Edge contour extraction method for post insulator strings in substations

Abstract

The present invention belongs to the field of target detection technology, and in particular, relates to a method for extracting edge contours of a substation post insulator string, characterized in that the method comprises the following steps: Step 1: image preprocessing operation to obtain the basic contour position of the insulator string; Step 2: extracting the left and right contours of the insulator string from the binary image, and calculating the coordinates on the line; Step 3: correcting the straight line coordinates of the center line, using Hough transform to detect the straight line on the center line, and then using the least squares method to fit to obtain the corrected center line coordinates; Step 4: correcting the left and right contour coordinates, and using the center line coordinates obtained after fitting to correct the left and right contour coordinates. The present invention can realize the accurate detection of the post insulator string in the substation by the inspection robot of the substation, and improve the video image detection and analysis capability of the inspection robot.

Description

Technical Field

[0001] The invention belongs to the technical field of target detection, and in particular relates to an edge contour extraction method for a post-type insulator string of a transformer substation. Background Art

[0002] For traditional insulator string detection methods, most of them use the texture, shape and other features of the insulator string to design algorithms to detect the insulator string; for insulator string detection methods based on deep learning, most of them output a rectangular box containing the position information of the insulator string, and further algorithms are often required to extract the precise contour coordinates of the insulator string. In the substation inspection task, since the visual sensor shoots from bottom to top, it is easy to look directly at the sun, resulting in partial overexposure of the captured image, so that part of it is missing. Some researchers have achieved accurate recognition of insulator strings in complex outdoor environments based on the fusion of feature information of lidar and camera. It obtains the environmental weighting coefficient through SVM probability output, which can eliminate the influence of light to a certain extent. However, this method does not solve the detection problem in the case of image loss caused by local overexposure, and the construction process of the target point cloud by lidar is relatively time-consuming. Summary of the invention

[0003] The purpose of the present invention is to provide an edge contour extraction method for a substation pillar insulator string. The coordinates of the center line of the insulator string can be calculated by detecting the left and right contours of the insulator string, and then the coordinates of the center line of the complete part of the image are detected by using the Hough transform. After fitting using the least squares method, the corrected center line coordinates are obtained. The positions of the left and right contours are corrected using the corrected center line coordinates, and then the corrected insulator string contour is obtained.

[0004] The objective of the present invention is achieved through the following technical solutions:

[0005] The edge contour extraction method for a transformer substation post insulator string of the present invention is characterized by comprising the following steps:

[0006] Step 1: Image preprocessing operation to obtain the basic contour position of the insulator string;

[0007] Step 2: Extract the left and right contours of the insulator string from the binary image and calculate the midline coordinates;

[0008] Step 3: Correct the straight line coordinates of the midline, use Hough transform to detect the straight line on the midline, and then use the least squares fitting method to obtain the corrected midline coordinates;

[0009] Step 4: Correct the left and right contour coordinates, and use the midline coordinates obtained after fitting to correct the left and right contour coordinates;

[0010] In step 2, after extracting the left and right contours of the insulator string, it is also necessary to calculate the centerline coordinates of the insulator string.

[0011] The line coordinates are calculated as follows

[0012]

[0013] Among them, Medge old is the set of midline points, are the x-axis and y-axis coordinates of the points on the line, is the coordinate point of the left edge contour of the insulator string, is the coordinate point of the right edge contour of the insulator string, n is the height of the image, and k represents the order of traversal of the coordinate points;

[0014] The step 3 corrects the linear coordinate operation of the center line, corrects the center line coordinates of the insulator string extracted in the previous step, and removes some incorrect center line points caused by local overexposure.

[0015] For the points on the midline, make the following transformations:

[0016] r=xcosθ+ysinθ

[0017] Where r is the distance from the origin to the line, θ is the angle between the line r and the x-axis, (x, y) are the coordinates of the midline point,

[0018] The method for finding the intersection point on (r, θ) is as follows: quantize (r, θ) into (m*n) grids, calculate each corresponding r value within θ∈(0°~180°) with the grid step length, record the position of the grid where its value falls, and add 1 to the cumulative counter of the corresponding grid. After the traversal is completed, the (r, θ) value corresponding to the grid with the largest count value corresponds to the desired straight line.

