Edge end-based insulator detection method in visible light image
By adopting edge-end-based visible light image detection method in power grid inspection, combining multiple features and clustering technologies, the accuracy and real-time problems of insulator detection in drone inspection are solved, and the system performance and detection efficiency are improved.
Patent Information
- Application Number
- CN202410866355.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-06-28
- Publication Date
- 2024-11-12
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
Existing multi-rotor drones have problems in power grid patrols with short battery life, limited transmission distance of image transmission and inability to analyze real-time data in real-time, and the rapid response and accuracy of image recognition algorithms under high-speed cruise conditions are difficult to meet real-time and reliability requirements.
Adopt the edge-end visible light image detection method, combining grayscale and texture features, shape features and plane features, and improve the accuracy of insulator detection through pixel mean, standard deviation, binarization, relative entropy and K-means clustering and other technologies. And reduce the computational complexity.
It realizes fast and accurate insulator object detection and recognition in complex environments, reduces data transmission load, improves the overall performance of the system, and is suitable for complex and variable lighting conditions and images under different resolutions.
Smart Images

Figure CN118941498A_ABST
Abstract
Description
Technical Field
[0001] The invention belongs to the field of digital image processing, and relates to a target confirmation method in a visible light image, in particular to an insulator detection method in a visible light image based on an edge end. Background Art
[0002] With the explosive growth of my country's power grid, the contradiction between the wide distribution range and long transmission distance of the power grid and the low inspection efficiency and high labor intensity at the current stage has become increasingly prominent. Therefore, it is urgent to explore a new inspection model to adapt to the needs of the rapid development of the power grid.
[0003] Drone intelligent inspection has emerged. The use of efficient and intelligent drone cruising mode can reduce the operation and maintenance pressure of the line, improve the safety of the power grid, enhance the control level of equipment operation status, and make up for the shortcomings of traditional manual inspection. However, there are still some problems with multi-rotor drone inspections: short flight time, limited transmission distance of image transmission signals, and inability to perform real-time data analysis.
[0004] On the other hand, the fast and accurate recognition technology of airborne AI edge images can transfer the calculation and recognition process to the edge of the device for real-time processing. However, under high-speed cruising conditions, the rapid response and accuracy of the image recognition algorithm become key challenges and need to be optimized for fast movement and unstable shooting conditions.
[0005] In order to solve the above problems, it is urgent to develop an efficient insulator target recognition method suitable for edge visible light images, which can quickly and accurately detect and identify insulator targets in complex environments, meet the real-time and reliability requirements of power grid inspection, and at the same time reduce the data transmission load and improve the overall performance of the system. Summary of the invention
[0006] The present invention proposes an insulator detection method in visible light images based on edge ends. According to the imaging feature information of the insulator target in the visible light image, multiple technologies such as the grayscale and texture feature description of the insulator, the shape feature and the insulator confirmation of the insulator plane features adapted to different backgrounds are combined, which effectively improves the accuracy of insulator detection in the visible light image and reduces the computational complexity and storage requirements.
[0007] The specific steps are as follows:
[0008] Step 1: input a visible light image and extract N insulator image slices to be inspected from it;
[0009] Each image slice contains only one potential insulator target; the potential insulator target of slice S is T';
[0010] Step 2: traverse all pixels of the current slice S and calculate the mean and standard deviation of the pixel grayscale values;
[0011] The calculation formulas for the pixel mean μ and variance σ are as follows:
[0012]
[0013] Where f(x,y) is the grayscale value of the pixel with coordinates (x,y) on slice S, and w and h represent the width and height of slice S respectively;
[0014] Step 3: binarize the image slice S using the mean and variance to obtain a binary image g', and further obtain the position of the centroid m0' of the potential insulator target T' in the binary image g';
[0015] The binary image calculation formula is as follows:
[0016] g'(x,y)=l{f(x,y)>μ+σ} (3)
[0017] Where g'(x,y) is the value corresponding to the pixel with coordinate (x,y) on the slice S after binarization, and l{·} is the binarization indicator function;
[0018] The center of mass m0' of the potential insulator target T' is located at (x0', y0') and is calculated as:
[0019]
[0020] Step 4: Traverse the slice S and calculate the probability p of each pixel's gray value appearing in the entire slice i :
[0021]
[0022] where h(i) x,y represents the indicator value of each pixel with gray value i on the slice S;
[0023] h(i) x,y =l{f(x,y)=i} (6)
[0024] Wherein f(x, y)=i represents that the gray value corresponding to the pixel point with coordinates (x, y) in the slice is i;
[0025] Step 5: Using probability, calculate the relative entropy between the random pixel Q of the slice S and the centroid m0', and divide the slice into two categories: potential target area and background area:
[0026] For a random pixel point Q, the relative entropy with the centroid point m0' is calculated as follows:
[0027]
[0028] is the probability of occurrence of the gray value of the centroid point m0', p Q is the probability of occurrence of the gray value of pixel Q;
[0029] Step 6: The relative entropy of each pixel in the slice and the centroid m0' is clustered using K-means, and the clusters with small center values are set to 1 in the binary image, and the updated binary image g is obtained to further obtain the position (x0, y0) of the centroid m0 of the potential insulator target T;
[0030] The coordinates T(x, y) of the potential insulator target T in the binary graph and the centroid m0 satisfy 8-neighborhood connectivity;
[0031] 8-neighborhood connectivity means that in a binary image, each pixel is regarded as a node, and if two pixels are adjacent in the horizontal, vertical or diagonal direction, they are considered to be 8-neighborhood connected.
