A method for identifying and segmenting infrared insulators and steel caps in complex environments

Abstract

The present invention relates to a method for identifying and segmenting infrared insulators and steel caps in a complex environment. The present invention first preliminarily identifies the area where the insulator string is located, narrows the scope, removes interference from larger obstructions such as electric poles, and performs angle correction operations to keep the insulator string horizontal; based on the preliminary identification, the area where the insulator string is located is accurately identified, the scope is further narrowed, the positions of the two lateral end points of the insulator string are determined, and interference from wires and connectors connecting the two ends of the insulator string is removed; finally, according to the morphological characteristics of the insulator string, a segmentation algorithm is established, mathematical modeling is performed on the accurately identified insulator string, and the steel caps on the insulator string are segmented. The present invention can promote the intelligent detection of live faults of insulators, reduce the safety issues of personnel in manual on-site detection, reduce the economic losses caused by power failure detection, avoid the real-time and labor costs of manual statistical analysis, and improve the safety of power grid operation.

Description

Technical Field

[0001] The invention belongs to the field of image processing and relates to a method for identifying and segmenting infrared insulators and steel caps in a complex environment. Background Art

[0002] Insulators are a special type of insulating control that can form good insulation between the current-carrying conductor and the ground in high-voltage transmission lines. The main function of insulators is to support and fix the current-carrying conductor and form good insulation between the current-carrying conductor and the ground. Insulators are usually composed of two parts, porcelain and metal. For this purpose, various electrical and mechanical performance requirements are specified, such as: it is not easy to break down or flash over along the surface under the rated operating voltage, lightning overvoltage and internal overvoltage; it does not cause destruction and damage under the specified long-term and short-term mechanical loads; after long-term operation under the specified mechanical and electrical loads and various environmental conditions, there is no obvious degradation.

[0003] Due to changes in environmental and electrical load conditions, the function of insulators will deteriorate. For example, due to mechanical stress, electromechanical load, atmospheric pollution and other factors, insulators are prone to flashover discharge. The functional degradation of insulators will damage the use and operating life of the entire line, so timely inspection and maintenance are required. At present, insulator flaw detection is divided into two methods: power-off detection and live detection. Among them, power-off detection requires power technicians to cut off the power supply and climb onto the transmission and transformation line for contact detection. Cutting off the main power supply line will bring great economic losses and affect industrial and domestic electricity consumption. Therefore, the power research field has been trying to find a way to detect live. Live detection is divided into contact live detection and non-contact detection. Contact detection uses manual contact detection, which is extremely dangerous, but has a higher accuracy rate. Non-contact detection mainly uses manual analysis based on infrared photography, which requires manual statistical analysis, requires professional photography, and requires a lot of calculation and analysis.

[0004] Currently, the use of intelligent live insulator detection has become an urgent need for the State Grid, but infrared imaging of insulators is difficult to accurately identify and cut due to factors such as equipment obstruction, weather influences, and mutual intersections. Summary of the invention

[0005] The purpose of the present invention is to realize the intelligence of the deteriorated insulator detection process, improve the anti-interference of the identification and segmentation technology, and improve the efficiency of insulator-related operation and maintenance detection in the power system detection, which specifically includes: mathematical modeling of insulators and steel caps, development of anti-disturbance technology, and then accurate identification and segmentation of insulators and steel caps. In industrial applications, based on the identification and segmentation results of the present invention, temperature matching and analysis can be performed, pixel-level temperature and position mapping can be established, temperature curves can be automatically generated, and an automatic identification method for deteriorated insulators can be established.

[0006] The steps of the method of the present invention include:

[0007] Step (1), obtaining the infrared thermal image of the insulator string, performing preliminary processing on the image, removing interference from larger obstructions, narrowing the range, performing preliminary identification on the position of the insulator string, and obtaining approximate position information of the insulator string;

[0008] Step (2), according to the position information of the insulator string obtained by preliminary identification in step (1), the position of the insulator string is determined in the original infrared thermal image of the area where the insulator string is located, and a partial image of the area where the insulator string is located is extracted to perform an angle correction operation;

[0009] Step (3), further accurately identifying the image in step (2), determining the positions of the two end points of the insulator string in the lateral direction, and removing interference from the wires and connectors connecting the two ends of the insulator string;

[0010] Step (4), according to the morphological characteristics of the insulator string, a segmentation algorithm is established, mathematical modeling is performed on the accurately identified insulator string, and the steel caps on the insulator string are segmented;

[0011] The present invention can identify insulators and steel caps on site and cut the steel caps, which is convenient for measuring the temperature of the steel caps and makes the cutting images more accurate to a certain extent, so as to accurately measure the temperature of the insulators.

