Study on the method of transmission line foreign body detection based on deep learning

Target detection methods based on region proposal are complicated and the detection speed is slow, which is difficult to meet the requirement of real-time detection of massive power inspection pictures. SSD (Single Shot MultiBox Detector) is simpler than the one that requires region proposal, eliminates the generation of proposals and subsequent feature resampling phases, and encapsulates all calculations in a single network. The model is more easily trained while meeting the high accuracy and real-time detection requirements. In this paper, we described the SSD method in detail, compared and analyzed the DPM (Deformable Parts Model), Faster R-CNN and SSD methods. The three methods are tested and verified by using the actual data set of transmission line foreign body, and the influence of different parameters on the identification results is discussed. The experimental results show that the SSD method can achieve 85.2% of the mAP(mean average precision) at 26FPS (GTX1050Ti), it can meet the real-time requirements while ensuring the accuracy of detection, and the feasibility of the recognition method based on deep learning is verified in real time foreign body detection of transmission line.