[0019] Then, the least squares method is used to fit the points on the detected straight line to eliminate noise.

[0020]

[0021] After solving the above formula, the corrected coordinates of the midline point are as follows:

[0022]

[0023] The step 4 of correcting the left and right contour coordinates is to correct the left and right contour coordinates using the coordinates of the corrected center line.

[0024] The corrected left contour coordinates are as follows,

[0025]

[0026] The corrected right contour coordinates are as follows,

[0027]

[0028] Among them, Ledge new is the coordinate of the corrected left contour, Redge new is the coordinate of the corrected right contour, and δ is the allowable error between the center line before correction and the center line after correction.

[0029] In the image preprocessing operation of step 1, the influence of environmental factors is filtered out and converted into a binary image to represent the insulator string and background information.

[0030] Advantages of the present invention:

[0031] The invention provides an edge contour extraction method for a substation post insulator string, and proposes a contour detection method for a substation post insulator string with local overexposure in the image. The method can be used to realize accurate detection of the post insulator strings in the substation by a substation inspection robot, thereby improving the video image detection and analysis capability of the inspection robot. DETAILED DESCRIPTION

[0032] The specific embodiments of the present invention are further described below.

[0033] The edge contour extraction method for a transformer substation post insulator string of the present invention is characterized by comprising the following steps:

[0034] Step 1: Image preprocessing operation to obtain the basic contour position of the insulator string;

[0035] Step 2: Extract the left and right contours of the insulator string from the binary image and calculate the midline coordinates;

[0036] Step 3: Correct the straight line coordinates of the midline, use Hough transform to detect the straight line on the midline, and then use the least squares fitting method to obtain the corrected midline coordinates;

[0037] Step 4: Correct the left and right contour coordinates and use the midline coordinates obtained after fitting to correct the left and right contour coordinates.

[0038] In the image preprocessing operation of step 1, the influence of environmental factors is filtered out and converted into a binary image to represent the insulator string and background information.

[0039] In step 2, after extracting the left and right contours of the insulator string, it is also necessary to calculate the centerline coordinates of the insulator string.

[0040] The line coordinates are calculated as follows

[0041]

[0042] Among them, Medge old is the set of midline points, are the x-axis and y-axis coordinates of the points on the line, is the coordinate point of the left edge contour of the insulator string, is the coordinate point of the right edge contour of the insulator string, n is the height of the image, and k represents the order of traversal of the coordinate points.

[0043] The step 3 corrects the linear coordinate operation of the center line, corrects the center line coordinates of the insulator string extracted in the previous step, and removes some incorrect center line points caused by local overexposure.

[0044] For the points on the midline, make the following transformations:

[0045] r=xcosθ+ysinθ

[0046] Where r is the distance from the origin to the line, θ is the angle between the line r and the x-axis, (x, y) are the coordinates of the midline point,

[0047] The method for finding the intersection point on (r, θ) is as follows: quantize (r, θ) into (m*n) grids, calculate each corresponding r value within θ∈(0°~180°) with the grid step length, record the position of the grid where its value falls, and add 1 to the cumulative counter of the corresponding grid. After the traversal is completed, the (r, θ) value corresponding to the grid with the largest count value corresponds to the desired straight line.

[0048] Then, the least squares method is used to fit the points on the detected straight line to eliminate noise.

[0049]

[0050] After solving the above formula, the corrected coordinates of the midline point are as follows:

[0051]

[0052] The step 4 of correcting the left and right contour coordinates is to correct the left and right contour coordinates using the coordinates of the corrected center line.

[0053] The corrected left contour coordinates are as follows,

[0054]

[0055] The corrected right contour coordinates are as follows,

[0056]

[0057] Among them, Ledge newis the coordinate of the corrected left contour, Redge new is the coordinate of the corrected right contour, and δ is the allowable error between the center line before correction and the center line after correction.

[0058] The present invention is a method for extracting the edge of a substation insulator string under local overexposure conditions, and the specific implementation steps are as follows:

[0059] Step 1: Image preprocessing operation, median filtering and then binarization operation, followed by morphological processing to eliminate the influence of tower lines, etc.

