An Improved Method for Rockfall Detection and Tracking Based on Video Stream

Rockfall events occur frequently in mountainous areas. To address the problems of missed detection, false detection, and trajectory interruption when using the deep learning-based online multiple object tracking methods to detect rockfalls, this paper proposes a rockfall detection and tracking method based on video streams. In the detection stage, three-frame difference method is utilized to obtain the moving targets from the video streams, and they are combined with the detection results of the rock detector obtained by the offline-trained YOLOX model. In the tracking stage, data association is firstly performed based on the rockfall detection results. For the existing trajectories that are not matched at the current moment, re-matching is performed by combining the moving object detection results to achieve accurate tracking of rockfalls. Simulations and field experiments prove that the detection method proposed in this paper can effectively separate the rockfalls in the video, and the detected rockfalls have high precision. Besides, it significantly improves the accuracy of rockfall tracking, effectively suppressing phenomena such as trajectory interruption during tracking.