Method for segmenting abnormal temperature region of infrared image of transformer

Abstract

The invention belongs to the technical field of transformer infrared image abnormal temperature region segmentation, and particularly relates to a transformer infrared image abnormal temperature region segmentation method. The method comprises the following steps: firstly, converting an infrared thermogram I into a gray level image G, denoising the gray level image, preliminarily segmenting the gray level image G by adopting an Otsu method to obtain an initial abnormal temperature region R1, obtaining a region edge pixel point value, analyzing a contour edge point pixel value, obtaining a region growth seed point, and performing region growth by taking the region growth seed point as a central point; the method comprises the following steps of: obtaining a new region R1, obtaining a new region R2, segmenting the new region by adopting a maximum between-class variance method to obtain an abnormal region R3, and carrying out intersection and union on the obtained new region and a region R1 obtained by initial segmentation to obtain a new region as an abnormal heating region of the transformer. The edge segmentation of the hot spot region is completed according to the difference between the hot spot region of the infrared image and the infrared image.

Description

Technical Field

[0001] The invention belongs to the technical field of abnormal temperature region segmentation of transformer infrared images, and in particular relates to a method for segmenting abnormal temperature regions of transformer infrared images. Background Art

[0002] Transformer equipment in power plants operates in a high voltage and high current environment for a long time, and the operating conditions are relatively harsh. At present, the daily maintenance of transformers is mainly completed by routine inspections by inspection personnel, which not only has high labor costs and low detection efficiency, but also has a high false detection rate. When a transformer fails, it usually manifests as abnormal local heating of the transformer. Based on this characteristic, infrared thermal imagers are used to monitor the transformer, which can intuitively find abnormal temperature points and guide maintenance personnel to perform repairs.

[0003] The infrared thermal image of the transformer during operation is obtained by an infrared thermal imager, and the fault point is segmented and identified according to the hot spots in different areas of the infrared thermal image to complete the evaluation of the transformer's operating status. However, the infrared image quality collected by the infrared thermal imager is poor and the signal-to-noise ratio is low, which is not conducive to the segmentation and identification of abnormal temperature areas.

[0004] Common image segmentation methods include edge detection, region segmentation, maximum entropy threshold method, and neural network deep learning segmentation method. Before using deep learning segmentation methods to process infrared images, a large number of sample infrared images need to be labeled and trained. In actual detection, there are generally not enough sample infrared images, and this method does not have better adaptability. Traditional methods such as watershed algorithm, Otsu method, and regional growing algorithms are used to process infrared images. The regional segmentation effect is greatly affected by the quality of infrared images. The present invention improves the image processing method to achieve a good effect of complete edge segmentation of abnormal heating areas. Summary of the invention

[0005] The object of the present invention is to provide a method for segmenting abnormal temperature regions of transformer infrared images, and to complete the edge segmentation of hot spot regions according to the hot spot regions and infrared differences of infrared images.

[0006] In order to achieve the above object, the technical solution adopted by the present invention is:

[0007] A method for segmenting abnormal temperature regions of transformer infrared images is provided. First, the infrared thermal image I is converted into a grayscale image G, the grayscale image is denoised, and the grayscale image G is preliminarily segmented by the maximum inter-class variance method to obtain an initial abnormal temperature region R1, and the pixel values ​​of the regional edge are obtained. The pixel values ​​of the contour edge points are analyzed to obtain the regional growth seed points, which are used as the center points for regional growth to obtain a new region R2. The new region is segmented by the maximum inter-class variance method to obtain an abnormal region R3, and the new region is intersected and unioned with the region R1 obtained by the initial segmentation to obtain the new region as the abnormal heating region of the transformer.

[0008] Histogram equalization is used to denoise the grayscale image.

