CN111652857B - A kind of infrared detection method of insulator defect - Google Patents

Abstract

The invention discloses an infrared detection method for insulator defects, and relates to the technical field of infrared identification of insulator defects; the method comprises the steps of S1 image preprocessing, S2 insulator segmentation and S3 insulator fault detection, wherein in the step S1 image preprocessing, acquired insulator aerial images are subjected to graying processing to obtain grayed images, the grayed images are subjected to piecewise linear stretching according to the gray rule of the grayed images to obtain grayed stretched images, and the grayed stretched images are subjected to filtering processing to obtain preprocessed images; the insulator defect identification is realized through the steps of S1 image preprocessing, S2 insulator segmentation, S3 insulator fault detection and the like.

Description

A kind of infrared detection method of insulator defect

technical field

The invention relates to the technical field of infrared identification of insulator defects, in particular to an infrared detection method of insulator defects.

Background technique

With the increasing development of smart grid construction, the requirements for intelligent inspection of transmission lines are getting higher and higher. The use of inspection robots and drones based on image recognition technology to gradually replace cumbersome manual inspections and expensive helicopter inspections has been become an inevitable trend. Insulators are the most common components in transmission lines. Once a fault occurs, it will lead to contact between transmission lines and transmission lines or between transmission lines and towers, resulting in interruption of power supply, and even large-scale power outages in severe cases. Due to long-term exposure to the wild, it is often eroded by harsh environments such as storms, rain, and snow. Insulators will inevitably experience various failures: self-explosion, flashover discharge, foreign objects, and dropped strings. Among them, the self-explosion problem of insulator sheets is the most common and urgently needed defect, which mainly occurs on glass insulators. Correctly identify and locate glass insulators, find self-explosion defects and take remedial measures in time, so as to facilitate the effective use of transmission lines and prolong their life. The traditional detection method of insulator self-explosion defect requires a large number of workers to manually interpret, which is time-consuming, labor-intensive, inefficient and expensive.

The traditional detection method of insulator self-explosion defect requires a large number of workers to manually interpret, which is time-consuming, labor-intensive, inefficient and expensive. In this context, drones have been widely used in power inspection. At present, related researches on insulator defect detection at home and abroad include deep learning technology, computer vision technology, image recognition technology, etc. However, there are still many problems in these insulator defect detection technologies, and it is difficult to meet the current fault identification requirements of insulator defects. A suitable infrared identification method for insulator defects is very necessary.

Existing technical problems and thinking:

How to solve the technical problem of identifying insulator defects.

SUMMARY OF THE INVENTION

The technical problem to be solved by the present invention is to provide an infrared detection method for insulator defects, which realizes the identification of insulator defects through the steps of S1 image preprocessing, S2 insulator segmentation, and S3 insulator fault detection.

In order to solve the above technical problems, the technical solution adopted in the present invention is: an infrared detection method for insulator defects includes the steps of S1 image preprocessing, S2 insulator segmentation and S3 insulator fault detection. The insulator aerial image is grayscaled to obtain a grayscale image, and the grayscale image is segmented linearly stretched according to the grayscale law of the grayscale image to obtain a grayscale stretched image. The stretched image is filtered and the preprocessed image is obtained.

A further technical solution is: in the step S2, insulator segmentation is performed, insulator string region positioning is performed on the preprocessed image, and an insulator image is extracted from the candidate region.

A further technical solution is that: the step S2 insulator segmentation includes the steps of S201 insulator identification and positioning and S202 insulator image segmentation, and in the step S201 insulator identification and positioning, according to the grayscale variance of the preprocessed image The image is classified and obtained. Classified images, classified images include simple background images and complex background images. When the grayscale variance is greater than 50, it is a simple background image, and when the grayscale variance is less than or equal to 50, it is a complex background image. The pixel point sparse algorithm is used to process the simple background image or the complex background image. The background image is converted to HSI color space, and then image filtering and image erosion are performed, and the skeletonized image is obtained by image operation.

A further technical solution is: in the step S201 insulator identification and positioning, after the skeletonized image is obtained, the Hough transform algorithm of fuzzy counting is used to detect and obtain the straight line in the skeletonized image, according to the set straight line length of the insulator string. Screen all straight lines and obtain candidate straight lines for the central axis of the insulator, eliminate interference by limiting the detection length, and determine the candidate area for the main axis of the insulator.

