UNet Network-Based Infrared Image Segmentation Method for Induction Motor

Intelligent analysis of infrared images has become an efficient means for fault diagnosis of electrical equipment, and the key technology is segmentation of the target equipment. In this paper, we take the overheating region of induction motor as the research object, use UNet neural network based on VGG16, construct a sample data set, build a semantic segmentation network to train a semantic segmentation model, and bifurcate the pixels of the high temperature region of the motor with the background pixels. The method in the paper is used to test 273 induction motor infrared images, and the experimental results show that the MIoU of the method is 96.0%, which can segment the induction motor high temperature region from the test images efficiently and pave the way for the subsequent intelligent diagnosis of induction motor faults.