Cable defect detection method and system based on image processing

Abstract

The invention relates to the technical field of image processing, in particular to a cable defect detection method and system based on image processing, and the method comprises the steps: obtaining a gray level image of a cable, correcting a preset strong edge threshold value according to the illumination influence degree of any row of pixel points in the gray level image, obtaining the strong edge threshold value of the row of pixel points, and obtaining the strong edge threshold value of the row of pixel points; and taking a median among the fifty-th percentile, the ninety-th percentile and the strong edge threshold values of the pixel point gradient values in the gray level image as the optimal strong edge threshold value of the pixel points in the row, and based on the optimal strong edge threshold value of the pixel points in each row and a preset weak edge threshold value, performing edge detection on the gray level image by using a # imgabs0 # algorithm, and detecting cable defects according to a comparison result of the obtained number of the connected domains and a preset threshold value. According to the invention, the accuracy of cable defect detection can be improved.

Description

Technical Field

[0001] The present invention relates to the field of image processing technology, and more specifically, to a cable defect detection method and system based on image processing. Background Art

[0002] Cables are key infrastructure in power transmission, communication and industrial applications, and their quality directly affects the stability and safety of the system. During the production process, due to factors such as materials, processes and equipment, cables may have defects such as outer skin cracks and surface scratches, which will affect the life of the cable and may cause equipment failure. Therefore, cable defect detection is very important, and with the development of image processing technology, image-based cable defect detection has become the focus of research and application.

[0003] In the related art, a method for detecting cable surface defects is disclosed in a patent application document with publication number CN115222741A, which includes: obtaining two frames of images and grayscale images of the same area, calculating a two-dimensional histogram of the grayscale-neighborhood gradient mean of the grayscale image, and comparing the corresponding histogram differences in the two images. Grayscale pixels with a difference greater than zero are regarded as target points, valid noise points in the target points are identified, and the filtered valid target noise points are clustered to determine the cluster area, and the size of the filter sliding window is determined according to the area. Finally, the filter sliding window is used to denoise the image and perform edge detection to identify the defective area.

[0004] However, the above solution does not fully consider the influence of illumination when applying edge detection algorithm to identify defective areas, resulting in the detected edge being the boundary of the illuminated area. In this case, the edge detection result may be inaccurate, resulting in deviation in the identification of defective areas, thereby reducing the accuracy of cable defect detection. Summary of the invention

[0005] In order to solve the problem of low accuracy in cable defect detection, the present invention provides a cable defect detection method and system based on image processing.

[0006] According to a first aspect of the present invention, there is provided a cable defect detection method based on image processing, comprising: Get a grayscale image of the cable; Calculate the strong edge threshold of any row of pixels in the grayscale image , To preset a strong edge threshold, For the The degree of illumination influence on the row pixels, is the default value, is the hyperbolic tangent function; The median between the 50th percentile, the 90th percentile and the strong edge threshold of the pixel gradient value in the grayscale image is used as the optimal strong edge threshold of the pixel row. Based on the optimal strong edge threshold of each pixel row and the preset weak edge threshold, the optimal strong edge threshold of each pixel row is used. The algorithm performs edge detection on the grayscale image to detect cable defects based on the comparison results of the number of connected domains obtained and the preset threshold; The method for obtaining the degree of illumination influence is as follows: calculating the average value of the correlation between the pixel points in the current row and the pixel points in the neighboring rows, and normalizing them to obtain the local similarity of the pixel points in the current row; The grayscale mean and standard deviation of the pixels in the current row are calculated, and the normalized value of the ratio of the grayscale mean to the product of the standard deviation and the local similarity is used as the degree of illumination influence of the pixels in the current row.

[0007] The present invention combines the degree of illumination influence of each row of pixels and calculates the strong edge threshold of the corresponding row of pixels, thereby reducing the influence of illumination factors and avoiding misjudging the edge of the illumination-affected area as the cable edge, thereby ensuring the accuracy of the edge detection results and improving the accuracy of cable defect detection.

