Human Motion Tracking Based on Unscented Kalman Filter in Sports Domain
Guojun Liu, Xianglong Tang · InTech eBooks · 2010
In this paper, we propose a novel computer vision system for tracking high-speed non-rigid skaters over a large playing area in short track speed skating competitions. Several important features distinguish the proposed approach from others: 1. Introducing the reference frames as a transition through which each frame can be mapped to the field model to reduce the error accumulation of the projection, and it's very important for a long video sequence and helpful for improving the precision of the system. 2. Incorporating the hierarchical model based on the contextual knowledge and multiple cues into the unscented Kalman filter to improve the tracking performance when occlusions occur. 3. Choosing the unscented Kalman filter for visual tracking in the sports domain, it is superior to EKF in theory, and is more efficient than particle filter. 4. Evaluating the relation between the accuracy of homography H t -1,t and tracking performance. Proposing a novel and objective evaluation method to measure the precision of our practical system. However, The main problem is remained in our current system: how to improve the tracking performance when skaters are moving in groups during a long and continuously full occlusion. In future, we can model the skater's uniform of different teams in each sub-region, and the uniform model can be used to assist in tracking the target occluded for a long time. 5.