belt surface image under each light intensity level are used as abnormal pixel points.

[0031] Further, the status warning of the electrical automation feeding equipment based on the defect detection result includes:

[0032] When a crack defect area is detected, the alarm system issues a warning signal, and the coordinates of the center point of the crack defect area are used as the positioning result, and the positioning result and the warning signal are synchronously sent to the terminal device of the staff.

[0033] This application has the following beneficial effects:

[0034] Considering that the stress generated by the conveyor belt during movement is mainly along its movement direction, when cracks appear on the surface of the conveyor belt, the corresponding cracks usually expand in the movement direction, that is, the crack direction on the surface of the conveyor belt is usually consistent with the running direction of the conveyor belt, so the gray level gradient corresponding to the crack area along the running direction is small; and because the overall area of the crack area is small and the distribution is relatively regular, the overall influence of the crack area on the gray value of each initial conveyor belt image block is small; according to this characteristic, the embodiment of the present application combines the change of the gray value along the running direction of the conveyor belt and the distribution of the overall gray value difference to obtain the contrast feature value of each surface defect merged image block. By calculating the contrast feature value, the subsequent distinction between the cracks and dirt on the surface of the conveyor belt is more accurate, and the accuracy of detecting the cracks on the surface of the conveyor belt is improved. Considering that the gray values of the crack area and the dirt area are affected differently under different illumination intensity levels, the present application obtains the contrast weight of each surface defect merged image block according to the distribution of the change of the gray value difference of the pixel points in the surface defect merged image block under different illumination intensity levels. By combining the contrast weight with the characteristics of different performances of the dirt area and the crack area under different illumination intensities, the detection accuracy of the crack area on the surface of the conveyor belt is higher. Further, the edge detection algorithm is improved according to the crack feature suppression degree obtained by the contrast feature value and the contrast weight, and the defect detection of the cracks on the surface of the conveyor belt is completed, realizing the non-stop detection during the operation of the electrical automation feeding equipment. Description of the Drawings

[0035] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required to be used in the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments of the present application. For those of ordinary skill in the art, without creative work, other drawings can also be obtained based on these drawings.

[0036] Figure 1 It is a flowchart of a detection method for an electrical automation feeding equipment provided by an embodiment of the present application. Detailed Embodiments

[0037] In order to further elaborate on the technical means and effects adopted by the present application to achieve the predetermined invention purpose, the following will, in combination with the accompanying drawings and preferred embodiments, detail the specific embodiments, structures, features and their effects of a detection method for an electrical automation feeding equipment proposed according to the present application. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. In addition, the specific features, structures or characteristics in one or more embodiments can be combined in any suitable form.

[0038] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the technical field to which this application belongs.

[0039] The following specifically describes the specific solution of a detection method for an electrical automation feeding device provided by this application with reference to the accompanying drawings.

[0040] Please refer to Figure 1 , which shows a flowchart of a detection method for an electrical automation feeding device provided by an embodiment of this application. The method includes:

[0041] Step S1: Obtain the surface image of the conveyor belt at each light intensity level, and divide the surface image of the conveyor belt into initial conveyor belt image blocks with a preset number of divisions.

[0042] This application uses devices such as industrial cameras to collect the surface data of the conveyor belt in the feeding device in real time, processes and analyzes the surface data to distinguish the oil stain area and crack defects on the conveyor belt, extracts the crack defect area from the surface data, and sends the coordinate information of the crack defect area to the operator and issues a warning signal at the same time, realizing real-time detection of the electrical automation feeding device without stopping the machine.

[0043] Specifically, considering the influence of the working environment where the electrical automation feeding device is located, such as coal mine transportation and dust workshops, etc., which will cause changes in the illumination brightness of the environment around the conveyor belt and affect the detection results. Therefore, the embodiments of this application first obtain the surface image of the conveyor belt at each light intensity level. The specific process is as follows: During the operation of the conveyor belt, fix the camera and adjust the shooting angle, and set a fixed light source around the camera so that the camera can clearly capture the surface of the conveyor belt. And as the conveyor belt runs, the captured video contains the complete surface of the conveyor belt. Further, according to each video frame, through image stitching technology, a complete surface image of the conveyor belt is obtained. Subsequently, the illumination intensity of the light source is changed in turn to obtain the surface image of the conveyor belt at each light intensity level. It should be noted that the image stitching technology is a well-known prior art to those skilled in the art and will not be further described here.

[0044] It should be noted that in the embodiments of the present application, the light intensity levels are divided as follows: the light intensity is divided into three light intensity levels according to the lux value. Among them, the weak light intensity level corresponds to 500 lux to 1000 lux, the medium light intensity level corresponds to 1000 lux to 2000 lux, and the strong light intensity level corresponds to 2000 lux to 10000 lux. In the embodiments of the present application, surface images of the conveyor belt are respectively obtained under the strong light intensity level of 4000 lux, the medium light intensity level of 1500 lux, and the weak light intensity level of 800 lux. In an embodiment of the present application, an industrial high-speed camera is selected as the camera, and a fluorescent lamp with adjustable brightness is used as the light source.