A further technical solution is: in the step S201 insulator identification and positioning, after determining the to-be-selected area of the insulator main axis, in the vertical direction of the slope of the candidate straight line, a generalized Hough transform algorithm is used to detect and obtain a disc-shaped curve, that is, a disc-shaped curve. Distributed points, detect and obtain the number of disk-shaped curves in the neighborhood of each candidate area, that is, the number of disks. When the number of disks in the neighborhood exceeds the set threshold, it is determined that the place is an insulator string and the insulator string is realized. fine positioning.

A further technical solution is: in the step S202 insulator image segmentation, extracting and obtaining the insulator pre-processing image from the pre-processing image according to the fine positioning, performing corrosion operation on the insulator pre-processing image, extracting the maximum connected domain and dilation operation processing, and then processing the insulator pre-processing image. Obtain an image of the insulator region.

A further technical solution is: in the step S3 insulator fault detection, based on the insulator area image, the insulator direction is corrected according to the insulator central axis equation, and the insulator string is divided according to the morphological characteristics to obtain a single insulator image.

A further technical solution is: in the step S3 insulator fault detection, after a single insulator image is obtained, the insulator self-explosion fault detection is performed by analyzing and calculating the variation law of the horizontal axis spacing of adjacent insulators.

A further technical solution is: in the step S3 insulator fault detection, after a single insulator image is obtained, the image grayscale co-occurrence matrix is calculated to obtain the texture features of the single insulator image, and the dust-covered stain defects and crack faults of the insulator are judged. Happening.

A further technical solution is: in the step S3 insulator fault detection, after a single insulator image is obtained, the zero-value insulator fault is detected according to the temperature-sensitive characteristic of the infrared image.

The beneficial effects produced by the above technical solutions are:

First, an infrared detection method for insulator defects includes the steps of S1 image preprocessing, S2 insulator segmentation, and S3 insulator fault detection. In the step S1 image preprocessing, the collected aerial images of insulators are subjected to grayscale processing to obtain grayscale images. According to the grayscale law of the grayscale image, piecewise linear stretch is performed on the grayscale image to obtain a grayscale stretched image, and the grayscale stretched image is filtered to obtain a preprocessed image. The technical solution realizes the identification of insulator defects through the steps of S1 image preprocessing, S2 insulator segmentation, and S3 insulator fault detection.

Second, in the step S2, insulator segmentation is performed, insulator string region positioning is performed on the preprocessed image, and an insulator image is extracted from the candidate region. The technical solution has better operation efficiency and higher recognition accuracy.

Third, the step S2 insulator segmentation includes the steps of S201 insulator identification and positioning and S202 insulator image segmentation. In the step S201 insulator identification and positioning, image classification is performed according to the grayscale variance of the preprocessed image and a classified image is obtained, The classified images include simple background images and complex background images. When the grayscale variance is greater than 50, it is a simple background image. When the grayscale variance is less than or equal to 50, it is a complex background image. The pixel sparse algorithm is used to process the simple background image or convert the complex background image. To the HSI color space, then image filtering and image erosion are performed, and the skeletonized image is obtained by image operation. The technical solution has better operation efficiency and higher recognition accuracy.

Fourth, in the step S201 insulator identification and positioning, after the skeletonized image is obtained, the Hough transform algorithm of fuzzy counting is used to detect and obtain the straight lines in the skeletonized image, and all straight lines are screened according to the set straight line length of the insulator string. And obtain the candidate straight line of the central axis of the insulator, eliminate the interference by limiting the detection length, and determine the candidate area of the main axis of the insulator. The technical solution has better operation efficiency and higher recognition accuracy.

Fifth, in the step S201 insulator identification and positioning, after determining the area to be selected for the main axis of the insulator, in the vertical direction of the slope of the candidate straight line, the generalized Hough transform algorithm is used to detect and obtain a disc-shaped curve, that is, a disc-shaped distribution point. , detect and obtain the number of disc-shaped curves in the neighborhood of each candidate area, that is, the number of discs. When the number of discs in the neighborhood exceeds the set threshold, it is determined that the place is an insulator string and the fine positioning of the insulator string is realized. . The technical solution has better operation efficiency and higher recognition accuracy.

Sixth, in the step S202 insulator image segmentation, extract and obtain the insulator pre-processing image from the pre-processing image according to the fine positioning, perform erosion operation on the insulator pre-processing image, extract the maximum connected domain and dilate operation processing, and obtain the insulator area. image. The technical solution has better operation efficiency and higher recognition accuracy.