[0009] The following steps are involved:

[0010] The first step is to convert the collected transformer infrared image into a grayscale image and perform histogram equalization calculation to finally obtain a new grayscale image G;

[0011] In the second step, the maximum inter-class variance method is used to complete the preliminary segmentation of the fault hotspot area on the obtained image G, and the segmentation threshold T is calculated to obtain the segmented area R1, where the maximum inter-class variance calculation formula is:

[0012] σ 2 =w0w1(μ0-μ1)

[0013] w0 is the ratio of foreground pixels to the entire image, with an average grayscale of μ0; the ratio of background pixels to the entire image is w1, with an average grayscale of μ1;

[0014] The third step is to draw the mapping contour of the region R1 on the grayscale image G according to the segmented region R1 to obtain the contour point coordinates and grayscale value P (x, y, z), where x, y are the set of contour coordinates, and z is the grayscale value of the pixel at the coordinate point;

[0015] The fourth step is to obtain the corresponding coordinate pixel values ​​on the grayscale image G according to the regional contour, calculate the absolute values ​​of the slopes of two adjacent pixel values, and take the coordinate points whose absolute values ​​of the slopes are greater than the average value as the regional growth seed points (x0, y0);

[0016] The fifth step is to set the regional growth seed point, taking the contour point coordinates (x0, y0) as the center, considering the grayscale values ​​of the pixels in the surrounding area of ​​(x0, y0) and the seed point. If the absolute value of the grayscale value difference is within the set threshold, it is considered that these points belong to the same region and are merged. The point is used as a new seed growth point and regional growth continues. Otherwise, the point with a grayscale difference greater than the set threshold is deleted, regional growth stops, and a new segmentation region R2 is obtained at the seed growth point.

[0017] In the sixth step, the region R2 obtained by region growing is segmented again using the maximum inter-class variance method to obtain a new region R3. Region R3 is merged with region R1 to obtain the region of abnormal hot spots in the final transformer infrared image.

[0018] In the second step, by traversing the image pixels, the threshold T that maximizes the inter-class variance is obtained as the segmentation threshold of the grayscale image G to obtain the segmented area R1.

[0019] The fifth step is to consider the grayscale values ​​of the pixels in the four areas around (x0, y0) and the seed point.

[0020] The first step is to perform histogram equalization calculation to change the grayscale of each pixel in the image by changing the histogram of the image, increase the distribution range of the image pixel grayscale, enhance the contrast of the infrared grayscale image, and finally obtain a new grayscale image G.

[0021] The beneficial effects achieved by the present invention are:

[0022] The existing technology has shortcomings such as unclear segmentation edges and segmentation errors when performing infrared thermal image segmentation. By using the maximum inter-class variance method and the improved region growing algorithm mentioned in the present invention, by setting region merging and elimination rules, pixel-level segmentation of hot spot areas can be achieved, thereby avoiding mis-segmentation and missed segmentation of hot spot areas in infrared thermal images. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 Schematic diagram of segmentation area R1 obtained by the first maximum inter-class variance;

[0024] Figure 2 Schematic diagram of a new region R2 obtained by growing the grayscale region according to the coordinates of the contour points;

[0025] Figure 3 The maximum inter-class variance segmentation is performed again for the newly generated region R2 to obtain the R3 region, and R3 is merged with the original region R1 to obtain the final infrared image abnormal temperature region segmentation schematic diagram. DETAILED DESCRIPTION

[0026] The present invention is described in detail below with reference to the accompanying drawings and specific embodiments.

[0027] The present invention first converts the infrared thermal image I into a grayscale image G, and uses histogram equalization to denoise the grayscale image. The maximum inter-class variance method is used to preliminarily segment the grayscale image G to obtain the initial abnormal temperature area R1, obtain the regional edge pixel value, analyze the contour edge point pixel value, and obtain the regional growth seed point. With this as the center point, regional growth is performed to obtain a new area R2, and the new area is segmented by the maximum inter-class variance method to obtain an abnormal area R3. The new area obtained is intersected and unioned with the area R1 obtained by the initial segmentation to obtain a new area as the abnormal heating area of ​​the transformer.