[0045] It should be further noted that the implementer can select different numbers of light sources according to the implementation situation to increase the light intensity, but it is necessary to ensure that the light is evenly distributed on the surface image of the conveyor belt; and the implementer can determine the division method of the light intensity levels and the specific value selection of the light intensity levels according to the specific implementation environment by himself / herself, and the present application does not make special restrictions on this.

[0046] It should be noted that since the subsequent analysis process is based on the gray values of each pixel point in the surface image of the conveyor belt, the surface image of the conveyor belt in the embodiments of the present application is grayscale processed to obtain the corresponding grayscale image. And for the convenience of description, all the surface images of the conveyor belt in the subsequent process are grayscale images, and no further elaboration will be made hereinafter.

[0047] Considering that when there are defective areas in the surface image of the conveyor belt, the area of the corresponding defective areas is usually very small, which results in a large amount of unnecessary calculations when analyzing the surface image of the conveyor belt. Moreover, when analyzing the entire surface image of the conveyor belt, due to the small area of the defective areas, the corresponding features are not obvious enough, which may lead to insufficient detection of the defective areas, further reducing the accuracy of defect detection. To avoid this situation, in the embodiments of the present application, the surface image of the conveyor belt is divided into a preset number of initial conveyor belt image blocks. By dividing the surface image of the conveyor belt into image blocks, it is possible to accurately detect each small block, and according to the difference in the gray value distribution between the defective areas and the normal areas, avoid the analysis and calculation of a large number of normal image blocks, thereby reducing the amount of calculation. In the embodiments of the present application, the preset number of divisions is set to 100, the initial conveyor belt image blocks are of the same size and shape, and the boundaries of the initial conveyor belt image blocks are parallel or perpendicular to the running direction of the conveyor belt. Considering that the surface image of the conveyor belt is usually a rectangular area, the surface image of the conveyor belt in the embodiments of the present application is divided into 100 rectangular initial conveyor belt image blocks of the same size and shape. It should be noted that the implementer can adjust the size of the preset number of divisions and the size and shape of the initial conveyor belt image blocks according to the specific implementation environment, which will not be elaborated further here.

[0048] Step S2: At each light intensity level, according to the gray value distribution of the adjacent pixel points between adjacent initial conveyor belt image blocks, all the initial conveyor belt image blocks are merged to obtain at least one surface defect merged image block; according to the change of the gray value of the pixel points in each surface defect merged image block along the running direction of the conveyor belt, and the overall gray value difference distribution, the contrast feature value of each surface defect merged image block at each light intensity level is obtained.

[0049] Considering that when dividing the surface image of the conveyor belt in the embodiments of the present application, it does not rely on the characteristics of the surface image of the conveyor belt itself, but only divides the surface image of the conveyor belt into multiple initial conveyor belt image blocks. Therefore, when there are defective areas in the surface image of the conveyor belt, some defective areas may be divided into multiple adjacent initial conveyor belt image blocks. Therefore, in order to accurately analyze each complete defective area, it is necessary to merge the multiple adjacent initial conveyor belt image blocks divided from the defective area. In the embodiments of the present application, at each light intensity level, according to the gray value distribution of the adjacent pixel points between adjacent initial conveyor belt image blocks, all the initial conveyor belt image blocks are merged to obtain at least one surface defect merged image block.

[0050] Preferably, merging all the initial conveyor belt image blocks to obtain at least one surface defect merged image block includes:

[0051] For any two adjacent initial conveyor belt image blocks: take one of the initial conveyor belt image blocks as the first initial conveyor belt image block, and take the other initial conveyor belt image block as the second initial conveyor belt image block; take the pixel points adjacent to the block boundary in the first initial conveyor belt image block as the first boundary pixel points; take the pixel points adjacent to the block boundary in the second initial conveyor belt image block as the second boundary pixel points; match the first boundary pixel points with the second boundary pixel points to obtain at least two matching binary groups, where the first boundary pixel point and the second boundary pixel point in the matching binary group are adjacent. In the embodiments of the present application, the dividing line between the first initial conveyor belt image block and the second initial conveyor belt image block, i.e., the block boundary, is a virtual straight line, that is, this straight line is not composed of pixel points, but only divides the boundaries of the first initial conveyor belt image block and the second initial conveyor belt image block. Therefore, the first boundary pixel points and the second boundary pixel points are adjacent along the block boundary, that is, the first boundary pixel point and the second boundary pixel point in the matching binary group are adjacent. Since the initial conveyor belt image blocks in the embodiments of the present application are rectangles with the same shape and size, the block boundary is a straight line, and the line segment between the first boundary pixel point and the second boundary pixel point in each corresponding matching binary group is perpendicular to the straight line corresponding to the block boundary.