[0032] Step 7: Use the area S of the potential insulator target T in the binary graph g T , and the position of the centroid m0', update the position (x0, y0) of the centroid m0.
[0033] Step 8: Calculate the aspect ratio r of the slice S and the distance from the centroid m0 to the center of the slice, obtain a sub-block s with a symmetrical relationship from the slice S according to the shape characteristics of the insulator, and calculate the position of the centroid m1 of the sub-block s;
[0034] The aspect ratio r and the distance from the centroid m0 to the center of the slice are calculated as follows:
[0035] r=w / h (8)
[0036] d x =|x0-w / 2| (9)
[0037] d y =|y0-h / 2| (10)
[0038] where d x is the distance between the centroid m0 and the slice center in the x-axis direction, d y is the distance between the centroid m0 and the center of the slice in the y-axis direction.
[0039] Divide into the following three cases to find the position of the center of mass m1:
[0040] (I) If 2 / 3<r<3 / 2 and d x <2,d y<2, then take a sub-block s from the binary image g of the slice S, which satisfies s{(x,y)|x<w / 2,y<h / 2}, and use the following formula to find the position (x1,y1) of its centroid m1:
[0041]
[0042] The distance from the position (x1, y1) of the centroid m1 to the center of the sub-block s:
[0043] sd x =|x1-w / 4|
[0044] sd y =|y1-h / 4| (12)
[0045] where sd x is the distance between the centroid m1 and the center of the sub-block s in the x-axis direction, sd y is the distance between the centroid m1 and the center point of the sub-block s in the y-axis direction;
[0046] If sd x ≥2 or sd y ≥2, then re-take the sub-block s{(x,y)|x<w / 2,h / 2≤y≤h} from the binary image g and use formula (11) to calculate its centroid m1.
[0047] (II) If r>3 / 2, then take s{(x,y)|x<w / 2} as a sub-block from the binary graph g and use formula (11) to calculate its centroid m1.
[0048] (III) If r < 2 / 3, take s{(x, y)|y < h / 2} as a sub-block from the binary graph g and use formula (11) to calculate its centroid m1.
[0049] Step 9: According to the positions of the two mass centers m0 and m1, the symmetry axis straight line of the potential insulator target T is calculated, and the position direction of the potential insulator target T in the slice S is obtained;
[0050] The slope k of the symmetry axis line is calculated as follows:
[0051] k=(y1-y0) / (x1-x0) (13)
[0052] The position and direction of the potential insulator target T in the slice S is calculated as follows:
[0053] y-y0=k(x-x0) (14)
[0054] Among them: the coordinates of the center of mass m0 are (x0, y0), and the coordinates of the center of mass m1 are (x1, y1);
[0055] Step 10: Taking the centroid m0 as the starting point, find the points e1 and e2 on both sides of the symmetry axis that belong to the potential insulator target T and are the farthest from the symmetry axis, and find the points v1 and v2 where the symmetry axis intersects the outer edge contour of the potential insulator target T;
[0056] Step 11: Calculate the area M of the minimum circumscribed rectangle of the potential insulator target T, and calculate the distances ed1, ed2, vd1 and vd2 from each point to the centroid m0 based on the points e1, e2, v1 and v2, and calculate the ratio based on the distances. If the area M and the ratio r ls If all of them meet the preset judgment conditions, the potential insulator target T is determined to be an insulator, otherwise it is determined that the potential insulator target T is not an insulator.