[0012] The present invention can promote the intelligence of insulator live fault detection, reduce the personnel safety issues of manual on-site detection, reduce the economic losses caused by power failure detection, avoid the real-time and labor costs of manual statistical analysis, and improve the safety of power grid operation. DETAILED DESCRIPTION

[0013] The present invention is further analyzed below in conjunction with specific embodiments.

[0014] Step (1), obtaining the infrared thermal image of the insulator string, performing preliminary processing on the image, removing interference from large obstructions such as electric poles, narrowing the range, performing preliminary identification on the position of the insulator string, and obtaining approximate position information of the insulator string;

[0015] 1-1 Obtain an infrared thermal image of the area where the insulator string is located; due to environmental and equipment factors, the captured infrared thermal image may contain certain noise, which will affect subsequent operations, so the image is filtered to remove the noise.

[0016] Preferably, the filtering operation adopts a median filtering method.

[0017] 1-2 grayscale the image denoised in step 1-1;

[0018] According to formula (1), the RGB color image in step 1-1 is converted into a grayscale image. The color of the insulator string in the infrared image is orange or bright yellow, and yellow and orange are mainly composed of red and green. Therefore, when performing the image grayscale operation, the weight of the value of the pixel in the 'R' channel and the 'G' channel of the RGB image in the conversion formula is set to a higher value, and the weight of the pixel in the 'B' channel in the conversion formula is set to a smaller value, so as to highlight the orange and bright yellow areas in the image.

[0019] The following formula is used to grayscale the infrared image:

[0020] I(x,y)=0.01×[10·B(x,y)+45·R(x,y)+45·G(x,y)] Formula (1) Where:

[0021] I(x,y) is the value of the pixel at coordinate (x,y) in the grayscale image;

[0022] B(x,y) is the value of the pixel at coordinate (x,y) in the 'Blue' channel of the RGB image;

[0023] R(x,y) is the value of the pixel at coordinates (x,y) in the 'Red' channel of the RGB image;

[0024] G(x,y) is the value of the pixel at coordinate (x,y) in the 'Green' channel of the RGB image;

[0025] 1-3 Perform threshold segmentation on the grayscale image in step 1-2 to obtain a binary image; specifically

[0026] The OTSU maximum inter-class variance threshold segmentation algorithm is used to perform threshold segmentation on the grayscale image, so that the brighter target foreground part in the grayscale image is white and the darker background part is black.

[0027] Assume that there is a threshold K that divides all pixels in the image into two categories. Pixels with grayscale values ​​less than K are classified as class A, and pixels with grayscale values ​​greater than K are classified as class B. Based on the concept of variance, write the inter-class variance expression, traverse 0-255 grayscale levels, and the threshold K that maximizes the inter-class variance is the OTSU threshold. The inter-class variance expression is as follows:

[0028] σ 2 =p1·p2·(m1-m2) 2 Formula (2)

[0029] in:

[0030] p1 is the probability that the pixel is classified as class A;

[0031] p2 is the probability that the pixel is classified as class B, p2 = 1-p1;

[0032] m1 is the mean gray value of class A pixels;

[0033] m2 is the mean gray value of class B pixels;

[0034] 1-4 performs morphological processing on the binary image in step 1-3 to eliminate the interference between the wires and the poles and obtain the area where the insulator string is located in the image.

[0035] When the image is binarized in the previous step, since the colors of the utility poles, wires and insulator strings are similar, these interfering objects may also be set as foreground parts. Therefore, the present invention adopts a series of morphological operations to eliminate the interference of these objects.

[0036] 1-4-1 Erode the binary image in step 1-3 to disconnect the thinner wires from the insulator string.

[0037] The erosion of the binary image is a conventional technique, so it will not be explained in detail.

[0038] 1-4-2 Since the diameter of the steel cap is small and may also be corroded, the image after corrosion in step 1-4-1 is expanded to reconnect the individual insulators that may be disconnected from each other due to the corrosion of the steel cap into a whole.