[0052] Take the matching binary groups in which both the first boundary pixel point and the second boundary pixel point are abnormal pixel points as abnormal matching binary groups; count the number of abnormal matching binary groups corresponding to the first initial conveyor belt image block and the second initial conveyor belt image block. Abnormal pixel points are the pixel points corresponding to the defective area. When the defective area is divided into two adjacent initial conveyor belt image blocks, the defective area will be truncated, and there should be a certain number of defective area pixel points on both sides of the corresponding truncated boundary. That is, when the defective area is divided into the first initial conveyor belt image block and the second initial conveyor belt image block, there should be a certain number of abnormal matching binary groups between the corresponding first initial conveyor belt image block and the second initial conveyor belt image block. However, considering that the image on the surface of the conveyor belt may be affected by other external factors such as noise, some of the abnormal matching binary groups may be irrelevant to the defective area. Therefore, it is necessary to further determine the number of abnormal matching binary groups to avoid the influence of external factors such as noise.

[0053] Preferably, the method for obtaining abnormal pixel points includes:

[0054] For the gray values of the pixel points in the surface image of the conveyor belt at each light intensity level, the corresponding segmentation threshold is obtained by the Otsu method; the pixel points in the surface image of the conveyor belt at each light intensity level with gray values less than the corresponding segmentation threshold are used as abnormal pixel points. Considering the crack defects on the conveyor belt to be detected in the embodiments of the present application, the gray values of the pixel points in the corresponding areas are always significantly less than those of the pixel points in other normal areas. Therefore, the defective areas with relatively small gray values can be clearly divided by threshold segmentation. However, considering that the gray values of the pixel points in some dirty areas are also always significantly less than those of the pixel points in other normal areas, the crack areas required in the embodiments of the present application cannot be screened out only by the Otsu method. Therefore, further analysis needs to be carried out on this basis.

[0055] When the block boundary is parallel to the running direction of the conveyor belt, the first initial conveyor belt image block and the second initial conveyor belt image block with the number of abnormal matching binary groups greater than the preset first threshold are merged to obtain the corresponding surface defect merged image block. In the embodiments of the present application, the preset first threshold is set to 5. Since the purpose of the embodiments of the present application is to detect defect areas in the crack areas, the initial conveyor belt image blocks corresponding to the crack areas should be merged as much as possible according to the characteristics of the cracks. When the boundary between the first initial conveyor belt image block and the second initial conveyor belt image block is parallel to the running direction of the conveyor belt, considering that the extending direction of the crack is also usually parallel to the running direction of the conveyor belt, when the crack area is divided into the first initial conveyor belt image block and the second initial conveyor belt image block, it is usually divided along the midline corresponding to the width of the crack area. Therefore, the number of corresponding abnormal matching binary groups is relatively large, and the value of the corresponding preset first threshold is relatively large.

[0056] When the block boundary is not parallel to the running direction of the conveyor belt, the first initial conveyor belt image block and the second initial conveyor belt image block with the number of abnormal matching binary groups greater than the preset second threshold are merged to obtain the corresponding surface defect merged image block. In the embodiments of the present application, the preset second threshold is set to 2. When the boundary between the first initial conveyor belt image block and the second initial conveyor belt image block is perpendicular to the running direction of the conveyor belt, if the crack area is divided into the first initial conveyor belt image block and the second initial conveyor belt image block, the crack area is usually truncated along the length direction, and the number of corresponding abnormal matching binary groups is usually less than or equal to the width of the crack area. Since the crack area is slender and its width is generally small, the number of corresponding abnormal matching binary groups is relatively small, and the value of the corresponding preset second threshold is relatively small.

[0057] It should be noted that when an initial conveyor belt image block simultaneously meets the merging conditions with multiple initial conveyor belt image blocks, the initial conveyor belt image block is merged with the corresponding multiple initial conveyor belt image blocks into a surface defect merged image block. That is, there may be more than two initial conveyor belt image blocks in a surface defect merged image block.

[0058] So far, the surface defect merged image blocks corresponding to each defect area are obtained. However, considering that there may be dirty areas and crack defects in the defect area, and the existence of the dirty area will not affect the safety and stability of the conveyor belt, so there is no need to detect the dirty area. While the existence of the crack defect will affect the stability of the conveyor belt and there are certain potential safety hazards. Therefore, in order to further complete the defect detection of the crack area, it is necessary to further analyze the surface defect merged image blocks.

[0059] Considering that the gray values of the pixel points corresponding to the crack defect area and the dirty area are smaller than those of the pixel points in the normal area, and the areas corresponding to the crack defect area and the dirty area are different, so the influences of the crack defect area and the dirty area on the overall gray value of the initial conveyor belt image block are different. And since the extension direction of the crack defect area is parallel to the running direction of the conveyor belt, while the shape corresponding to the dirty area is irregular, so the variation of the gray values of the pixel points in the crack defect area and the dirty area along the running direction of the conveyor belt is different. In the embodiment of the present application, according to the variation of the gray values of the pixel points in each surface defect merged image block along the running direction of the conveyor belt and the overall gray value difference distribution, the contrast feature value of each surface defect merged image block at each illumination intensity level is obtained.

