Inland waterway vessel speed monitoring based on YOLO and DeepSORT

Normally, methods for monitoring the speed of inland ships include sensors such as the AIS system and radar. We propose a method for detecting the speed of inland ships based on YOLO and DeepSORT to better ensure the safety of inland ship navigation. First, we use the YOLOv7x and DeepSORT algorithms in deep learning to detect and track inland ships. Secondly, we have established an instance segmentation model based on visual images to map pixel coordinates to world coordinates. Finally, using this mapping relationship, we convert the motion pixel vector of the ship in pixel coordinates into the displacement in the world coordinate system and calculate the speed of the ship. Experiments have shown that the average accuracy of our method for measuring the speed of the ship is more than 95%.