[0039] The expansion operation is a conventional technique and thus will not be explained in detail.

[0040] 1-4-3 Perform Hough straight line transform on the image expanded in step 1-4-2, detect straight line segments whose length is greater than 0.3 times the height of the input image, calculate their slopes, obtain the position information of straight line segments whose absolute values ​​of slopes are greater than 1, and draw black straight lines with a width of 0.1 times the width of the input image at the locations of these straight line segments. In this way, the area where the telephone poles are located can be set to a black background to a large extent to reduce interference.

[0041] The interference poles in the image are relatively thick and will remain after the second step. When the infrared photo is taken at a relatively horizontal angle, due to the connection method between the cable where the insulator string is located and the pole, the horizontal inclination angle of the insulator string is less than 45 degrees, while the horizontal inclination angle of the pole is greater than 45 degrees. Therefore, if a straight line with a slope greater than 1 is detected, it can be determined that the straight line is located at the pole and this part is removed.

[0042] 1-4-4 Use a contour detection algorithm based on a nested contour hierarchy to find all contours in the image processed in step 1-4-3, where the contour with the largest area surrounding the white area is the location of the insulator string.

[0043] Step (2), based on the position information of the insulator string obtained by preliminary identification in step (1), the position of the insulator string is determined in the original infrared thermal image of the area where the insulator string is located, and a partial image of the area where the insulator string is located is extracted to perform an angle correction operation.

[0044] 2-1 The shape of the insulator string in the image is approximately a rectangle, and its horizontal side length is longer than the vertical side length. After the morphological processing in the previous step, the overall shape of the insulator string has not changed much and can still be regarded as a rectangle. The minimum circumscribed rectangle is used to obtain the length, width, rotation angle and position information of the minimum enclosing rectangle of the area where the insulator string is located, and the area enclosed by the rectangle is extracted from the input image using a mask operation. According to the overall morphological characteristics of the insulator string and the inclination angle of the minimum enclosing rectangle, the extracted area is corrected to the horizontal direction, that is, the longer side of the rectangle is in the horizontal direction, completing the preliminary recognition of the insulator string.

[0045] Step (3), further accurately identify the image of step (2), determine the positions of the two lateral end points of the insulator string, and remove interference from the wires and connectors connecting the two ends of the insulator string.

[0046] 3-1 Grayscale the image in step (2) according to formula (3) to obtain a grayscale image; then use formula (2) Otsu's maximum inter-class variance method to perform threshold segmentation on the grayscale image to obtain a binary image, so that the area where the insulator string with larger pixel intensity is located in the grayscale image is white, and the background area with smaller pixel intensity is black.

[0047] I(x,y)=0.01×[20·B(x,y)+40·R(x,y)+40·G(x,y)] Formula (3)

[0048] 3-2 Traverse the pixels in the binary image vertically, find the upper boundary pixels in the black and white area, and record the coordinates of these pixels into the height change function h(x), where x is the horizontal coordinate of the edge pixel and h(x) is the vertical coordinate of the edge pixel.

[0049] 3-3 In order to further refine the range, the end point positions of the insulator string are identified according to the morphological characteristics of the insulator string.

[0050] After the preliminary identification in the previous steps, the area where the insulator string is located has been basically determined. Now it is only necessary to distinguish the endpoints of the insulator from the wires and connectors connected to both ends. Since the wires and connectors are thinner than the insulator string and have smaller height fluctuations, the h(x) function is traversed from left to right or from right to left to analyze the height fluctuation. The dividing point between the interval with smaller height fluctuation and the interval with larger height fluctuation is the endpoint of the insulator string, and the horizontal coordinates of the two endpoints are marked as st_left and st_right.

[0051] Step (4): according to the morphological characteristics of the insulator string, a segmentation algorithm is established to mathematically model the accurately identified insulator string and segment the steel caps on the insulator string.

[0052] 4-1 Traverse the function h(x) in the interval [st_left,st_right], find the peaks and troughs of the height change function h(x) according to the judgment conditions of the local maximum and minimum points, and save them to the peak array top[] and the trough array bottom[] respectively, where top[k] stores the x-coordinate of the k-th peak pixel, and bottom[k] stores the x-coordinate of the k-th trough pixel.

[0053] 4-2 According to the morphological characteristics of the steel cap in the insulator string, determine the horizontal distribution position and width of the steel cap.

