Intelligent Diagnosis Of Composite Insulator Based On Image Recognition Technology

In order to deal with the bad operating environment of composite insulators on transmission lines and the frequent occurrence of accidents,and to solve the problem of low efficiency of manual judgment of faults by grass-roots inspectors,an intelligent diagnosis method for composite insulator defects is proposed in this paper,which includes the positioning of insulator sheds,the identification of different defects,the influence of various environmental factors on abnormal temperature rise and the inversion of infrared spectrum temperature To conduct in-depth research and discussion.Faster RCNN network is used to locate the target insulator.After image enhancement and other preprocessing methods,ellipse detection is used to find the insulator umbrella skirt,and the area only containing composite insulator umbrella skirt is segmented.Different processing methods are adopted to identify and classify the external defects such as damage,contamination and pulverization.At the same time,the thermal defects of composite insulators can be located by the maximum local temperature in the infrared image,which can not only judge the internal defects of insulators,but also be used as a supplement to the visible light image defect recognition method.In view of the problem that the external environment may cause the defective insulator to have no obvious heating,the influence of temperature and humidity,wind speed and electric field strength on the defect temperature rise of composite insulator is explored,which lays a good foundation for accurate positioning of the insulator with abnormal temperature rise.This method is not only suitable for the normalization of temperature rise of composite insulator in multiple environments,but also has strong generalization for other power equipment.In view of the problem that different infrared cameras lead to various infrared spectrum formats and can not be read directly,the temperature value of each pixel in the infrared image is deduced by using the color bar color.The temperature inversion of the infrared spectrum is realized,and the temperature values of each point of the infrared spectrum can be measured in large quantities without purchasing the secondary development right of various infrared manufacturers’ software,which lays a good foundation for the establishment of the infrared database.