Research On Electrical Equipments Faults Diagnosis Methods Based On Image Processing Technology

The safe operation of electrical equipment and power grid, is not only related to People’sdaily life, but also seriously affects a country’s social and economic activities. Thus power failuresearly prevention and maintenance is very important, the fault diagnosis of electrical equipmentshas been a hot topic. Traditional method of electrical equipments fault diagnosis is usually madeby electrical engineers, who periodically use instruments or meters to get electrical equipmentoperation parameters and analyse if it exists faults. This method is not only unable to get real-timeresults, but also spends a lot of time and manpower costs. Infrared thermal imaging technique is anon-destructive technique. In recent years, it’s gradually used widely in electric power system.This paper proposes two feasible methods for fault diagnosis based on image processingtechnology with infrared thermal imager to obtain the thermal images of electrical equipments.Details are as follows:At first, this paper puts forward a kind of extended Mean Shift algorithm, based on thetraditional iterative idea of Mean Shift algorithm. The histogram is used to describe the target andthe weighted vector is used to locate the target position, then the problem of fault diagnosis ofelectrical equipments is transferred to the detection of modal by extended Mean shift algorithm. Inthe beginning, choose an initial pixel to calculate the Mean Shift vector, the Mean Shift vector willlead to more close to the location of the target modal; Secondly when detecting target modal, theMean Shift kernel function is used to get the real position of the image. Finally, iterative functionof Mean Shift algorithm is used to find all the target modal in the whole image, and to do regionsegmentation of all the true locations. Programming and simulation of the extended Mean Shiftalgorithm is done on MATLAB platform, and fault diagnosis of the qualitative results are obtained.The algorithm features simple calculation, fast calculation speed, and can be applied to real-timemonitoring of electrical equipments.The second method in this paper is the improved MSER algorithm. It takes advantage of theMSER excellent features such as affine invariant. The first step is feature extraction on the thermal images of electric equipment and find out the most stable extreme regions. The next step is usingellipse to fit for the extreme regions. The final step is standardizing and segmentation. Theimproved MSER algorithm can not only get the qualitative diagnosis results, but also obtain thequantitative data such as temperature, temperature rate from the connection between the imagegrey value and temperature. Then the quantitative data is judged according to the internationalelectrical testing association (NETA) standard to determine the diagnosis.This paper adapts Visual C++, MATLAB programming, combined with OpenCV to design aelectrical equipments fault diagnosis platform. We use the platform to diagnose the failures oftransformers, coils and overhead cables. During the process, we implement the above twoalgorithm and finish the electrical equipments fault diagnosis. The platform verifies theuniversality of the algorithms. Through100experiments, the diagnosis accuracy is about98%.The diagnosis platform is simple, practical, and has a strong reliability.