Research on transformer bushing fault recognition based on image identification and support vector machine

The infrared images recorded by the inspection robot can accurately locate the heating defects of the equipment and have an important role in the fault diagnosis of the equipment. The infrared thermal image is completed by the robot in the shooting stage, while the processing and analysis of the later period still need artificial progress. In this paper, the infrared thermal image of the transformer bushing is processed by the method of image recognition and pattern recognition, which is concerned with the over-reliance on the artificial characteristics of the infrared thermal image processing and analysis process. Firstly, The Normalized Cross Correlation (NCC) template matching method and Otsu threshold segmentation method are used to get the region of the interests (ROI) of the bushing. Then the maximum temperature rise, temperature mean value, temperature variance, temperature gradient and other characteristics of the ROI are extracted. Finally, support vector machine is used to identify the status of bushing. The results show that the proposed model can reduce the manual interference of infrared thermal image and has high accuracy, which is suitable for engineering application.