Pin-CasNet: Detecting pin status in transmission lines based on cascade network

Bolt-pin is an important connection component in transmission lines, and missing pins may disintegrate key components in transmission lines. Because of the extremely small size of the transmission line pins and the complex environment, simply applying a universal object detector cannot well detect the missing pins. This work proposes an automatic method for detecting normal and missing pins in transmission lines based on a cascade network. The two-stage networks can respectively detect local regions that may contain pins and pins in local regions step by step. In addition, this work designs a dual branch selective block (DBSB) attention module for pin detection, which can make the second-stage network extract more robust pin features. Experiments on the transmission line pin datasets show that our method has strong robustness and adaptability in complex environments, and can achieve efficient pin detection.