[0028] The following steps are involved:

[0029] In the first step, the collected transformer infrared image is converted into a grayscale image, and histogram equalization is performed to increase the distribution range of image pixel grayscale and enhance image contrast. Finally, a new grayscale image G is obtained.

[0030] In the second step, the maximum inter-class variance method is used to complete the initial segmentation of the fault hotspot area on the obtained image G, and the threshold T is calculated to obtain the segmented area R1. The maximum inter-class variance calculation formula is:

[0031] σ 2 =w0w1(μ0-μ1)

[0032] Where w0 is the ratio of foreground pixels to the entire image, and the average grayscale is μ0. The ratio of background pixels to the entire image is w1, and the average grayscale is μ1. By traversing the image pixels, the threshold T that maximizes the inter-class variance is obtained as the segmentation threshold of the grayscale image G, and the segmented region R1 is obtained.

[0033] The third step is to obtain the corresponding coordinate pixel values ​​on the grayscale image G according to the regional contour, calculate the absolute values ​​of the slopes of each pair of adjacent pixel values, and use the coordinate points whose absolute values ​​of the slopes are greater than the average value as the regional growth seed points (X0, Y0).

[0034] The fourth step is to set the regional growth seed point, taking the contour point coordinates (x0, y0) as the center, considering the grayscale values ​​of the pixels in the four areas around (x0, y0) and the seed point. If the absolute value of the grayscale value difference is within the set threshold, it is considered that these points belong to the same region and merge them, and the point is used as a new seed growth point to continue regional growth. Otherwise, the points with grayscale difference greater than the set threshold are deleted, regional growth stops, and a new segmentation region R2 is obtained at the seed growth point.

[0035] In the fifth step, the region obtained by region growing is segmented again using the maximum inter-class variance method to obtain a new region R3. Region R3 is merged with region R1 to obtain the region of abnormal hot spots in the final transformer infrared image.

[0036] The following steps are involved:

[0037] In the first step, the collected transformer infrared image is converted into a grayscale image and histogram equalization is performed. By changing the image histogram, the grayscale of each pixel in the image is changed, the distribution range of the image pixel grayscale is increased, and the contrast of the infrared grayscale image is enhanced. Finally, a new grayscale image G is obtained.

[0038] In the second step, the maximum inter-class variance method is used to complete the preliminary segmentation of the fault hotspot area on the obtained image G. The segmentation threshold T is calculated by traversing all pixels to obtain the segmented area R1. The maximum inter-class variance calculation formula is:

[0039] σ 2 =w0w1(μ0-μ1)

[0040] Where w0 is the ratio of foreground pixels to the entire image, and the average grayscale is μ0. The ratio of background pixels to the entire image is w1, and the average grayscale is μ1. By traversing the image pixels, the threshold T that maximizes the inter-class variance is obtained as the segmentation threshold of the grayscale image G, and the segmented region R1 is obtained.

[0041] The third step is to draw the mapping contour of the region R1 on the grayscale image G according to the segmented region R1 to obtain the contour point coordinates and grayscale value P (x, y, z), where x and y are the set of contour coordinates, and z is the grayscale value of the pixel at the coordinate point.

[0042] The fourth step is to calculate the absolute values ​​of the slopes of two adjacent pixel values ​​according to the coordinate pixel values ​​of the regional contour on the grayscale image, and take the coordinate points whose absolute values ​​of the slopes are greater than the average of all slopes as the regional growth seed points (x0, y0).

[0043] The fifth step is to set it as the regional growth seed point, taking the contour point coordinates (x0, y0) as the center, considering the grayscale values ​​of the pixels in the four areas around (x0, y0) and the seed point. If the absolute value of the grayscale value difference is within the set threshold, it is considered that these points belong to the same region and are merged, and the point is used as a new seed growth point to continue regional growth. Otherwise, the point with a grayscale difference greater than the set threshold is deleted, the regional growth stops, and a new region R2 is obtained at the seed growth point.

