Real-time Detection of Pantograph Using Improved CenterNet

With the rapid development of China's railway industry, inspection technology for safe railway operations has become very overwhelming. The pantograph is one of the important components of the high-speed rail power supply system. It is therefore very necessary to accurately monitor the state of the pantograph in real time. In this paper, a fast and accurate pantograph detection model based on improved CenterNet is presented. By replacing the original backbone with lightweight MobileNet and integrating with feature pyramid networks (FPN), our model can achieve the best speed-accuracy trade-off in real-time detection of pantograph.