Seventh, in the step S3 insulator fault detection, based on the insulator area image, the insulator direction is corrected according to the insulator central axis equation, and the insulator string is divided according to the morphological characteristics to obtain a single insulator image. The technical solution has better operation efficiency and higher recognition accuracy.

Eighth, in the step S3 insulator fault detection, after a single insulator image is obtained, the insulator self-explosion fault detection is performed by analyzing and calculating the variation law of the spacing between the horizontal axes of adjacent insulators. The technical solution has better operation efficiency and higher recognition accuracy.

Ninth, in the step S3 insulator fault detection, after a single insulator image is obtained, the grayscale co-occurrence matrix of the image is calculated to obtain the texture features of the single insulator image, and the dust and stain defects and crack faults of the insulator are judged. The technical solution has better operation efficiency and higher recognition accuracy.

Tenth, in the step S3 insulator fault detection, after a single insulator image is obtained, the zero-value insulator fault is detected according to the temperature-sensitive characteristic of the infrared image. The technical solution has better operation efficiency and higher recognition accuracy.

For details, please refer to the description in the detailed description.

Description of drawings

Fig. 1 is the flow chart of the present invention;

Fig. 2 is the flow chart of insulator segmentation algorithm in the present invention;

Fig. 3 is the flow chart of fine positioning algorithm in the present invention;

FIG. 4 is a flow chart of the sparse algorithm in the present invention.

Detailed ways

The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, but not all of the embodiments. The following description of at least one exemplary embodiment is merely illustrative in nature and is in no way intended to limit the application, its application, or uses. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present application.

In the following description, many specific details are set forth to facilitate a full understanding of the present application, but the present application can also be implemented in other ways different from those described herein, and those skilled in the art can do so without departing from the connotation of the present application Similar promotion, therefore, the present application is not limited by the specific embodiments disclosed below.

As shown in FIG. 1 to FIG. 4 , the present invention discloses an infrared detection method for insulator defects, which includes the following steps:

S1 image preprocessing

The collected aerial images of insulators are subjected to grayscale processing to obtain a grayscale image. According to the grayscale law of the grayscale image, the grayscale image is segmented and linearly stretched to obtain a grayscale stretched image. The scaled stretched image is filtered and the preprocessed image is obtained.

S2 insulator segmentation

Perform insulator string area localization on the preprocessed image, and extract the insulator image from the candidate area.

S201 insulator identification and positioning

Image classification is performed according to the grayscale variance of the preprocessed image to obtain the classified image. The classified images include simple background images and complex background images. When the grayscale variance >50, it is a simple background image, and when the grayscale variance is less than or equal to 50, it is a complex background image. The pixel point sparse algorithm processes simple background images or converts complex background images to HSI color space, then performs image filtering and image erosion, uses image operations to obtain skeletonized images, and uses Hough transform algorithm for fuzzy counting to detect and obtain skeletonized images. For the straight line in the image, screen all straight lines according to the set straight line length of the insulator string and obtain the candidate straight line of the central axis of the insulator. By limiting the detection length, the interference is eliminated and the candidate area of the main axis of the insulator is determined. In the vertical direction of the slope of the candidate straight line, the generalized Hough is adopted. The transformation algorithm detects and obtains the disc-shaped curve, that is, the points of the disc-shaped distribution. The number of disc-shaped curves detected and obtained in the neighborhood of each candidate area is the number of discs. When the number of discs in the neighborhood exceeds the set value. If the threshold is set, it is determined that this place is an insulator string and the fine positioning of the insulator string is realized.

S202 insulator image segmentation

The insulator preprocessing image is extracted and obtained from the preprocessing image according to the fine positioning, the insulator preprocessing image is corroded, the maximum connected domain is extracted and the dilation operation is processed, and the insulator region image is obtained.

S3 insulator fault detection

Based on the image of the insulator area, the direction of the insulator is corrected according to the equation of the central axis of the insulator, the insulator string is divided according to the morphological characteristics, and a single insulator image is obtained. The grayscale co-occurrence matrix of the image is calculated to obtain the texture characteristics of a single insulator image, and the dust and stain defects and crack faults of the insulator are judged. According to the temperature-sensitive characteristics of the infrared image, the fault of the zero-value insulator is detected.