[0008] Preferably, the correlation between the pixel point in the current row and the pixel point in the neighboring row is a comprehensive correlation, and a method for obtaining the comprehensive correlation includes: Taking the current row as the center, preset the neighboring rows of the current row, and calculate the absolute value of the difference between the grayscale mean of the pixel point in the current row and the pixel point in any neighboring row; The Pearson correlation coefficient between the pixel point in the current row and any pixel point in the neighboring row is obtained, and the ratio of the Pearson correlation coefficient to the absolute value is used as the comprehensive correlation between the pixel point in the current row and any pixel point in the neighboring row.

[0009] The present invention takes into account that under the influence of light, the changes in the grayscale values ​​of pixels in different rows will be relatively similar, affecting the accuracy of the correlation between pixels in different rows. The difference in the grayscale values ​​of pixels in different rows is used to further limit the correlation between the corresponding two rows of pixels.

[0010] Preferably, the local similarity of the pixels in the current row satisfies the following relationship: ; In the formula, Represents the grayscale image Local similarity of row pixels; Indicates the number of neighboring rows of the row; Indicates The row pixel and the Pearson correlation coefficient of pixels in neighboring rows; Indicates Grayscale mean of row pixels; Indicates the row The grayscale mean of pixels in neighboring rows; Indicates the absolute value symbol; Represents the normalization function.

[0011] The present invention limits the value range of local similarity to 0-1, which can reduce the amount of calculation and thus reduce the difficulty of calculation.

[0012] Preferably, the degree of illumination influence satisfies the following relationship: ; In the formula, For the The degree of illumination influence on the row pixels; For the Grayscale mean of row pixels; For the The standard deviation of the grayscale values ​​of the row pixels; For the Local similarity of row pixels is the normalization function.

[0013] The present invention combines various data to measure the illumination influence degree of row pixels, thus ensuring the accuracy of the calculation result.

[0014] Preferably, use The algorithm performs edge detection on grayscale images and also includes: Calculate half of the optimal strong edge threshold of each row of pixels to obtain the optimal weak edge threshold of the corresponding row of pixels; The optimal strong edge threshold and the optimal weak edge threshold are used as The algorithm uses dual thresholds to distinguish edge pixel types and obtain strong edge pixels and weak edge pixels in grayscale images.

[0015] The present invention can avoid misjudging the edge of the illumination-affected area as the edge of the cable, thereby ensuring the accuracy of the edge detection result.

[0016] Preferably, the method for obtaining the connected domain includes: The weak edge pixels adjacent to any strong edge pixel are taken as valid pixels, and multiple independent regions connected by the strong edge pixels and the valid pixels are obtained to obtain a connected domain.

[0017] Preferably, detecting cable defects according to a comparison result of the obtained number of connected domains and a preset threshold value includes: Obtaining a preset threshold, and when the number of connected domains obtained is greater than the preset threshold, it is determined that the cable has defects; When the number of connected domains obtained is less than or equal to a preset threshold, the cable is determined to be intact.

[0018] According to a second aspect of the present invention, a cable defect detection system based on image processing is provided. The system includes a memory and a processor. A computer program is stored in the memory. The processor executes the computer program to implement the steps of the first aspect of the present invention.

[0019] The present invention has the following effects: 1. The present invention utilizes When performing edge detection on the grayscale image of the cable, the influence of the lighting factor is reduced, and the edge of the lighting-affected area can be avoided from being misjudged as the edge of the cable, thereby ensuring the accuracy of the edge detection result, thereby improving the accuracy of cable defect detection.

[0020] 2. The present invention not only calculates the strong edge threshold based on the degree of illumination influence, but also combines the 90th percentile and the 50th percentile of the pixel gradient to screen the optimal strong edge threshold, which can avoid the situation where the strong edge threshold is too large or too small, thereby further ensuring the accuracy of the edge detection results.