[0044] In the sixth step, the region R2 obtained by region growing is segmented again using the maximum inter-class variance method to obtain a new region R3. Region R3 is merged with region R1 to obtain the region of abnormal hot spots in the final transformer infrared image.

Claims

1. A method for segmenting abnormal temperature regions of transformer infrared images, characterized in that: Firstly, the infrared thermal image I is converted into a grayscale image G, and the grayscale image is denoised. The maximum inter-class variance method is used to preliminarily segment the grayscale image G to obtain the initial abnormal temperature area R1, and the pixel value of the regional edge is obtained. The pixel value of the contour edge point is analyzed to obtain the regional growth seed point, which is used as the center point for regional growth to obtain a new area R2. The new area is segmented by the maximum inter-class variance method to obtain the abnormal area R3. The new area is intersected and unioned with the area R1 obtained by the initial segmentation to obtain the new area as the abnormal heating area of ​​the transformer.

2. The method for segmenting abnormal temperature regions of transformer infrared images according to claim 1 is characterized in that: Histogram equalization is used to denoise the grayscale image.

3. The method for segmenting abnormal temperature regions of transformer infrared images according to claim 1 is characterized in that: The following steps are involved: The first step is to convert the collected transformer infrared image into a grayscale image and perform histogram equalization calculation to finally obtain a new grayscale image G; In the second step, the maximum inter-class variance method is used to complete the preliminary segmentation of the fault hotspot area on the obtained image G, and the segmentation threshold T is calculated to obtain the segmented area R1, where the maximum inter-class variance calculation formula is: s 2 =w0w1(μ0-μ1) w0 is the ratio of foreground pixels to the entire image, with an average grayscale of μ0; the ratio of background pixels to the entire image is w1, with an average grayscale of μ1; The third step is to draw the mapping contour of the region R1 on the grayscale image G according to the segmented region R1 to obtain the contour point coordinates and grayscale value P (x, y, z), where x, y are the set of contour coordinates, and z is the pixel grayscale value of the coordinate point; The fourth step is to obtain the corresponding coordinate pixel values ​​on the grayscale image G according to the regional contour, calculate the absolute values ​​of the slopes of two adjacent pixel values, and take the coordinate points whose absolute values ​​of the slopes are greater than the average value as the regional growth seed points (x0, y0); The fifth step is to set the regional growth seed point, taking the contour point coordinates (x0, y0) as the center, considering the grayscale values ​​of the pixels in the surrounding area of ​​(x0, y0) and the seed point. If the absolute value of the grayscale value difference is within the set threshold, it is considered that these points belong to the same region and are merged. The point is used as a new seed growth point and regional growth continues. Otherwise, the point with a grayscale difference greater than the set threshold is deleted, regional growth stops, and a new segmentation region R2 is obtained at the seed growth point. In the sixth step, the region R2 obtained by region growing is segmented again using the maximum inter-class variance method to obtain a new region R3. Region R3 is merged with region R1 to obtain the region of abnormal hot spots in the final transformer infrared image.

4. The method for segmenting abnormal temperature regions of transformer infrared images according to claim 3 is characterized in that: In the second step, by traversing the image pixels, the threshold T that maximizes the inter-class variance is obtained as the segmentation threshold of the grayscale image G to obtain the segmented area R1.

5. The method for segmenting abnormal temperature regions of transformer infrared images according to claim 3 is characterized in that: The fifth step is to consider the grayscale values ​​of the pixels in the four areas around (x0, y0) and the seed point.

6. The method for segmenting abnormal temperature regions of transformer infrared images according to claim 3 is characterized in that: The first step is to perform histogram equalization calculation to change the grayscale of each pixel in the image by changing the histogram of the image.

7. The method for segmenting abnormal temperature regions of transformer infrared images according to claim 6 is characterized in that: Increase the distribution range of image pixel grayscale, enhance the contrast of infrared grayscale image, and finally obtain a new grayscale image G.

Images