[0054] Determine the horizontal coordinate center_x of the center of the kth steel cap and the width w_steel_cap of the kth steel cap according to the following formula.

[0055]

[0056]

[0057] 4-3 According to the horizontal coordinate of the center of the steel cap, determine the thickness of the steel cap in the vertical direction. In the binary image after threshold segmentation, starting from the pixel point with coordinates (hmax-h[center_x]+2, center_x), traverse the pixel points directly below it until a pixel point with coordinates (m, center_x) is found, whose value is 1. The thickness of the steel cap is calculated according to the following formula.

[0058] det_y=m-(hmax-h[center_x]+2)

[0059] in:

[0060] hmax is the height of the binary image,

[0061] h[center_x] is the height of the edge pixel with the horizontal coordinate center_x, that is, the height of the center position of the upper end of the steel cap.

[0062] 4-4 Based on the length, thickness and position information of the steel cap, the position of the steel cap on the insulator string is framed with a rectangle and numbered to complete the segmentation of the steel cap.

Claims

1. A method for identifying and segmenting infrared insulators and steel caps in a complex environment, the method comprising the following steps: Step (1), obtaining the infrared thermal image of the insulator string, performing preliminary processing on the image, removing interference from larger obstructions, narrowing the range, performing preliminary identification on the position of the insulator string, and obtaining approximate position information of the insulator string; 1-1 Obtain an infrared thermal image of the area where the insulator string is located, and perform a filtering operation on the image to remove noise; 1-2 grayscale the image denoised in step 1-1; According to formula (1), the RGB color image in step 1-1 is converted into a grayscale image, the weights of the values ​​of the pixels in the 'R' channel and the 'G' channel of the RGB image in the conversion formula are set to higher values, and the weights of the values ​​of the pixels in the 'B' channel in the conversion formula are set to lower values; The following formula is used to grayscale the infrared image: I(x,y)=0.01×[10·B(x,y)+45·R(x,y)+45·G(x,y)] Formula (1) Where: I(x,y) is the value of the pixel at coordinate (x,y) in the grayscale image; B(x,y) is the value of the pixel at coordinate (x,y) in the 'Blue' channel of the RGB image; R(x,y) is the value of the pixel at coordinates (x,y) in the 'Red' channel of the RGB image; G(x,y) is the value of the pixel at coordinate (x,y) in the 'Green' channel of the RGB image; 1-3 Perform threshold segmentation on the grayscale image in step 1-2 to obtain a binary image; specifically: The OTSU maximum inter-class variance threshold segmentation algorithm is used to perform threshold segmentation on the grayscale image, so that the brighter target foreground part of the grayscale image is white and the darker background part is black; Assume that there is a threshold K that divides all pixels in the image into two categories, and the pixels with gray values ​​less than K are classified as class A, and the pixels with gray values ​​greater than K are classified as class B; traversing 0-255 gray levels, the threshold K that maximizes the inter-class variance is the OTSU threshold; the inter-class variance expression is as follows: σ 2 =p1·p2·(m1-m2) 2 Formula (2) in: p1 is the probability that the pixel is classified as class A; p2 is the probability that the pixel is classified as class B, p2 = 1-p1; m1 is the mean gray value of class A pixels; m2 is the mean gray value of class B pixels; 1-4 performs morphological processing on the binary image in step 1-3 to eliminate the interference between the wires and the poles and obtain the area where the insulator string is located in the image; specifically: 1-4-1 corrode the binary image in step 1-3 to disconnect the thinner wires from the insulator string; 1-4-2 performing an expansion operation on the image after corrosion in step 1-4-1, so that the individual insulators that may be disconnected from each other due to corrosion of the steel caps are reconnected as a whole; 1-4-3 Perform Hough straight line transform on the image after dilation in step 1-4-2, detect straight line segments whose length is greater than 0.3 times the height of the input image, calculate their slopes, obtain the position information of straight line segments whose absolute values ​​of slopes are greater than 1, and draw black straight lines whose width is 0.1 times the width of the input image at the positions of these straight line segments; 1-4-4 Use a contour detection algorithm based on a nested contour hierarchy to find all contours in the image processed in step 1-4-3, where the contour with the largest area surrounding the white area is the location of the insulator string; Step (2), according to the position information of the insulator string obtained by the preliminary identification in step (1), the position of the insulator string is determined in the original infrared thermal image of the area where the insulator string is located, and a partial image of the area where the insulator string is located is extracted to perform an angle correction