[0021] 3. The present invention calculates the degree of illumination influence of row pixels by integrating various factors, thereby ensuring the accuracy of the calculation results. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] By reading 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 accompanying drawings, several embodiments of the present invention are shown in an exemplary and non-limiting manner, and the same or corresponding reference numerals represent the same or corresponding parts, wherein: Figure 1 1 is a schematic diagram of the steps of a cable defect detection method based on image processing according to an embodiment of the present invention; Figure 2 is a grayscale image of the cable obtained in an embodiment of the present invention; Figure 3 In the embodiment of the present invention, Schematic diagram showing the result of the algorithm misjudging the edge of the illumination-affected area in the grayscale image as the edge of the cable. DETAILED DESCRIPTION

[0023] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0024] The specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0025] Reference Figure 1 The cable defect detection method based on image processing includes steps S1 to S4, which are as follows: S1: Get a grayscale image of the cable.

[0026] Specifically, an RGB image of the cable may be captured by a camera, and the RGB image may be converted into a grayscale image for the convenience of subsequent processing.

[0027] It should be noted that, since the cable image contains not only the cable but also the background, the present invention uses the semantic segmentation network to preprocess the grayscale image of the cable to obtain a grayscale image containing only the cable, such as Figure 2 shown.

[0028] The process of using the semantic segmentation network to obtain a grayscale image containing only cables is as follows: first, the semantic segmentation network is trained. The training content includes: taking the grayscale images of cables with different backgrounds as sample sets, and setting labels, marking the cable pixels as 1, and marking the background pixels as 0, and then training the semantic segmentation network. The loss function used in training is the cross entropy loss function. When the loss function tends to converge, a trained semantic segmentation network is obtained.

[0029] Then, the grayscale image containing the background and the cable is input into the trained semantic segmentation network, and a binary image of the grayscale image is output; finally, the product of the output result and the corresponding position data in the grayscale image is calculated to obtain a grayscale image containing only the cable.

[0030] In another embodiment, other methods, such as threshold segmentation, edge detection and contour extraction, etc., may also be used to extract the grayscale image containing only the cables. The present invention does not specifically limit the method of segmenting the foreground and background in the grayscale image.

[0031] S2: Calculate the degree of illumination influence of each row of pixels in the grayscale image.

[0032] Step S2 includes steps S20 and S21, which are specifically as follows: S20: Calculate the average value of the correlation between the pixel points in the current row and the pixel points in the neighboring rows, and normalize them to obtain the local similarity of the pixel points in the current row; It should be noted that, due to the smooth surface of the cable, the difference between the incident angle and the reflection angle at the same horizontal position (i.e., the same row of pixels) is small, so the brightness at the same horizontal position is similar; for pixels in different rows, since the cable is a cylinder, the incident angle and the reflection angle corresponding to different rows are different, and the difference between pixels in different rows is relatively large at the boundary of the area with greater illumination influence. Therefore, the present invention uses this feature to calculate the local similarity of each row of pixels, which can measure the influence of illumination on each row of pixels.

[0033] In an exemplary embodiment of the present invention, the correlation between the current row pixel and the neighboring row pixel is a comprehensive correlation, and the comprehensive correlation can be determined by the following steps: Taking the current row as the center, preset the neighboring rows of the current row, and calculate the absolute value of the difference between the grayscale mean of the current row pixel and any neighboring row pixel; obtain the Pearson correlation coefficient between the current row pixel and any neighboring row pixel, and use the ratio of the Pearson correlation coefficient to the absolute value as the comprehensive correlation between the current row pixel and any neighboring row pixel.

[0034] It should be noted that due to the influence of illumination, the grayscale value of the pixels in the same row changes little, which may lead to similar grayscale distributions of pixels in different rows, thereby making the correlation between pixels in different rows greater. Therefore, in order to more accurately measure the correlation between pixels in different rows, the present invention combines the grayscale differences of pixels in different rows to calculate the comprehensive correlation of pixels in corresponding rows.

[0035] Optional, grayscale image The neighboring rows of a row can be the first row, which is the same as the grayscale image Any line between the lines. It should be noted that when the , or When the number of rows is greater than the maximum number of rows in the grayscale image, the neighboring rows of the corresponding part are discarded and the neighboring rows of the other part are used as the first This embodiment does not specifically limit the number of selected neighboring rows.

