Identification and Optimization of Infrared Image Data of Power Equipment

At present, mainly rely on inspection personnel handheld infrared thermal imager to take and collect images of all kinds of electrical equipment, and then analyze and judge the state of power equipment, so as to ensure the safe operation of power system. The infrared image of power equipment after data amplification still has defects such as poor resolution, which will affect the results of subsequent image detection and recognition. Therefore, this paper uses the super-resolution reconstruction algorithm to process the amplified image to improve the resolution and contrast. This paper mainly implements super-resolution reconstruction and data augmentation of infrared images of power equipment. Firstly, the super-resolution reconstruction algorithm FSRCNN (fast super-resolution convolutional neural networks) is used to reconstruct the infrared image of power equipment to improve the resolution and contrast of the image, and YOLOv4 network is used to realize image detection and recognition before and after enhancement, which verifies the effectiveness of the algorithm. Then, a variety of online and offline data augmentation methods are used to expand the infrared images of power equipment to improve the generalization ability and robustness of the network. The above preprocessing operations of the infrared image of power equipment pave the way for the subsequent target location and recognition.