Research On Fault Feature Extraction Method Of Substation Equipment Based On Multispectral Image

With the continuous improvement of national comprehensive strength and the continuous progress of science and technology.Daily production and life for power demand is also increasing,the State Grid system is also developing towards a more intelligent direction,under the premise of production and life is not affected,how to efficiently and stably ensure the continuous operation of transmission lines,has been the goal of China’s power system.In order to achieve this goal,it is necessary to regularly detect the operation status of substation equipment and risk early warning.The common inspection means can not adapt to the development direction of smart grid.Using the multispectral image features of substation equipment,the information of the equipment can be obtained conveniently and quickly,and the fault can be identified.Using image processing technology to replace the traditional manual detection can not only improve the efficiency,but also improve the accuracy of fault diagnosis.At the same time,it can get higher return with lower cost,which is more suitable for the development of smart grid.In this paper,the multispectral image(visible light image and infrared image)of the insulator in the substation equipment is used as the data source,and the insulator region is extracted by using the image processing theory to detect the insulator fault of different image types.The main contents of this paper are as follows:(1)Image preprocessing,first of all,graying the collected visible and infrared images to reduce the amount of data.In the process of image acquisition,it may be affected by noise,resulting in image blur.On the basis of gray image,median filtering method is used to denoise visible and infrared images.(2)For multispectral image matching and recognition,this paper presents a method combining binarization,morphological operation and scale invariant feature transform(SIFT)algorithm.The image is segmented based on PCNN segmentation method to obtain binary image,and then the binary image is divided into multiple suspected target regions by morphological operation,Finally,the suspected target region is matched with the template image by SIFT,and whethe r it is the target region is judged according to the matching points.This method combines the scale rotation invariance of sift method,and can solve the problem of low recognition rate caused by different shooting angles and distances.It also provides a flow method for the matching and recognition of visible and infrared images of insulators.(3)In order to solve the problem of missing disk in the insulator of substation equipment,the original LBP method is improved.The improved LBP features can obtain more texture features of the image.The improved LBP features and hog features are fused in series.The fused method can not only extract the contour features of the object,but also extract the texture features of the object.The improved lbp-hog algorithm is combined with SVM classifier for insulator fault diagnosis.Based on the characteristics of the infrared image,according to the relationship between the temperature and the gray value of the pixel in the infrared image,the threshold method is used to mark the fault area of the substation equipment,and the relative temperature difference method is used to diagnose the heating fault of the equipment.