[0036] Furthermore, after determining the comprehensive correlation between any row of pixels and the neighboring row of pixels, the local similarity of the row of pixels can be calculated. Specifically, the local similarity of any row of pixels satisfies the following relationship: ; In the formula, Represents the grayscale image Local similarity of row pixels; Indicates the number of neighboring rows of the row. In this embodiment The maximum value of is 10; Indicates The row pixel and the Pearson correlation coefficient of pixels in neighboring rows; Indicates Grayscale mean of row pixels; Indicates the row The grayscale mean of pixels in neighboring rows; Indicates the absolute value symbol; Represents the normalization function.

[0037] in, Reflects the grayscale image The row pixel is The comprehensive correlation between the pixels in the neighboring rows. The larger the value, the The row pixel and the The more similar the pixels in the neighboring rows are, the The greater the local similarity of row pixels, the greater the probability that the row pixels are in an area less affected by light.

[0038] In another embodiment, other methods may be used to perform normalization processing, such as not using The function is normalized, and the deformed exponential function is used for normalization, or the hyperbolic tangent function is used for normalization.

[0039] S21: Calculate the grayscale mean and standard deviation of the pixels in the current row, and use the normalized value of the ratio of the grayscale mean to the product of the standard deviation and the local similarity as the illumination influence degree of the pixels in the current row.

[0040] Specifically, the degree of illumination influence of any row of pixels satisfies the following relationship: ; In the formula, For the The degree of illumination influence on the row pixels; For the Grayscale mean of row pixels; For the The standard deviation of the grayscale values ​​of the row pixels; For the Local similarity of row pixels is the normalization function.

[0041] in, The smaller it is, the more the pixels in this row are affected by the light. The larger the value, the higher the overall grayscale value of the pixels in this row. In areas with stronger light intensity, the higher the grayscale value of the pixels, the greater the influence of light on the pixels in this row. When it is smaller, it means that the grayscale values ​​of the pixels in this row are closer, and the overall level of the grayscale values ​​of the pixels in this row is more reliable.

[0042] S3: Calculate the strong edge threshold of any row of pixels in the grayscale image.

[0043] It should be noted that since there are illumination-affected areas in the grayscale image and the edge gradient of the illumination-affected area is large, if the same strong edge threshold is used for pixels in different areas, the edge of the illumination-affected area may be misjudged as the edge of the cable. Figure 3 Therefore, the present invention combines the illumination influence degree of each row of pixels and calculates the strong edge threshold of the corresponding row of pixels, thereby reducing the influence of illumination factors.

[0044] Specifically, the strong edge threshold of any row of pixels satisfies the following relationship: ; In the formula, The grayscale image Strong edge threshold of row pixels; is a preset strong edge threshold. In this embodiment, the preset strong edge threshold is the 75th percentile of the pixel gradient value in the grayscale image. For the The degree of illumination influence on the row pixels, is a preset value. , is the hyperbolic tangent function.

[0045] Among them, when When , it means that the pixel points in this row are greatly affected by the light. In order to avoid the edge of the area affected by the light being misjudged as the edge of the cable, it is necessary to increase the strong edge threshold of the pixel points in this row; on the contrary, when , it means that the pixels in this row are less affected by the light, and it is necessary to lower the strong edge threshold of the pixels in this row.

[0046] Optionally, the Sobel operator or the Prewitt operator may be used to obtain the gradient value of each pixel in the grayscale image. The calculation process of the gradient value of the pixel is a prior art and will not be described in detail in this embodiment.

[0047] The process of obtaining the 75th percentile of the gradient value of a pixel point in a grayscale image is as follows: First, arrange the gradient values ​​of all pixels in the grayscale image in ascending order.

[0048] Then, using the calculation formula , calculate the position of the 75th percentile, where is the amount of data in the gradient value set of all pixels.

[0049] Finally, judge Is it an integer? If it is an integer, the gradient value corresponding to the position is taken as the 75th percentile; if it is a decimal, the calculation formula is used , calculate the gradient value at that location and get the 75th percentile. , are the first Data and data, for The integer part of for The decimal part of .