[0060] Preferably, the method for obtaining the contrast feature value includes:

[0061] Select an arbitrary surface defect merged image block at any light intensity level as the target surface defect merged image block; use the initial conveyor belt image block contained in the target surface defect merged image block as the reference initial conveyor belt image block; calculate the average gray gradient value of the pixel points in each reference initial conveyor belt image block along the running direction of the conveyor belt. Considering that there are certain gray differences between the defect areas corresponding to cracks and dirt and the normal areas, there will usually be certain gray gradient features in each reference initial conveyor belt image block. In each reference initial conveyor belt image block, since the overall extension direction of the crack is always consistent with the conveyor belt direction and is relatively slender, along the running direction of the conveyor belt, the number of pixel points with gray differences is usually small, so the corresponding average gray gradient value is usually small; while for the dirt area, due to its irregular shape, the number of pixel points with gray differences along the running direction of the conveyor belt is usually much larger than that of the crack defect area, so the corresponding average gray gradient value is usually larger than that of the crack defect area. Therefore, the crack area can be better detected for defects by calculating the average gray gradient value.

[0062] Calculate the variance of the gray values corresponding to all pixel points in each reference initial conveyor belt image block. Since the gray values of the pixel points in the normal area are usually consistent, while the gray values of the pixel points corresponding to the defect area are relatively small compared to the normal area, there will be a certain variance in the gray values of all pixel points in each reference initial conveyor belt image block. The defect characteristics of the crack area are not obvious, and the corresponding defect area is usually small, that is, the number of pixel points with gray differences from the pixel points in the normal area is small, so the corresponding variance of the gray values is usually small. While for the dirt area, due to its irregular shape and usually relatively large corresponding area, that is, when there is a dirt area in the reference initial conveyor belt image block, the number of pixel points with gray differences from the pixel points in the normal area is large, so the corresponding variance of the gray values is larger than that of the crack defect area. Therefore, the crack area can be better detected for defects by calculating the variance of the gray values.

[0063] As can be seen so far, in the target surface defect merged image block, the greater the variance of the gray value in each reference initial conveyor belt image block, the greater the average gray gradient along the running direction of the conveyor belt, and the less significant the crack feature in the corresponding defect area. In the embodiments of the present application, the crack feature of the defect area is characterized by the contrast feature value, that is, the smaller the contrast feature value, the more significant the crack feature in the defect area of the corresponding surface defect merged image block. In the embodiments of the present application, the product of the maximized value of the variance of the gray value corresponding to each reference initial image block and the corresponding average gray gradient is used as the average gradient amplitude corresponding to each reference initial conveyor belt image block; the average value of all the average gradient amplitudes of the target surface defect merged image block is used as the contrast feature value of the target surface defect merged image block. It should be noted that the implementer can select other methods besides maximization to process the variance of the gray value according to the specific implementation environment, such as calculating the normalized value of the variance of the gray value, which will not be elaborated further here.

[0064] In the embodiments of the present application, the method for obtaining the contrast feature value of the target surface defect merged image block is expressed in formula as:

[0065]

[0066] Wherein, is the contrast feature value of the target surface defect merged image block, is the variance of the gray value corresponding to the th reference initial conveyor belt image block in the target surface defect merged image block, is the maximum value of the variances of the gray values corresponding to all the reference initial conveyor belt image blocks corresponding to the target surface defect merged image block, is the average gray gradient of the th reference initial conveyor belt image block in the target surface defect merged image block along the running direction of the conveyor belt, is the number of reference initial conveyor belt image blocks in the target surface defect merged image block, is the maximum value selection function, is the maximized value of the variance of the gray value corresponding to the th reference initial conveyor belt image block in the target surface defect merged image block.

[0067] It should be noted that since there are abnormal pixel points with gray values different from those of the normal area pixel points in each initial conveyor belt image block corresponding to the target surface defect merged image block, the maximum value of the corresponding variance of the gray value cannot be 0. Further, according to the method for obtaining the contrast feature value of the target surface defect merged image block, the contrast feature value of each surface defect merged image block under each illumination intensity level is obtained.

[0068] In addition, the implementer can also obtain the contrast feature value of the target surface defect merged image block through other forms of formulas. For example:

[0069]

[0070] Among them, is a normalization function, that is, is the normalized value of the variance of the gray value of the th reference initial conveyor belt image block in the target surface defect merged image block. The meanings of the remaining parameters are the same as those in the formula corresponding to the method for obtaining the contrast feature value of the target surface defect merged image block in the embodiments of the present application, and will not be further elaborated here.

[0071] It should be noted that in the embodiments of the present application, the normalization method adopts linear normalization. The implementer can adopt other normalization methods according to the specific implementation environment, and linear normalization is a well-known prior art to those skilled in the art, and will not be further elaborated here.

[0072] Step S3: Obtain the contrast weight of each surface defect merged image block at each illumination intensity level according to the distribution of the gray value differences of the pixel points in the surface defect merged image block changing under different illumination intensity levels; obtain the crack feature suppression degree of each surface defect merged image block according to the contrast feature value and the contrast weight.

[0073] Since under different illumination conditions, the gray features of the defect areas corresponding to the same surface defect merged image block are different, analyzing the conveyor belt surface image only at one illumination intensity level may cause misjudgment of defect detection. Therefore, the embodiments of the present application obtain the contrast feature value of each surface defect merged image block at each illumination intensity level. Considering that when the illumination intensity changes, the gray values of both the dirty area and the crack defect area will change, but there are certain differences in the change situations of the gray values of the dirty area and the crack defect area. The embodiments of the present application obtain the contrast weight of each surface defect merged image block at each illumination intensity level according to the distribution of the gray value differences of the pixel points in the surface defect merged image block changing under different illumination intensity levels. Through the contrast weight, further detection of the crack defects on the conveyor belt surface is carried out according to the characteristics that the gray value changes of the dirty area and the crack defect area are different under different illumination intensities.

