Power equipment defect detection method based on CenterNet

The infrared images collected by robots during the inspection process are difficult to reflect the texture information of the equipment. Artificial methods or traditional machine learning methods cannot accurately identify and classify power equipment defects, and other environmental factors can easily lead to misjudgment. In response to this issue, a CenterNet based defect detection method for power equipment is proposed in this paper. This method uses CenterNet and combines structured positioning to collect and train on-site infrared image data samples, achieving high accuracy in identifying and locating different substation equipment and components from complex infrared images. Based on the surface temperature range of equipment components and the identified substation equipment type, combined with relevant temperature specifications, infrared image defect detection of power equipment is achieved. The experimental results show that this method improves the detection accuracy of infrared image defect detection in power equipment.