Improved Cue Fusion for Object Tracking Algorithm Based on Particle Filter

The traditional object tracking with cue fusion is inaccurate under complex background. Especially when some blocks exist, the targets may be lost. To solve this problem, improved cue fusion for object tracking algorithm based on particle filter is proposed. It uses color and motion as the observation information source. Color is the main observation information and motion is the auxiliary information. It weights particles followed by the order of information. Block detection, particle filter and mean-shift are used together to track the interest targets. The experimental results show that in complex scene, when the number of particles of the proposed method is half of the traditional cue fusion, the proposed method can improve effectively the accuracy of target tracking, and track object stably when the shape is changing. So the proposed method is more robust and real-time.