operation; specifically: the shape of the insulator string in the image is approximately a rectangle, and its horizontal side length is greater than the vertical side length. After morphological processing, the overall shape of the insulator string does not change much; the length, width, rotation angle and position information of the minimum enclosing rectangle of the outline of the area where the insulator string is located are obtained by using the minimum circumscribed rectangle, and the area enclosed by the rectangle is extracted from the input image by using a mask operation; according to the overall morphological characteristics of the insulator string and the inclination angle of the minimum enclosing rectangle, the extracted area is corrected to the horizontal direction, that is, the longer side of the rectangle is placed in the horizontal direction, thereby completing the preliminary identification of the insulator string; Step (3), further accurately identifying the image in step (2), determining the positions of the two end points of the insulator string in the lateral direction, and removing interference from the wires and connectors connecting the two ends of the insulator string; specifically: 3-1 grayscale the image in step (2) according to formula (3) to obtain a grayscale image; then perform threshold segmentation on the grayscale image using formula (2) Otsu maximum inter-class variance method to obtain a binary image, so that the area where the insulator string with larger pixel intensity is located in the grayscale image is white, and the background area with smaller pixel intensity is black; I(x,y)=0.01×[20·B(x,y)+40·R(x,y)+40·G(x,y)] Formula (3) 3-2 Traverse the pixels in the binary image vertically, find the upper boundary pixels in the black and white area, and record the coordinates of these pixels into the height change function h(x), where x is the horizontal coordinate of the edge pixel and h(x) is the vertical coordinate of the edge pixel; 3-3 is to further refine the range and identify the end point position of the insulator string according to the morphological characteristics of the insulator string; Step (4), according to the morphological characteristics of the insulator string, a segmentation algorithm is established, mathematical modeling is performed on the accurately identified insulator string, and the steel caps on the insulator string are segmented; specifically: 4-1 Traverse the function h(x) in the interval [st_left,st_right], find the peaks and troughs of the height change function h(x) according to the judgment conditions of the local maximum and minimum points, and save them to the peak array top[] and the trough array bottom[] respectively, where top[k] stores the x-coordinate of the k-th peak pixel, and bottom[k] stores the x-coordinate of the k-th trough pixel; 4-2 According to the morphological characteristics of the steel cap in the insulator string, determine the distribution position and width of the steel cap in the horizontal direction; Determine the horizontal coordinate center_x of the center of the k-th steel cap and the width w_steel_cap of the k-th steel cap according to the following formula; 4-3 According to the horizontal coordinate of the center of the steel cap, determine the thickness of the steel cap in the vertical direction; in the binary image after threshold segmentation, start from the pixel point with coordinates (hmax-h[center_x]+2,center_x), traverse the pixel points directly below it, until a pixel point with coordinates (m,center_x) is found, whose value is 1, and the thickness of the steel cap is calculated according to the following formula; det_y=m-(hmax-h[center_x]+2) in: hmax is the height of the binary image; h[center_x] is the height of the edge pixel with the horizontal coordinate center_x, that is, the height of the center position of the upper end of the steel cap; 4-4 Based on the length, thickness and position information of the steel cap, the position of the steel cap on the insulator string is framed with a rectangle and numbered to complete the segmentation of the steel cap.

2. The method according to claim 1, characterized in that The filtering operation in step 1-1 adopts a median filtering method.

3. The method according to claim 1, characterized in that Step 3-3 specifically traverses the h(x) function from left to right or from right to left to analyze the height fluctuation. The dividing point between the interval with smaller height fluctuation and the interval with larger height fluctuation is the endpoint of the insulator string. The horizontal coordinates of the two endpoints are marked as st_left and st_right.

4. An electronic device, characterized in that: The method comprises a processor and a memory, wherein the memory stores machine executable instructions that can be executed by the processor, and the processor executes the machine executable instructions to implement the method according to any one of claims 1 to 3.

5. A machine-readable storage medium, characterized in that: The machine-readable storage medium stores machine-executable instructions. When the machine-executable instructions are called and executed by a processor, the machine-executable instructions prompt the processor to implement the method described in any one of claims 1 to 3.

Images