Technical features of this application:

The invention aims at the infrared identification method of insulator defects, which specifically includes three main aspects: preprocessing of insulator aerial photography images, segmentation of insulator aerial photography images, and infrared detection of insulator faults based on aerial photography images, and solves the related technical problems.

1. Preprocessing of aerial photography images of insulators: This patent firstly reduces the amount of computational data by performing grayscale processing on the collected image data. Then, according to the grayscale law of the insulator data, the image is segmented and linearly stretched to highlight the target insulator area. Finally, the image is filtered, and the noise that may appear in the process of image data acquisition and transmission is denoised.

2. Segmentation of aerial photography images of insulators: This patent combines the specific features of insulator images and introduces the grayscale variance of images to classify images. For grayscale variance, simple background images have larger values, all greater than 50, while complex background images are all less than 50. For simple background images, a pixel sparse algorithm is used to reduce the amount of computational data; for complex background images, the image quality is optimized by converting the image to HSI color space, and then image filtering and image erosion are used to highlight the target features. Aiming at the problem of image quality after complex background processing, the Hough transform vote counting method is improved. This patent proposes a search algorithm in the candidate area to finally locate the insulator string. Finally, the hybrid algorithm of image opening operation and maximum connected domain extraction is used in the central axis area, and finally the segmentation of the insulator area is realized.

3. Infrared detection of insulator faults based on aerial images: First, the direction correction based on the slope of the central axis is performed on the central axis of the insulator in the image. Then, according to the distribution law of the horizontal axis of the insulator, the insulator string is divided into a single independent insulator. For the self-explosion fault, the monotonicity of the insulator horizontal axis spacing is used to detect and determine the irregularities; for the insulator stain fault, an algorithm for horizontal comparison of the insulator grayscale difference is used to detect the fault area; for the abnormal temperature of the insulator fault, Using the characteristics of temperature-sensitive conversion of infrared images into grayscale transformations, a recognition algorithm is designed to detect abnormalities.

Technical contributions of this application:

The invention provides an infrared identification method for insulator defects, which specifically includes three main aspects: preprocessing of insulator aerial images, segmentation of insulator aerial images, and infrared detection of insulator faults based on aerial images.

1. Preprocessing of aerial images of insulators. Due to some instability when the drone is equipped with a camera, coupled with the weather, sunshine and wind direction, the obtained image data may have some quality defects, including noise interference and insufficient contrast. In order to The accuracy of the subsequent steps requires preprocessing of the collected images. Since the collected image is a three-dimensional RGB image, the color information of the image has little effect on the detection of insulator self-explosion, stains and abnormal temperature faults. In order to reduce the amount of data, this patent uses grayscale images when processing simple background images. In order to make the insulator defects more obvious, this patent proposes an experience-based linear transformation to enhance the image. For noisy images, denoise the image.

2. Segmentation of aerial images of insulators. This patent combines the specific features of the insulator image and introduces the grayscale variance of the image for image classification. For grayscale variance, simple background images have larger values, all greater than 50, while complex background images are all less than 50. For simple background images, a pixel sparse algorithm is used to reduce the amount of computational data; for complex background images, the image quality is optimized by converting the image to HSI color space, and then image filtering and image erosion are used to highlight the target features. Aiming at the problem of image quality after complex background processing, the Hough transform vote counting method is improved. This patent proposes a search algorithm in the candidate area to finally locate the insulator string. Finally, the hybrid algorithm of image opening operation and maximum connected domain extraction is used in the central axis area, and finally the segmentation of the insulator area is realized.

3. Infrared detection of insulator faults based on aerial images. This patent uses visible light and infrared images taken at the same location at the same time to realize comprehensive detection of insulator faults. Firstly, the direction correction based on the slope of the central axis is performed on the central axis of the insulator in the image. Then, according to the distribution law of the horizontal axis of the insulator, the insulator string is divided into a single independent insulator. For the self-explosion fault, the monotonicity of the insulator horizontal axis spacing is used to detect and determine the irregularities; for the insulator stain fault, an algorithm for horizontal comparison of the insulator grayscale difference is used to detect the fault area; for the abnormal temperature of the insulator fault, Using the characteristics of temperature-sensitive conversion of infrared images into grayscale transformations, a recognition algorithm is designed to detect abnormalities.

