The object detection method based on deep learning is widely used in Unmanned Aerial Vehicle(UAV)power line inspection.However,there are few researches on pin defects detection since the pin is too small.According to the characteristics of pins in UAV images,this paper proposes an improved feature extraction network based on Faster RCNN to detect pins on transmission lines.Specifically,due to the pins only appear on the link fittings.We detect the link fittings firstly to narrow the region of interest.Next,in order to highlight the pin target,the deconvolution modules are introduced into the original network to obtain significant image features.In addition,the connection between different layers in the ResNet101 network are modified to fuse the features of different resolutions.Finally,we detect the screws on the link fittings to determine whether there are pins on it.The actual transmission line images taken during the UAV inspection are considered as the experimental data.The experimental results demonstrate that the mean average precision(mAP)value of the improved algorithm is 95.32%,which is better than 90.74%of the original network.It also verifies that the improved algorithm has greater application value in the missing pins detection.