Autonomous Diagnosis Method for Defects of Cable Accessories Based on YOLOv3 and Mean-Shift Algorithm by Infrared Images

In daily inspection, infrared images are often used to measure the temperature of cable accessories. However, in the face of a large number of inspection images, traditional manual diagnosis is time-consuming and laborious, and relies too much on manual experience. Therefore, an automated infrared infrared image diagnosis method for cable accessory defects based on YOLOv3 and Mean Shift is proposed. Firstly, the method takes YOLOv3 as the basic model, and adds Mosaic technology to the input to enhance the training effect of the model, realize the recognition and positioning of diagnostic targets, and eliminate the impact of complex background images and non diagnostic foreground images on subsequent processing. Then the Mean Shift clustering algorithm is used to segment the image for the diagnosis object, so as to extract the overheated area quickly and accurately. Finally, the temperature information of overheated area and non overheated area is extracted, and the status diagnosis of cable accessories is realized according to the corresponding diagnostic criteria. The research has certain reference value for the defect diagnosis of cable accessories in practical projects.