Technical solution description:

As shown in Figure 1, it is the flow chart of the research route of this patent. The research object of this patent is the visible light and infrared images of aerial insulators. The main research contents are as follows:

1) Preprocessing of aerial images

According to the composition characteristics of aerial photography images of insulators, combined with the background characteristics of specific data sources, appropriate image preprocessing methods are adopted to reduce unnecessary information by graying the images. Enhance the image by grayscale stretching to make the features of the target area more obvious. After image enhancement, filtering is performed to remove noise interference, which lays the foundation for subsequent further processing.

2) Insulator identification and positioning

Firstly, the image is divided into two categories according to the background complexity area of the image, and the targeted extraction algorithm is adopted for different background situations. Then several traditional image segmentation algorithms are simulated, and a segmentation algorithm suitable for text research data is obtained by comparison. Then, the image skeleton is obtained by image operation, and the straight line in the skeletonized image is detected by the Hough transform improved by fuzzy vote counting, and the central axis of the insulator is collected accordingly. Then, the generalized Hough transform detection curve is used to detect the number of disk-like curves in the neighborhood of each candidate area, so as to realize the fine positioning of the insulator string.

3) Insulator image segmentation

The improved image addition operation is performed on the identified insulator string region image, that is, the binarized insulator string region image is first subjected to erosion operation, and then the maximum connected domain of the processed image is extracted, and finally the image is expanded. Segmentation of insulator strings.

4) Insulator defect detection

This patent adopts visible light and infrared images of the target insulator in the same space at the same time to realize the detection of various defects. Firstly, the direction of the insulator is corrected according to the equation of the central axis of the insulator, and the insulator string is divided according to the morphological characteristics to obtain the image of a single insulator. The self-explosion fault detection of insulators is carried out by analyzing and calculating the variation law of the horizontal axis spacing of adjacent insulators; by calculating the gray-scale co-occurrence matrix of the images, the texture features of a single insulator image are obtained to judge the dust-covered stain defects and crack faults of the insulators; finally Based on the temperature-sensitive characteristics of infrared images, zero-value insulator faults are detected.

As shown in Figure 2, the flow chart of the insulator segmentation algorithm. The method first pre-classifies the insulator images, then uses a specific algorithm to locate the insulator string region for images with different background complexities, and then extracts the insulator images from the candidate regions. The method can distinguish the geometrical difference of conductors, towers and insulator strings with similar gray levels, improve the recognition accuracy, and does not need to classify the pixels of the whole image, with less computation and faster running speed. The insulator strings have obvious geometric characteristics in the image. This patent combines its geometric characteristics and according to the above characteristics, firstly skeletonizes the data picture to highlight the linear characteristics of the insulator central axis, and then globally performs line detection based on Hough transform on the processed image. According to the geometric characteristics of the insulator image, A straight line whose length is within a certain range is selected as the candidate straight line for the central axis of the insulator string. Then, in the field with a certain pixel value in the vertical direction of the slope of the candidate line, the point of the disk-shaped distribution is detected based on the generalized Hough transform. If the number of disks in the neighborhood is more than a set threshold, it is determined that the place is an insulator string. Finally, the fine positioning of the insulator string is realized.

As shown in Figure 3, it is the flow chart of the fine positioning algorithm.

Step1: Build a candidate region library. Calculate the length ln of the straight line according to the endpoint coordinates of the candidate central axis obtained by the rough positioning of the insulator, where n is the total number of identified straight lines. The coordinate pairs whose length ln is greater than the shortest length lmin are stored in a new matrix as a candidate matrix. The 4 columns of each row respectively store the starting horizontal and vertical coordinates and the stopping horizontal and vertical coordinates of the central axis of each candidate area. Since the starting and stopping directions of the straight line do not affect the final result, there is no sequence.

Step2: Determine the candidate regions in turn. Set the search width d. Perform the following operations on the candidate lines in the array in sequence: Calculate the slope k of the line according to the coordinates of the endpoints. At the two endpoints, find the two points that are d/2 in length from the endpoint and on a straight line with a slope of -1/k passing through the endpoint. Each vertex constructs a rectangular area, which is the area to be inspected for each central axis.

Step3: Determine the area to be inspected. Set the threshold p. Disk-shaped detection based on generalized Hough transform is performed on the regions to be detected in turn. If the number of discs actually detected exceeds the threshold p, the area is determined to be an insulator string area, and the coordinate endpoints are stored in a new result matrix; if the number of discs detected is less than the threshold p, it is determined to be an interference area Carry out the detection of the next area to be inspected. The area of the final insulator string is the area enclosed by the endpoints stored in the result matrix.