[0057] The judgment conditions are:
[0058] (I) If 1<M / S T <thre1 and r ls >thre2, the potential insulator target T is determined to be an insulator;
[0059] (II) If (I) is not met, it is determined that the potential insulator target T is not an insulator.
[0060] Among them, thre1 and thre2 are thresholds set according to the relationship between the real insulator and the resolution.
[0061] The advantages of the present invention are:
[0062] 1. The present invention provides an insulator detection method based on the visible light image of the edge end. The calculation complexity and amount used in calculating the planar geometric features of the insulator are low, the data required to be saved are small, and the method is suitable for insulator confirmation.
[0063] 2. The present invention discloses an insulator detection method in a visible light image based on an edge end, which uses relative entropy to measure the distance between the current pixel point and the cluster center. It has a better segmentation effect when accurately approaching the boundary and can more accurately distinguish potential targets from the background in the slice.
[0064] 3. The present invention provides an insulator detection method in a visible light image based on an edge end. The planar geometric features of the insulator are universal and change little under various lighting backgrounds. It can be used in complex and changeable environments and images with different resolutions to correctly distinguish between insulator targets and non-insulator targets. BRIEF DESCRIPTION OF THE DRAWINGS
[0065] Figure 1 It is a flow chart of the insulator detection method based on the visible light image of the edge end of the present invention. DETAILED DESCRIPTION
[0066] In order to facilitate those skilled in the art to understand and implement the present invention, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. Obviously, the described embodiments are only partial embodiments of the present invention, rather than all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work should fall within the scope of protection of the present invention.
[0067] The present invention discloses an insulator detection method in a visible light image based on an edge end, such as Figure 1 As shown, the specific steps are as follows:
[0068] Step 1: input a visible light image and extract N insulator image slices to be inspected from it;
[0069] Each image slice contains only one potential insulator target; the potential insulator target of slice S is T'; and the pending insulator image slice is an image slice obtained after preliminary extraction and is uncertain whether it is an insulator target.
[0070] Step 2: traverse all pixels of the current slice S and calculate the mean and standard deviation of the pixel grayscale values;
[0071] The calculation formulas for the pixel mean μ and variance σ are as follows:
[0072]
[0073] Where f(x,y) is the grayscale value of the pixel coordinate (x,y) on slice S, w and h represent the width and height of slice S respectively; Carmack fast square root algorithm is used to calculate the square root. Where f(x,y) and f(x,y) 2 The summation operation is calculated in the same loop, and the product operation of the width w and height h of the insulator image slice to be inspected is also calculated only once.
[0074] Step 3: Use the mean and variance to binarize the image slice S to obtain a binary image g' of the slice, and further obtain the position of the centroid m0' of the potential insulator target T' in the binary image g';
[0075] The threshold used in the binarization operation is the sum of the mean and the standard deviation; the binary image calculation formula is as follows:
[0076] g'(x,y)=l{f(x,y)>μ+σ} (3)
[0077] Where g'(x,y) is the value corresponding to the pixel with coordinate (x,y) on the slice S after binarization, l{·} is the binarization indicator function; its value selection rule is: l{expression with true value}=1, l{expression with false value}=0.
[0078] The potential target T'{(x,y)|g'(x,y)=1} is specified.
[0079] The center of mass m0′ of the potential insulator target T' is located at (x0', y0') and is calculated as:
[0080]
[0081] All summations involving g'(x,y) are performed in the same loop.
[0082] Step 4: Traverse the slice S and calculate the probability p of each pixel's gray value appearing in the entire slice i :
[0083] Use the 8-neighborhood to traverse the binarization result and find the area of the centroid m0'∈T';
[0084]
[0085] where h(i) x,y represents the indicator value of each pixel with gray value i on the slice S;
[0086] h(i) x,y =l{f(x,y)=i} (6)
[0087] Wherein f(x, y)=i represents that the gray value corresponding to the pixel point with coordinates (x, y) in the slice is i;
[0088] Step 5: Using probability, calculate the relative entropy between the random pixel Q of the slice S and the centroid m0', and divide the slice into two categories: potential target area and background area:
[0089] For a random pixel point Q, the relative entropy with the centroid point m0' is calculated as follows:
[0090]
[0091] is the probability of occurrence of the gray value of the centroid point m0', p Q is the probability of occurrence of the gray value of pixel Q; the logarithm is calculated with base 2 for faster calculation.
