The Subway Pantograph Detection Using Modified Faster R-CNN

To ensure the safe operation of the train in metro system, catenary anomaly detection and alerting security have become a major issue to be resolved. Moreover, effective pantograph detection is an important foundation of catenary anomaly detection. In this paper, we present a novel computer vision pantograph detection system involving Faster R-CNN object detection method. Based on the architecture of deep Convolution Neural Network (CNN), we modify the Faster R-CNN to real-time detect the subway pantograph. It combines region proposal generation with object detection. The results reveal that the approach achieves inspiring detection accuracy with over 94.9%. The system can work in different environment of the subway’s train, at different times throughout the day. It provides important reference for subsequent anomaly detection of catenary.