Through the above algorithm, we can select a straight line with a specific geometric shape feature from a plurality of straight lines to be selected, and such a straight line has a high probability of being the central axis of the insulator.

As shown in Figure 4, it is the flow chart of the sparse algorithm. The algorithm idea is to use a method similar to filtering for the binary image after skeletonization, using sliding pixel blocks to move in the image, sparse pixels, and reduce the number of data pixels without affecting the overall outline of the image. On the basis of the shape, the effect of reducing the number of calculations in the Hough transform process is achieved, so as to achieve the purpose of shortening the operation time and improving the efficiency of the algorithm.

After the application has been run confidentially for a period of time, the benefits of feedback from on-site technicians are:

According to the characteristics of aerial images, appropriate preprocessing methods are selected. Including graying the image to reduce the amount of computational data; eradicating the grayscale characteristics of target insulators and interference factors in the image, setting appropriate grayscale thresholds to stretch the image piecewise linearly to highlight the main body of the image; The noise interference generated during transmission is subjected to targeted filtering processing. In the research on the location of insulator strings in aerial images, the images are classified according to the complexity of the image background, and the data set sparse algorithm is used for the insulators with simple backgrounds, which reduces the running time and improves the algorithm efficiency. The image skeleton map with higher quality is obtained by means of filtering and image erosion, and then the traditional Hough transform vote counting mode is improved to finally obtain the main axis area of the image. Accurate positioning of the insulator string is obtained through the circular detection algorithm and the search algorithm in the candidate main shaft area. The identified insulator strings are corrected to remember the main axis direction, and then the image is divided into individual insulator images according to the horizontal axis distribution. Identify the self-explosion fault of the insulator according to the change rule of the main shaft spacing; identify the stain fault according to the grayscale gradient law; identify the abnormal temperature fault of the insulator according to the grayscale reflected temperature. This patent proposes a fault detection method that does not depend on a large amount of data and can be applied to a variety of faults, which improves the intelligence level of power transmission line inspection.

First, an infrared detection method for insulator defects includes the steps of S1 image preprocessing, S2 insulator segmentation, and S3 insulator fault detection. In the step S1 image preprocessing, the collected aerial images of insulators are subjected to grayscale processing to obtain grayscale images. According to the grayscale law of the grayscale image, the grayscaled image is linearly stretched piecewise to obtain a grayscale stretched image, and the grayscale stretched image is filtered to obtain a preprocessed image. Through the steps of S1 image preprocessing, S2 insulator segmentation and S3 insulator fault detection, etc., the identification of insulator defects is realized.

Second, in the step S2, insulator segmentation is performed, insulator string region positioning is performed on the preprocessed image, and insulator images are extracted from the candidate regions, which has better computing efficiency and higher recognition accuracy.

Third, the step S2 insulator segmentation includes the steps of S201 insulator identification and positioning and S202 insulator image segmentation. In the step S201 insulator identification and positioning, image classification is performed according to the grayscale variance of the preprocessed image and a classified image is obtained, The classified images include simple background images and complex background images. When the grayscale variance is greater than 50, it is a simple background image. When the grayscale variance is less than or equal to 50, it is a complex background image. The pixel sparse algorithm is used to process the simple background image or convert the complex background image. To the HSI color space, then image filtering and image erosion are performed, and the skeletonized image is obtained by image operation. The operation efficiency is good, and the recognition accuracy rate is high.

Fourth, in the step S201 insulator identification and positioning, after the skeletonized image is obtained, the Hough transform algorithm of fuzzy counting is used to detect and obtain the straight lines in the skeletonized image, and all straight lines are screened according to the set straight line length of the insulator string. And the candidate straight line of the insulator central axis is obtained, the interference is eliminated by limiting the detection length and the candidate area of the insulator main shaft is determined, the calculation efficiency is good, and the recognition accuracy rate is high.

Fifth, in the step S201 insulator identification and positioning, after determining the area to be selected for the main axis of the insulator, in the vertical direction of the slope of the candidate straight line, the generalized Hough transform algorithm is used to detect and obtain a disc-shaped curve, that is, a disc-shaped distribution point. , detect and obtain the number of disc-shaped curves in the neighborhood of each candidate area, that is, the number of discs. When the number of discs in the neighborhood exceeds the set threshold, it is determined that the place is an insulator string and the fine positioning of the insulator string is realized. , the operation efficiency is better, and the recognition accuracy rate is higher.