[0074] Preferably, the method for obtaining the contrast weight includes:

[0075] In each merged image block of surface defects at each light intensity level, the average gray value of all abnormal pixel points is used as the first gray average value, the average gray value of all other pixel points outside the abnormal pixel points is used as the second gray average value, and the difference between the second gray average value and the first gray average value is used as the corresponding between-class gray difference. The between-class gray differences corresponding to each merged image block of surface defects at each light intensity level are maximized to obtain the contrast weight of each merged image block of surface defects at each light intensity level.

[0076] Since in the embodiments of the present application, the abnormal pixel points are the pixel points less than the corresponding segmentation threshold after being divided by the maximum between-class variance method, that is, the pixel points corresponding to the defect area, so all other pixel points outside the abnormal pixel points are normal pixel points. Therefore, the between-class gray variance is the difference in the overall gray values corresponding to the normal pixel points and the abnormal pixel points at each light intensity level.

[0077] As the light intensity increases, the gray values of the pixel points on the surface of the conveyor belt generally increase. However, since the crack defect area is different from the dirty area, the crack defect is a depression passing through the conveyor belt, and although the gray value of the corresponding pixel points also increases with the increase of the light intensity, compared with the dirty area and other normal pixel point areas, the speed of increase of its corresponding gray value is slower, resulting in a larger between-class gray difference corresponding to the crack defect area when the light intensity is greater. And the defect corresponding to the dirty area adheres to the surface of the conveyor belt. As the light intensity increases, the speed of increase of the gray values of the pixel points in the dirty area is similar to that of the pixel points in the normal area, resulting in little change in the between-class gray difference corresponding to the dirty area when the light intensity is greater.

[0078] On the other hand, since the change in the between-class gray difference corresponding to the dirty area is small, after maximizing the maximum between-class gray differences of the corresponding merged image blocks of surface defects at different light intensities, the contrast weights of the merged image blocks of surface defects corresponding to the dirty area at different light intensity levels are numerically large and similar. And the between-class gray difference of the merged image block corresponding to the crack defect area increases with the increase of the light intensity. Therefore, after maximizing the between-class gray differences of the corresponding merged image blocks of surface defects at different light intensities, the contrast weights of the merged image blocks of surface defects corresponding to the crack defect area at different light intensity levels are positively correlated with the light intensity level in terms of numerical value, and there are certain differences in the contrast weights at different light intensity levels. Therefore, the defect detection of the crack area in the conveyor belt can be further carried out by calculating the distribution difference of the contrast weights of each merged image block of surface defects at different light intensity levels.

[0079] In the embodiments of the present application, at the th light intensity level, the merged image block of surface defects The method for obtaining the contrast weight is expressed in formula as follows:

[0080]

[0081] Wherein, is the contrast weight of the surface defect merged image block at the -th light intensity level, ; is the second gray mean value of the surface defect merged image block at the -th light intensity level, ; is the first gray mean value of the surface defect merged image block at the -th light intensity level, ; is the second gray mean value of the surface defect merged image block ; is the first gray mean value of the surface defect merged image block ; is the maximum value selection function, is the gray class between-class difference of the surface defect merged image block at the -th light intensity level, ; is the maximum value of the gray class between-class differences corresponding to the surface defect merged image blocks at all light intensity levels .

[0082] It should be noted that since the overall contrast weights corresponding to the surface defect merged image blocks in the crack area are smaller than those in the dirt area, the larger the overall contrast weights of the corresponding surface defect merged image blocks, the less significant the corresponding crack defect features. Further, according to the method for obtaining the contrast weight of the surface defect merged image block at the -th light intensity level, the contrast weights of each surface defect merged image block at each light intensity level are obtained. It should be noted that since there must be abnormal pixel points and normal pixel points with different gray values in the surface defect merged image block, the maximum value of the corresponding gray class between-class difference must not be 0.

[0083] So far, the contrast weights and contrast feature values of each surface defect merged image block at each light intensity level, which characterize the significant features of the crack area, are obtained. Therefore, the crack feature suppression degree of each surface defect merged image block can be further obtained based on the contrast feature value and the contrast weight. The crack feature non-significance of the surface defect merged image block is characterized by the crack feature suppression degree, that is, the greater the crack feature suppression degree, the less significant the crack features of the corresponding surface defect merged image block.

