Target Detection in Infrared Image of Transmission Line Based on Faster-RCNN

Compared with the visual image, the infrared image of the transmission line has lost some image characteristics and the image resolution is lower. In this paper, an improved Faster-RCNN method is used to locate the target in the infrared image of the transmission line. We first construct the infrared image data set of the transmission line and extract the image features by comparing different network models; then we increase the scale and candidate frame when generating target candidate regions in the region proposal network according to the small target features of the infrared image data set. The accuracy of the insulator string (AP) is improved by about 8.4%, and the average accuracy (mAP) is improved by about 3%. Experiments show that this method has higher recognition accuracy when detecting infrared image targets with lower resolution.