A Vision System for Shot Tracking and Thrown Distance Measurement

This paper presents a vision system for the tracking of the shot and the measure of the thrown distance in a shot put competition. The pan-tilt-zoom (PTZ) camera is used to capture images of the shot and track the shot to the landing position, and the lidar is used to measure the thrown distance once the landing position is determined. A new algorithm based on Hough gradient is developed in order to detect object quickly. The algorithm is further accelerated through GPU parallel computing and implemented on NVIDIA TX2. The control of the system is realized with an onboard arm processor. The system is able to handle video steam of 75 frame per second (FPS), where the image resolution is $640 \times 480$ pixels.