[0084] Preferably, the method for obtaining the degree of crack feature suppression includes:

[0085] Multiply the contrast feature value by the contrast weight as the contrast feature intensity of each surface defect merged image block at each illumination intensity level; take the sum of all contrast feature intensities corresponding to each abnormal merged image as the degree of crack feature suppression of each surface defect merged image block. For each abnormal merged image block, since it corresponds to a contrast weight and a contrast feature value at each illumination intensity level, and the difference in contrast weight between the dirty area and the crack defect can only be analyzed as a whole. To better reflect the difference characteristics of the dirty area and the crack defect in the contrast weight, in the embodiment of the present application, after multiplying the contrast weight by the contrast feature value at each illumination intensity level, the corresponding product is accumulated with other products, that is, the sum of all contrast feature intensities corresponding to each abnormal merged image block is accumulated, so that the obtained degree of crack feature suppression can better combine the corresponding difference characteristics represented by the contrast weight and the contrast feature value. That is, the smaller the corresponding contrast feature value and the overall smaller the contrast weight, the smaller the corresponding degree of crack feature suppression, and the more significant the crack feature of the corresponding surface defect merged image block, that is, the more likely there is a crack defect.

[0086] In the embodiment of the present application, the surface defect merged image block The method for obtaining the degree of crack feature suppression is expressed in the formula as:

[0087]

[0088] Wherein, Is the degree of crack feature suppression of the surface defect merged image block ; Is the Th contrast weight of the surface defect merged image block At the th illumination intensity level; Is the Th contrast feature value of the surface defect merged image block At the th illumination intensity level; Is the number of preset illumination intensity levels. In the implementation of the present application, the number of illumination intensity levels is 3, that is, the weak illumination intensity level with an illumination intensity of 800 lux, the medium illumination intensity level with an illumination intensity of 1500 lux, and the strong illumination intensity level with an illumination intensity of 4000 lux.

[0089] In addition, the implementer can also obtain the degree of crack feature suppression of the surface defect merged image block Through other forms of formulas, but it is necessary to ensure that the surface defect merged image block The overall corresponding contrast feature values are positively correlated with the degree of crack feature suppression, and the overall contrast weights are positively correlated with the degree of crack feature suppression. For example:

[0090]

[0091] Among them, the meanings of the respective parameters are the same as the formula corresponding to the method for obtaining the degree of crack feature suppression of the surface defect merged image block in the embodiment of the present application and will not be further elaborated here.

[0092] Step S4: Perform edge detection on each surface defect merged image block through the degree of crack feature suppression to obtain an improved edge detection result for each surface defect merged image block, and perform conveyor belt defect detection based on the improved edge detection result.

[0093] So far, the degree of crack feature suppression of each surface defect merged image block in the conveyor belt surface image has been obtained, and the degree of crack feature suppression characterizes the insignificance of the crack area. Therefore, the edge texture features of each surface defect merged image block can be further enhanced or suppressed according to the degree of crack feature suppression, that is, the edge texture of the surface defect merged image block with strong crack feature significance is enhanced, and the edge texture of the surface defect merged image block with weak crack feature significance is weakened, so that the accuracy of detecting crack defects on the conveyor belt surface is higher. In the embodiment of the present application, edge detection is performed on each surface defect merged image block through the degree of crack feature suppression to obtain an improved edge detection result for each surface defect merged image block. Since there is an obvious difference between the gray value of the crack defect corresponding to the embodiment of the present application and the normal area, an edge detection algorithm is usually used for defect detection. Since there is also an obvious difference between the gray value of the dirty area and the normal area, the improved edge detection result is more accurate for detecting crack defects through the degree of crack feature suppression.

[0094] Preferably, the method for obtaining the improved edge detection result includes:

[0095] In the process of the Canny edge detection algorithm processing each surface defect merged image block, the product of the original gray gradient value of each pixel point after non-maximum suppression and the negative correlation mapping value of the crack feature suppression degree of the surface defect merged image block where it is located is used as the improved gray gradient value of each pixel point. According to the improved gray gradient value and the corresponding gradient direction, through double-threshold detection, the improved edge detection result corresponding to each surface defect merged image block is obtained. When the crack feature suppression degree is greater, the crack feature of the corresponding surface defect merged image block is less significant. Therefore, it is necessary to weaken the edge texture feature of the surface defect merged image block. On the contrary, when the crack feature suppression degree is smaller, the crack feature of the corresponding surface defect merged image block is more significant. Therefore, it is necessary to enhance the edge texture feature of the surface defect merged image block or weaken it less relative to the dirty area. Since the magnitude of the gray gradient value can represent the obviousness of the edge texture feature, the crack feature suppression degree of each surface defect merged image block is negatively correlated and mapped and then multiplied by the corresponding gray gradient value to obtain the improved gray gradient value.

[0096] It should be noted that the specific process of the Canny edge detection algorithm mainly includes Gaussian filtering, image gradient calculation, non-maximum suppression, and double-threshold detection. In the embodiments of the present application, only the gray gradient value of each pixel point after non-maximum suppression is improved, and the rest of the processes are well-known prior arts to those skilled in the art and will not be further elaborated here.

[0097] Furthermore, it should be noted that the illumination intensity level corresponding to the surface defect merged image block for edge detection in the embodiments of the present application is set to a medium illumination intensity level with an illumination intensity of 1500 lux. Since the magnitude of the illumination intensity level does not affect the position of abnormal pixel points in the surface defect merged image block, the implementer can select the surface defect merged image block under other illumination intensity levels for edge detection according to the specific implementation environment and will not be further elaborated here.

