Athletes Tracking using Homography Method: a Preliminary Study

Abstract

Particle tracking has been used widely to track a single particle motion or trajectory in a medium. One of the applications of the particle tracking is in sports analysis. The tracking method can be divided into, the wearable device based system and the image based system. The wearable device based system utilizing global (GPS) and local (LPM) positioning system to track player movement. The image based system use video image processing to track the player movement and recorded is frames basis. However, the image processing method in a football match requires correction as it is normally recorded from the side of the field. Thus, to solve this problem a set of mathematical solution is needed to convert the image coordinate system (pixels) to the actual coordinate system (meters). The most commonly used is the homography method. The technique requires at least 4 reference points to transform the image coordinate into the actual coordinate system. In this project, a futsal game was recorded. The image coordinate of the player were marked in each frame with respect to the time. The image coordinate data were converted into the actual coordinate using homography matrix. Comparisons were made between the homography technique method and open-source available image warp processing method for validation. Based on the result, the homography coordinate transformation system produce a good agreement with actual player activity on the field.