Sixth, in the step S202 insulator image segmentation, extract and obtain the insulator pre-processing image from the pre-processing image according to the fine positioning, perform erosion operation on the insulator pre-processing image, extract the maximum connected domain and dilate operation processing, and obtain the insulator area. The image has better computing efficiency and higher recognition accuracy.

Seventh, in the step S3 insulator fault detection, based on the insulator area image, the insulator direction is corrected according to the insulator central axis equation, and the insulator string is segmented according to the morphological characteristics to obtain a single insulator image. higher.

Eighth, in the step S3 insulator fault detection, after obtaining a single insulator image, by analyzing and calculating the variation law of the horizontal axis spacing of adjacent insulators, the self-explosion fault detection of the insulator is performed. The calculation efficiency is good, and the identification accuracy rate is high.

Ninth, in the step S3 insulator fault detection, after a single insulator image is obtained, the grayscale co-occurrence matrix of the image is calculated to obtain the texture features of the single insulator image, and the dust and stain defects and crack faults of the insulator are judged. The efficiency is better, and the recognition accuracy rate is higher.

Tenth, in the step S3 insulator fault detection, after a single insulator image is obtained, the zero-value insulator fault is detected according to the temperature-sensitive characteristic of the infrared image, the calculation efficiency is good, and the identification accuracy rate is high.

Claims (1)

1. an insulator defect infrared detection method, it is characterized in that: comprise the steps of S1 image preprocessing, S2 insulator segmentation and S3 insulator fault detection, described step S1 image preprocessing, the collected insulator aerial image is grayed out Process and obtain a grayscale image, perform piecewise linear stretch on the grayscale image according to the grayscale law of the grayscale image, and obtain a grayscale stretched image, filter the grayscale stretched image, and obtain a pre-scaled image. process images; In the step S2 insulator segmentation, the preprocessed image is located in the insulator string area, and the insulator image is extracted from the candidate area; the step S2 insulator segmentation includes the steps of S201 insulator identification and location and S202 insulator image segmentation, In the step S201 insulator identification and positioning, image classification is performed according to the grayscale variance of the preprocessed image to obtain a classified image. The classified image includes a simple background image and a complex background image. When the grayscale variance > 50, it is a simple background image. If the grayscale variance is less than or equal to 50, it is a complex background image. The pixel point sparse algorithm is used to process the simple background image or convert the complex background image to the HSI color space, and then perform image filtering and image erosion, and use the image operation to obtain the skeletonized image; obtain the skeleton. After transforming the image, the Hough transform algorithm of fuzzy counting is used to detect and obtain the straight lines in the skeletonized image, screen all the straight lines according to the set straight line length of the insulator string, and obtain the candidate straight line of the central axis of the insulator, and eliminate the interference by limiting the detection length. Determine the candidate area for the main axis of the insulator; after determining the candidate area for the main axis of the insulator, in the vertical direction of the slope of the candidate straight line, use the generalized Hough transform algorithm to detect and obtain the disc-shaped curve, that is, the points of the disc-shaped distribution, for each candidate area The number of disc-shaped curves detected and obtained in the neighborhood is the number of discs. When the number of discs in the neighborhood exceeds the set threshold, it is determined that the place is an insulator string and the fine positioning of the insulator string is realized; In the step S202 insulator image segmentation, extracting and obtaining the insulator preprocessing image from the preprocessing image according to the fine positioning, performing erosion operation on the insulator preprocessing image, extracting the maximum connected domain and dilation operation processing, and obtaining the insulator area image; In the step S3 insulator fault detection, based on the insulator area image, the insulator direction is corrected according to the insulator central axis equation, and the insulator string is divided according to the morphological characteristics to obtain a single insulator image; after the single insulator image is obtained, the phase is analyzed and calculated The variation law of the distance between the horizontal axes of adjacent insulators is used to detect the self-explosion fault of the insulator; or after obtaining the image of a single insulator, the grayscale co-occurrence matrix of the image is calculated to obtain the texture characteristics of the single insulator image, and the dust and stain defects and crack faults of the insulator are judged. situation; or after obtaining a single insulator image, according to the temperature-sensitive characteristic of the infrared image, the zero-value insulator fault is detected.

Images