[0098] In the embodiments of the present application, the surface defect merged image block the pixel point The method for obtaining the corresponding improved gray gradient value is expressed in the formula as:

[0099]

[0100] Wherein, is the improved gray gradient value corresponding to the pixel point in the surface defect merged image block corresponding, is the original gray gradient value corresponding to the pixel point in the surface defect merged image block corresponding, For the surface defect merged image block The corresponding degree of crack feature suppression Is an exponential function with the natural constant e as the base Is a normalization function. In the embodiments of the present application, the normalization function adopts linear normalization, and linear normalization is a conventional technical means, which will not be further elaborated here. For the surface defect merged image block The negative correlation mapping value of the corresponding degree of crack feature suppression

[0101] Further, according to the surface defect merged image block The pixels in The corresponding method for obtaining the improved gray gradient value is used to obtain the improved gray gradient value of each pixel in each surface defect merged image block. Since the degree of crack feature suppression of the surface defect merged image block corresponding to the dirty area is greater than that of the crack defect, the gray gradient value of the pixels in the surface defect merged image block corresponding to the dirty area is weakened to a higher degree. Therefore, by obtaining the improved edge detection result, the interference of the dirty area on the crack defect detection will be significantly reduced.

[0102] Further, the defect detection can be completed according to the improved edge detection result. In the embodiments of the present application, the conveyor belt defect detection is performed according to the improved edge detection result.

[0103] Preferably, the conveyor belt defect detection according to the improved edge detection result includes:

[0104] When continuous edges appear in the improved edge detection result, the area corresponding to the continuous edges is used as the crack defect area on the conveyor belt; when no continuous edges appear in the improved edge detection result, there is no crack defect on the corresponding conveyor belt. Since the edge detection result is improved by the degree of crack feature suppression in the embodiments of the present application, the edge textures in the surface image of the conveyor belt are all weakened, and only the surface defect merged image block with a significantly higher crack feature is weakened to a smaller degree. Therefore, the continuous edges that appear in the improved edge detection result are usually the corresponding crack defect areas.

[0105] Further, when the crack defect area is detected, the alarm system issues a warning signal, and the coordinates of the center point of the crack defect area are used as the positioning result. The positioning result and the warning signal are synchronously sent to the terminal device of the staff, so that the staff can repair the crack defect in time, avoid unnecessary safety problems, and realize the non-stop detection during the operation of the electrical automation feeding equipment.

[0106] It should be noted that the implementer can set up an alarm system according to the specific implementation environment. The alarm system can be communicatively connected to a computer device, and the communication methods include but are not limited to network cable communication, Bluetooth communication, etc. The alarm system consists of a communication module, an acoustic-optical warning device such as a buzzer, a flash lamp, etc., a signal transmission module, and a power interface. Among them, the communication module is used to communicatively connect to the computer device, the acoustic-optical warning device is used to emit warning signals, the signal transmission module is used to transmit and receive signals with the terminal device of the staff; the power interface is used to connect to a power supply.

[0107] In summary, after dividing the conveyor belt into multiple initial conveyor belt image blocks, the present application merges them according to the gray-scale distribution of adjacent pixel points between the initial conveyor belt image blocks, so that the obtained surface defect merged image blocks can contain complete defect regions. Further, according to the gray-scale gradient distribution characteristics and gray-scale value distribution characteristics of the crack defect region in the surface defect merged image blocks, the corresponding contrast feature values are obtained. According to the change of the gray-scale value difference of the crack defect in the surface defect merged image blocks under different illumination intensity levels, the contrast weight is obtained. According to the contrast feature values and the contrast weight, the crack feature suppression degree representing the insignificance of the crack region is obtained. According to the crack feature suppression degree, the edge detection algorithm is improved to complete the detection of the conveyor belt defects. The present application has higher accuracy in detecting surface crack defects of the conveyor belt.

[0108] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements on some of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the various embodiments of the present application, and should all be included in the protection scope of the present application.

[0109] Each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other. The key points of each embodiment are to illustrate the differences from other embodiments.

Claims

1. A detection method for an electrical automation feeding device, characterized in that, The method includes: Obtaining the surface image of the conveyor belt at each light intensity level, and dividing the surface image of the conveyor belt into initial conveyor belt image blocks with a preset number of divisions; At each light intensity level, according to the gray value distribution of the adjacent pixel points between adjacent initial conveyor belt image blocks, merging the initial conveyor belt image blocks to obtain at least one surface defect merged image block; according to the change of the gray value of the pixel points in each surface defect merged image block along the running direction of the conveyor belt, and the overall gray value difference distribution, obtaining the contrast feature value of each surface defect merged image block at each light intensity level; According to the distribution of the gray value difference of the pixel points in the surface defect merged image block changing under different light intensity levels, obtaining the contrast weight of each surface defect merged image block at each light intensity level; according to the contrast feature value and the contrast weight, obtaining the crack feature suppression degree of each surface defect merged image block; Performing edge detection on each surface defect merged image block through the crack feature suppression degree to obtain the improved edge detection result of each surface defect merged image block, performing conveyor belt defect detection according to the improved edge detection result, and completing the status warning of the electrical automation feeding equipment based on the defect detection result.

