Research On Image Processing Algorithms For Non-destructive Testing Of GIS Equipment

Gas-insulated metal fully enclosed combined electrical appliances(GIS)are the main switchgear in the power grid,and even more important high-voltage electrical equipment in the UHV power grid.GIS equipment is widely used in power systems because of its advantages such as convenient configuration,small size,reliable operation and long service life.However,in the large-scale and long-term use of GIS equipment,substation accidents caused by GIS are gradually increasing,which has a serious impact on the stable operation of the power grid.Therefore,how to visually inspect the internal status of GIS equipment to prevent it from happening is one of the research hotpots in the field of power grid operation and maintenance.At present,X-ray non-destructive testing technology is widely used in GIS equipment fault detection,but the technology relies on the inspection personnel's experience and technical level to analyze GIS images.Due to the subjectivity of the human eye,it is difficult to guarantee fault analysis under special circumstances Judgment accuracy and low detection efficiency.Aiming at this situation,this paper uses image processing technology to realize the fault detection of the internal components of GIS equipment,reduce the workload of the staff to read the film,improve the detection efficiency,and achieve the purpose of predicting potential faults in advance.This article first analyzes the causes of GIS equipment failures,and classifies the abnormal state of the five components in the GIS that are more common and cause failures.Secondly,the fast median filter is used to smooth the GIS.For the problem of low contrast between the target object and the background in the GIS image,different algorithms are used for grayscale correction.The application results of the algorithm show that: logarithmic transformation,histogram equalization Compared with the method,the gamma transform can significantly improve the image contrast,and an adaptive threshold segmentation algorithm based on C clustering is proposed to segment the GIS image and remove redundant information as much as possible.Then,for edge faults,this paper proposes a Sobel-Gabor edge detection method.Considering the Sobel algorithm for a certain pixel,only considering the horizontal and vertical gradients,the Gabor transform is introduced to increase the multi-directional texture feature extraction of each pixel to achieve Accurate detection and identification of edge faults;for local faults,LBP operator is used to extract texture features,calculate four feature parameters: energy,entropy,contrast,inverse variance,set the range of feature parameters when the component is normal,anddetermine the feature parameters Value to determine whether the component is faulty;for tilt faults,the slope value and the distance change from the edge of the conductive rod to the edge are calculated by mathematical geometry to determine whether the fault is faulty.Finally,in the development environment of VS2017,a GIS non-destructive testing system was established to achieve image acquisition,preprocessing and fault detection and recognition of GIS images.