[0060] Step 2: Extract the left and right contours of the insulator string from the binary image and calculate their midline coordinates.

[0061] Step 3: Correct the straight line coordinates of the center line, use Hough transform to detect the straight line on the center line, and then use the least squares fitting method to obtain the corrected center line coordinates.

[0062] Step 4: Correct the left and right contour coordinates.

[0063] Step 1: preprocess the image, and obtain a binary image with white outline of the insulator string and black background.

[0064] Step 2: Calculate the centerline coordinates of the insulator string using the following formula.

[0065]

[0066] Step 3: Calculate the corrected centerline coordinates using Hough transform and least squares method. For the points on the straight line detected by Hough transform, use the least squares method to fit and obtain the corrected centerline equation.

[0067] y=kx+b

[0068] Step 4: Use the corrected center line to correct the coordinates of the left and right contours. The corrected coordinates are as follows.

[0069]

[0070]

[0071] The present invention solves the problem of how to detect the contour of an insulator string when a part of it is missing due to partial overexposure when using a deep learning detection method. The coordinates of the centerline of the detectable left and right contours of the insulator string are calculated, and then the coordinates of the centerline of the complete part of the image are detected using the Hough transform. After fitting using the least squares method, the corrected centerline coordinates are obtained, and the positions of the left and right contours are corrected using the corrected centerline coordinates to obtain the corrected insulator string contour. This method can be used to realize the accurate detection of the substation inspection robot on the post insulator string in the station, and improve the video image detection and analysis capabilities of the inspection robot.

Claims

1. A method for extracting edge contours of a post insulator string in a substation, characterized in that: The following steps are involved: Step 1: Image preprocessing operation to obtain the basic contour position of the insulator string; Step 2: Extract the left and right contours of the insulator string from the binary image and calculate the midline coordinates; Step 3: Correct the straight line coordinates of the midline, use Hough transform to detect the straight line on the midline, and then use the least squares fitting method to obtain the corrected midline coordinates; Step 4: Correct the left and right contour coordinates, and use the midline coordinates obtained after fitting to correct the left and right contour coordinates; In step 2, after extracting the left and right contours of the insulator string, it is also necessary to calculate the centerline coordinates of the insulator string. The line coordinates are calculated as follows Among them, Medge old is the set of midline points, are the x-axis and y-axis coordinates of the points on the line, is the coordinate point of the left edge contour of the insulator string, is the coordinate point of the right edge contour of the insulator string, n is the height of the image, and k represents the order of traversal of the coordinate points; The step 3 corrects the linear coordinate operation of the center line, corrects the center line coordinates of the insulator string extracted in the previous step, and removes some incorrect center line points caused by local overexposure. For the points on the midline, make the following transformations: r=xcosθ+ysinθ Where r is the distance from the origin to the line, θ is the angle between the line r and the x-axis, (x, y) are the coordinates of the midline point, The method to find the intersection point on (r,θ) is as follows: quantize (r,θ) into (n r *n θ ) grids, the specific number is set by the experiment, and each corresponding r value is calculated within θ∈(0°~180°) with the step length of the grid, and the position of the grid where the value falls is recorded, so that the cumulative counter of the corresponding grid is increased by 1. After the traversal is completed, the (r,θ) value corresponding to the grid with the largest count value corresponds to the desired straight line. Then, the least squares method is used to fit the points on the detected straight line to eliminate noise. After solving the above formula, the corrected coordinates of the midline point are as follows: The step 4 of correcting the left and right contour coordinates is to correct the left and right contour coordinates using the coordinates of the corrected center line. The corrected left contour coordinates are as follows, The corrected right contour coordinates are as follows, Among them, Ledge new is the coordinate of the corrected left contour, Redge new is the coordinate of the corrected right contour, and δ is the allowable error between the center line before correction and the center line after correction.

2. The edge contour extraction method for a substation post insulator string according to claim 1 is characterized in that In the image preprocessing operation of step 1, the influence of environmental factors is filtered out and converted into a binary image to represent the insulator string and background information.

Images