[0092] Step 6: The relative entropy of each pixel in the slice and the centroid m0' is clustered using K-means, and the clusters with small center values are set to 1 in the binary image, and the updated binary image g is obtained to further obtain the position (x0, y0) of the centroid m0 of the potential insulator target T;
[0093] Using the obtained position of the centroid m0' and the binary graph g', traverse the binary graph g' according to the 8-domain to obtain the connected domain containing the centroid m0', and use the probability of occurrence of the gray value in the insulator slice S to be detected to calculate the relative entropy with base 2 between each pixel and the centroid m0'. Based on the calculation result of the relative entropy, use the K-means algorithm to cluster the points with smaller values and the centroid m0' into one category, and set these points to 1 in the binary graph g' to obtain the updated binary graph g, potential insulator target T and centroid m0.
[0094] Find the connected area to which the centroid m0' belongs and use the K-means algorithm to accurately approximate the boundary of the area based on the relative entropy results of each pixel and the centroid m0'. Cluster the pixels with smaller relative entropy and the centroid m0' into one category to obtain the updated binary graph g of the potential insulator target T{(x, y)|can be connected to the centroid m0 in 8 neighborhoods}; and record the area S of the potential insulator target T T , the search for connected regions uses 8 fields; first, the probability of occurrence of each gray value of the slice is counted, and then the relative entropy is calculated with the centroid and a random point outside the area where the centroid is located as the center to divide the slice into two categories: potential target area and background area;
[0095] 8-neighborhood connectivity means that in a binary image, each pixel is regarded as a node, and if two pixels are adjacent in the horizontal, vertical or diagonal direction, they are considered to be 8-neighborhood connected.
[0096] Step 7: Use the area S of the potential insulator target T in the binary graph g T , and the position of the centroid m0', update the position (x0, y0) of the centroid m0.
[0097] Step 8: Calculate the aspect ratio r of the slice S and the distance from the centroid m0 to the center of the slice, obtain a sub-block s with a symmetrical relationship from the slice S according to the shape characteristics of the insulator, and calculate the position of the centroid m1 of the sub-block s;
[0098] The aspect ratio r and the distance from the centroid m0 to the center of the slice are calculated as follows:
[0099] r=w / h (8)
[0100] d x =|x0-w / 2| (9)
[0101] d y =|y0-h / 2| (10)
[0102] where d x is the distance between the centroid m0 and the slice center in the x-axis direction, d y is the distance between the centroid m0 and the center of the slice in the y-axis direction.
[0103] Divide into the following three cases to find the position of the center of mass m1:
[0104] (I) If 2 / 3<r<3 / 2 and d x <2,d y <2, then take a sub-block s from the binary image g of the slice S, which satisfies s{(x,y)|x<w / 2,y<h / 2}, and use the following formula to find the position (x1,y1) of its centroid m1:
[0105]
[0106] The distance from the position (x1, y1) of the centroid m1 to the center of the sub-block s:
[0107] sd x =|x1-w / 4|
[0108] sd y =|y1-h / 4| (12)
[0109] where sd x is the distance between the centroid m1 and the center of the sub-block s in the x-axis direction, sd y is the distance between the centroid m1 and the center point of the sub-block s in the y-axis direction;
[0110] If sd x ≥2 or sd y ≥2, then re-take the sub-block s{(x,y)|x<w / 2,h / 2≤y≤h} from the binary image g and use formula (11) to calculate its centroid m1.
[0111] (II) If r>3 / 2, then take s{(x,y)|x<w / 2} as a sub-block from the binary graph g and use formula (11) to calculate its centroid m1.
[0112] (III) If r < 2 / 3, take s{(x, y)|y < h / 2} as a sub-block from the binary graph g and use formula (11) to calculate its centroid m1.