2. The detection method for an electrical automation feeding device according to claim 1, characterized in that, The merging of the initial conveyor belt image blocks to obtain at least one surface defect merged image block includes: For any two adjacent initial conveyor belt image blocks: Taking one of the initial conveyor belt image blocks as the first initial conveyor belt image block, and taking the other initial conveyor belt image block as the second initial conveyor belt image block; taking the pixel points adjacent to the block boundary in the first initial conveyor belt image block as the first boundary pixel points; taking the pixel points adjacent to the block boundary in the second initial conveyor belt image block as the second boundary pixel points; matching the first boundary pixel points with the second boundary pixel points to obtain at least two matching binary groups, where the first boundary pixel point and the second boundary pixel point in the matching binary group are adjacent; taking the matching binary group in which both the first boundary pixel point and the second boundary pixel point are abnormal pixel points as the abnormal matching binary group; counting the number of abnormal matching binary groups corresponding to the first initial conveyor belt image block and the second initial conveyor belt image block; When the block boundary is parallel to the running direction of the conveyor belt, merging the first initial conveyor belt image block and the second initial conveyor belt image block with the number of abnormal matching binary groups greater than a preset first threshold to obtain the corresponding surface defect merged image block; When the block boundary is not parallel to the running direction of the conveyor belt, merging the first initial conveyor belt image block and the second initial conveyor belt image block with the number of abnormal matching binary groups greater than a preset second threshold to obtain the corresponding surface defect merged image block; Wherein, the preset first threshold is greater than the preset second threshold; The method for obtaining the abnormal pixel points includes: For the gray values of the pixel points in the surface image of the conveyor belt at each light intensity level, the corresponding segmentation threshold is obtained by the maximum inter-class variance method; the pixel points in the surface image of the conveyor belt at each light intensity level with gray values less than the corresponding segmentation threshold are taken as abnormal pixel points.

3. A detection method for an electrical automation feeding device according to claim 1, characterized in that, The method for obtaining the contrast feature value includes: Arbitrarily select a surface defect merged image block at any light intensity level as the target surface defect merged image block; take the initial conveyor belt image block included in the target surface defect merged image block as the reference initial conveyor belt image block; calculate the average gray gradient value of the pixel points in each reference initial conveyor belt image block along the running direction of the conveyor belt; calculate the variance of the gray values corresponding to all the pixel points in each reference initial conveyor belt image block; Take the product of the maximized value of the variance of the gray values corresponding to each reference initial conveyor belt image block and the corresponding average gray gradient value as the average gradient amplitude corresponding to each reference initial conveyor belt image block; take the mean value of the average gradient amplitudes of all the reference initial conveyor belt image blocks corresponding to the target surface defect merged image block as the contrast feature value of the target surface defect merged image block.

4. A detection method for an electrical automation feeding device according to claim 2, characterized in that, The method for obtaining the contrast weight includes: In each surface defect merged image block at each light intensity level, take the mean value of the gray values of all the abnormal pixel points as the first gray mean value, take the mean value of the gray values of all the other pixel points outside the abnormal pixel points as the second gray mean value, and take the difference between the second gray mean value and the first gray mean value as the corresponding inter-class gray difference; Maximize the inter-class gray difference corresponding to each surface defect merged image block at each light intensity level to obtain the contrast weight of each surface defect merged image block at each light intensity level.

5. A detection method for an electrical automation feeding device according to claim 1, characterized in that The method for obtaining the crack feature suppression degree includes: Take the product of the contrast feature value and the contrast weight as the contrast feature intensity of each surface defect merged image block at each light intensity level; take the sum of the contrast feature intensities of all the light intensity levels corresponding to each surface defect merged image block as the crack feature suppression degree of each surface defect merged image block.

6. A detection method for an electrical automation feeding device according to claim 1, characterized in that, The method for obtaining the improved edge detection result includes: During the process of the edge detection algorithm processing each surface defect merged image block, take the product of the original gray gradient value of each pixel point after non-maximum suppression and the normalized value of the negative correlation mapping value of the crack feature suppression degree of the surface defect merged image block where it is located as the improved gray gradient value of each pixel point, and obtain the improved edge detection result corresponding to each surface defect merged image block through double-threshold detection according to the improved gray gradient value and the corresponding gradient direction.

7. A detection method for an electrical automation feeding device according to claim 6, characterized in that, The edge detection algorithm is the canny edge detection algorithm.

8. A detection method for an electrical automation feeding device according to claim 1, characterized in that, The conveyor belt defect detection based on the improved edge detection result includes: When there are continuous edges in the improved edge detection results, the area corresponding to the continuous edges is regarded as the crack defect area on the conveyor belt; when there are no continuous edges in the improved edge detection results, there are no crack defects on the corresponding conveyor belt.

9. A detection method for an electrical automation feeding device according to claim 1, characterized in that, The status warning of the electrical automation feeding equipment based on the defect detection results includes: When a crack defect area is detected, the alarm system issues a warning signal, and the coordinates of the center point of the crack defect area are used as the positioning result, and the positioning result and the warning signal are synchronously sent to the terminal device of the staff.

Images