[0050] S4: The median between the 50th percentile, the 90th percentile and the strong edge threshold of the pixel gradient value in the grayscale image is used as the optimal strong edge threshold of the pixel row, based on the optimal strong edge threshold of each pixel row and the preset weak edge threshold, using The algorithm performs edge detection on the grayscale image to detect cable defects based on the comparison results of the number of connected domains obtained and the preset threshold.

[0051] Among them, the calculation process of the percentile of the data set is a prior art. The present invention only describes in detail the calculation process of the 75th percentile in step S3, and does not describe in detail the 50th percentile and the 90th percentile.

[0052] Optionally, the 50th percentile and the 90th percentile of the gradient value of the pixel in the grayscale image can be recorded as as well as , then , as well as Arrange them in order from large to small, and take the median of the ordered sequence as the first The optimal strong edge threshold for a row of pixels.

[0053] It should be noted that the optimal strong edge threshold determination method adopted by the present invention can effectively avoid the strong edge threshold being too large or too small, thereby ensuring the accuracy of edge detection.

[0054] Optionally, the preset weak edge threshold may be selected within a range of 0.4 times to 0.6 times the optimal strong edge threshold. This embodiment does not specifically limit the setting of the preset weak edge threshold.

[0055] In an exemplary embodiment of the present invention, using The algorithm also includes: Calculate half of the optimal strong edge threshold for each row of pixels to obtain the optimal weak edge threshold for the corresponding row of pixels; use the optimal strong edge threshold and the optimal weak edge threshold as The algorithm uses dual thresholds to distinguish edge pixel types and obtain strong edge pixels and weak edge pixels in grayscale images.

[0056] Among them, using The specific process of the algorithm to distinguish the types of edge pixels is as follows: for any pixel in any row of the grayscale image, if the gradient value of the pixel is greater than the optimal edge threshold of the pixels in the row, the pixel is judged as a strong edge pixel.

[0057] Optionally, if the gradient value of the pixel is between the optimal weak edge pixel and the optimal strong edge pixel of the row of pixels, the pixel is determined to be a weak edge pixel, thereby achieving differentiation of the types of edge pixels.

[0058] In an exemplary embodiment of the present invention, the determination of the connected domain may be achieved by the following steps: The weak edge pixels adjacent to any strong edge pixel are taken as valid pixels, and multiple independent regions connected by the strong edge pixels and the valid pixels are obtained to obtain a connected domain.

[0059] The weak edge pixel points adjacent to any strong edge pixel point are weak edge pixel points within the eight-neighborhood range of the strong edge pixel point.

[0060] Optional, when using When the number of connected domains detected by the algorithm is large, it means that there are more edges in the grayscale image, and there may be cracks and damage on the cable surface. Therefore, whether the cable has defects can be evaluated based on the number of connected domains detected.

[0061] In an exemplary embodiment of the present invention, cable defects may be detected by the following steps: A preset threshold is obtained. When the number of connected domains obtained is greater than the preset threshold, it is determined that the cable has defects; when the number of connected domains obtained is less than or equal to the preset threshold, it is determined that the cable is intact.

[0062] Optionally, the preset threshold can be set to 50. When the number of detected connected domains is greater than 50, it means that there are too many edges in the grayscale image of the cable, and it is determined that there are defects in the cable. Conversely, when the number of detected connected domains is less than or equal to 50, it means that the edges in the grayscale image of the cable are normal edges of the cable, and the cable is determined to be intact, thereby enabling the detection of cable defects.

[0063] The present invention also provides a cable defect detection system based on image processing. The system includes a memory and a processor, and a computer program is stored in the memory. The computer program integrates the functions of the cable defect detection method based on image processing. When the computer program is executed, the accuracy of cable defect detection can be improved through the cable defect detection method based on image processing.

[0064] In the description of this specification, "plurality" or "several" means at least two, such as two, three or more, etc., unless otherwise clearly and specifically defined.

[0065] Although this specification has shown and described a number of embodiments of the present invention, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Those skilled in the art will conceive of many modifications, changes and alternatives without departing from the ideas and spirit of the present invention. It should be understood that in the practice of the present invention, various alternatives to the embodiments of the present invention described herein may be employed.