[0113] Step 9: According to the positions of the two mass centers m0 and m1, the symmetry axis straight line of the potential insulator target T is calculated, and the position direction of the potential insulator target T in the slice S is obtained;
[0114] Axis of symmetry l: Ax+By+C=0;
[0115] (I) If |x0-x1|>0.1, use the following formula to calculate the axis of symmetry:
[0116]
[0117] A=-k, B=1, C=-b (15)
[0118] (II) In addition to satisfying (I), the symmetry axis is calculated using the following formula:
[0119] A=1,B=0,C=-x0 (16)
[0120] The slope k of the symmetry axis line is calculated as follows:
[0121] k=(y1-y0) / (x1-x0)
[0122] The position and direction of the potential insulator target T in the slice S is calculated as follows:
[0123] y-y0=k(x-x0)
[0124] Among them: the coordinates of the center of mass m0 are (x0, y0), and the coordinates of the center of mass m1 are (x1, y1);
[0125] Step 10. Taking the centroid m0 as the starting point, according to the slope of the symmetry axis l, calculate the points e1 and e2 on both sides of the symmetry axis, which belong to the potential insulator target T and are the farthest from the symmetry axis, and the distances ed1 and ed2 from these two points to m0 in the following two cases:
[0126] (I) If B = 0, traverse along the y axis to the left and right with x0 as the center to find points e1 and e2, record their distances ed1 and ed2 respectively, and use formula (17) to calculate the distance from the traversed point (x, y) to the symmetry axis:
[0127] d(x,y)=|y-y0| (17)
[0128] (II) If B≠0 and , then use equation (18) to calculate the equation of the line l' passing through point m0 and perpendicular to the axis of symmetry:
[0129]
[0130] And find the points in e{(x,y)|A'x+B'y+C'<1,1≤x≤w,1≤y≤h} that belong to the potential insulator target T. With m0 as the center, use formula (19) to calculate points e1 and e2 and ed1 and ed2 at both ends:
[0131]
[0132] Find the points in e{(x,y)|Ax+By+C<1,1≤x≤w,1≤y≤h} that belong to the potential insulator target T, and use formula (19) to find the points v1 and v2 farthest from the centroid m0 in e∩T and their distances vd1 and vd2 from m0, respectively, along the symmetry axis upward and downward.
[0133] Step 11: Calculate the area M of the minimum circumscribed rectangle of the potential insulator target T, and calculate the ratio r based on the distance ls , if the area M and the ratio r ls If all of them meet the preset judgment conditions, the potential insulator target T is determined to be an insulator, otherwise it is determined that the potential insulator target T is not an insulator.
[0134] M=(ed1+ed2)×(vd1+vd2) (20)
[0135]
[0136] Based on the area M and the area S of the potential insulator target T T The ratio of ls The following judgment is made for the value of:
[0137] (I) If 1<M / S T <thre1 and r ls >thre2, the potential insulator target T is determined to be an insulator;
[0138] (II) If (I) is not met, it is determined that the potential insulator target T is not an insulator.
[0139] Among them, thre1 and thre2 are preferably 1.2 and 2 according to the relationship between the real insulator and the resolution.
[0140] Example:
[0141] This application conducted an algorithm experiment on a PC platform. During the experiment, a total of 20,240 slices of undetermined insulators extracted from visible light images with resolutions of 1 meter, 2 meters, and 5 meters were used. The size of each image slice was different, and 5,060 slices actually contained insulators.
[0142] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope of the conclusion of the present invention, which should be covered by the protection scope of the present invention.
Claims
1. An insulator detection method in a visible light image based on an edge, characterized in that: The specific steps are as follows: Step 1: input a visible light image and extract N insulator image slices to be inspected from it; Each image slice contains only one potential insulator target; the potential insulator target of slice S is T'; Step 2: traverse all pixels for the current slice S, calculate the mean and standard deviation of the pixel grayscale value, perform a binarization operation on the image slice S to obtain a binary image g', and further calculate the position of the centroid m0' of the potential insulator target T' in the binary image g'; Step 3: Traverse the slice S and calculate the probability p of each pixel's gray value appearing in the entire slice i , and then the relative entropy between the random pixel point Q of the slice S and the centroid point m0' is obtained, and the slice is divided into two categories: potential target area and background area: For a random pixel point Q, the relative entropy with the centroid point m0' is calculated as follows: is the probability of occurrence of the gray value of the centroid point m0', p Q is the probability of occurrence of the gray value of pixel Q; Step 4: The relative entropy of each pixel in the slice and the centroid m0' is used to cluster the relative entropies of the two types of areas using K-means, and the clusters with small