Research on Bird Nest Image Recognition and Detection Technology of Transmission Lines Based on Improved Faster-RCNN Algorithm

Under certain circumstances, the branches, iron wire, and firewood used by birds in the transmission line will reduce the insulation level of the line, cause the discharge of the wire and cause the trip of the line, which poses a serious threat to the safe and stable operation of the power grid. The traditional bird nest identification and detection algorithm has poor generalization ability, and classification detection accuracy is low. In order to improve the above problems, this paper uses the Inception-v3 feature extraction network to replace the original Faster-RCNN feature extraction network, improves some branches in the Inception-v3 module, and introduces a dual attention detection model between the essential network layers. The model will focus on more information channels to realize the lightweight improvement of the Faster-RCNN algorithm. The experimental results show that compared with the original feature extraction network VGG16, the improved Faster-RCNN algorithm mAP increases by 5.24%, and the recall rate increases by 4.61%. This method effectively improves the recognition rate and detection accuracy of transmission line bird nest detection.