Claims

1. A cable defect detection method based on image processing, characterized in that: include: Get a grayscale image of the cable; Calculate the strong edge threshold of any row of pixels in the grayscale image , To preset a strong edge threshold, For the The degree of illumination influence on the row pixels, is the default value, is the hyperbolic tangent function; The median between the 50th percentile, the 90th percentile and the strong edge threshold of the pixel gradient value in the grayscale image is used as the optimal strong edge threshold of the row of pixels. Based on the optimal strong edge threshold of each row of pixels and the preset weak edge threshold, the optimal strong edge threshold of each row of pixels is used. The algorithm performs edge detection on the grayscale image to detect cable defects based on the comparison result of the obtained number of connected domains and a preset threshold value; The method for obtaining the degree of illumination influence is: calculating the average value of the correlation between the pixel points in the current row and the pixel points in the neighboring rows, and normalizing them to obtain the local similarity of the pixel points in the current row; The grayscale mean and standard deviation of the pixels in the current row are calculated, and a normalized value of the ratio of the grayscale mean to the product of the standard deviation and the local similarity is used as the illumination influence degree of the pixels in the current row.

2. The cable defect detection method based on image processing according to claim 1, characterized in that: The correlation between the pixel point in the current row and the pixel point in the neighboring row is a comprehensive correlation, and the method for obtaining the comprehensive correlation includes: Taking the current row as the center, preset the neighboring rows of the current row, and calculate the absolute value of the difference between the grayscale mean of the pixel point in the current row and the pixel point in any neighboring row; The Pearson correlation coefficient between the pixel point in the current row and the pixel point in any adjacent row is obtained, and the ratio of the Pearson correlation coefficient to the absolute value is used as the comprehensive correlation between the pixel point in the current row and the pixel point in any adjacent row.

3. The cable defect detection method based on image processing according to claim 2 is characterized in that: The local similarity of the pixels in the current row satisfies the following relationship: ； In the formula, Represents the grayscale image Local similarity of row pixels; Indicates the number of neighboring rows of the row; Indicates The row pixel and the Pearson correlation coefficient of pixels in neighboring rows; Indicates Grayscale mean of row pixels; Indicates the row The grayscale mean of pixels in neighboring rows; Indicates the absolute value symbol; Represents the normalization function.

4. The cable defect detection method based on image processing according to claim 1, characterized in that: The degree of illumination influence satisfies the following relationship: ； In the formula, For the The degree of illumination influence on the row pixels; For the Grayscale mean of row pixels; For the The standard deviation of the grayscale values ​​of the row pixels; For the The local similarity of row pixels, is the normalization function.

5. The cable defect detection method based on image processing according to claim 1, characterized in that: The utilization The algorithm performs edge detection on the grayscale image, and also includes: Calculate half of the optimal strong edge threshold of each row of pixels to obtain the optimal weak edge threshold of the corresponding row of pixels; The optimal strong edge threshold and the optimal weak edge threshold are used as The algorithm distinguishes the double thresholds of edge pixel types and obtains strong edge pixel points and weak edge pixel points in the grayscale image.

6. The cable defect detection method based on image processing according to claim 5 is characterized in that: The method for obtaining the connected domain includes: The weak edge pixel points adjacent to any strong edge pixel point are taken as valid pixel points, and a plurality of independent regions formed by connecting the strong edge pixel points and the valid pixel points are obtained to obtain the connected domain.

7. The cable defect detection method based on image processing according to claim 1, characterized in that: The detecting of cable defects according to the comparison result of the obtained number of connected domains and a preset threshold value comprises: Obtaining a preset threshold, and when the number of connected domains obtained is greater than the preset threshold, determining that the cable has defects; When the number of connected domains obtained is less than or equal to the preset threshold, it is determined that the cable is intact.

8. Cable defect detection system based on image processing, characterized in that: The cable defect detection system based on image processing comprises a memory and a processor, wherein a computer program is stored in the memory, and the processor executes the computer program to implement the steps of the method according to any one of claims 1 to 7.

Images