center values are set to 1 in the binary image, and the updated binary image g is obtained, and the position (x0, y0) of the centroid m0 of the potential insulator target T is further obtained; The coordinates T(x, y) of the potential insulator target T in the binary graph and the centroid m0 satisfy 8-neighborhood connectivity; Step 5: Use the area S of the potential insulator target T in the binary graph g T , and the position of the centroid m0', update the position (x0, y0) of the centroid m0; Step 6: Calculate the aspect ratio r of the slice S and the distance from the centroid m0 to the center of the slice respectively, obtain a sub-block s with a symmetrical relationship from the slice S according to the shape characteristics of the insulator, and calculate the position of the centroid m1 of the sub-block s; The aspect ratio r and the distance from the centroid m0 to the center of the slice are calculated as follows: r=w / h <h2 style=";text-align:left;direction:ltr">d<h2 style=";text-align:left;direction:ltr"> x <h2 style=";text-align:left;direction:ltr"> =|x0-w / 2| d y =|y0-h / 2| Where w and h represent the width and height of slice S respectively, d x is the distance between the centroid m0 and the center of the slice in the x-axis direction, d y is the distance between the centroid m0 and the center of the slice in the y-axis direction; the coordinates of the centroid m0 are (x0, y0); Divide into the following three cases to find the position of the center of mass m1: (I) If 2 / 3<r<3 / 2 and d x <2,d y <2, then take a sub-block s from the binary image g of the slice S, which satisfies s{(x,y)|x<w / 2,y<h / 2}, and use the following formula to find the position (x1,y1) of its centroid m1: The distance from the position (x1, y1) of the centroid m1 to the center of the sub-block s: sd x =|x1-w / 4| sd y =|y1-h / 4| where sd x is the distance between the centroid m1 and the center of the sub-block s in the x-axis direction, sd y is the distance between the centroid m1 and the center point of the sub-block s in the y-axis direction; If sd x ≥2 or sd y ≥2, then re-take the sub-block s{(x,y)|x<w / 2,h / 2≤y≤h} from the binary image g and use the above formula to calculate its centroid m1; (II) If r>3 / 2, take s{(x,y)|x<w / 2} as a sub-block from the binary graph g and use the above formula to find its centroid m1; (III) If r < 2 / 3, take s{(x, y)|y < h / 2} as a sub-block from the binary graph g and use the above formula to find its centroid m1; Step 7: According to the positions of the two mass centers m0 and m1, the symmetry axis straight line of the potential insulator target T is calculated, and the position direction of the potential insulator target T in the slice S is obtained; Step 8: Taking the centroid m0 as the starting point, find the points e1 and e2 on both sides of the symmetry axis that belong to the potential insulator target T and are the farthest from the symmetry axis, and find the points v1 and v2 where the symmetry axis intersects the outer edge contour of the potential insulator target T; Step 9: Calculate the area M of the minimum circumscribed rectangle of the potential insulator target T, and calculate the distances ed1, ed2, vd1 and vd2 of each point to the center of mass m0 based on the points e1, e2, v1 and v2, and calculate the ratio based on the distances. If the area M and the ratio r ls If all of them meet the preset judgment conditions, the potential insulator target T is determined to be an insulator, otherwise it is determined that the potential insulator target T is not an insulator.
2. The insulator detection method in the visible light image based on the edge as claimed in claim 1, characterized in that: In step 2, the calculation formulas for the pixel mean μ and variance σ are as follows: Where f(x,y) is the grayscale value of the pixel with coordinates (x,y) on slice S; The binary image calculation formula is as follows: g'(x,y)=l{f(x,y)>μ+σ} Where g'(x,y) is the value corresponding to the pixel with coordinate (x,y) on the slice S after binarization, and l{·} is the binarization indicator function; The center of mass m0' of the potential insulator target T' is located at (x0', y0') and is calculated as:
3. The insulator detection method in the visible light image based on the edge as claimed in claim 1, characterized in that: In step 3, the probability p i The calculation formula is: where h(i) x,y represents the indicator value of each pixel with gray value i on the slice S; h(i) x,y =l{f(x,y)=i} Wherein f(x, y)=i represents that the gray value corresponding to the pixel point with coordinates (x, y) in the slice is i.
4. The insulator detection method in the visible light image based on the edge as claimed in claim 1, characterized in that: In step nine, the judgment condition is: (I) If 1<M / S T <thre1 and r ls >thre2, then the potential insulator target T is determined to be an insulator; (II) if (I) is not met, it is determined that the potential insulator target T is not an insulator; Among them, thre1 and thre2 are thresholds set according to the relationship between the real insulator and the resolution.
