Multi-target tracking based on multi-feature associated matrix reasoning

In the video-based multi-target tracking, the uncertainty of the target moving states is the difficulty of multi-target tracking. At present, data association algorithm is commonly used to resolve this problem. Aiming at multi-target state judgment and occlusion problem, this paper proposes a multi-target tracking algorithm based on multi-feature associated matrix reasoning. This algorithm first applies the fusion of multiple visual features for target modeling. These features are complementary and redundancy to each other. And then deduce multi-target states through establishing multi-feature associated matrix. Finally judge and deal with concrete occlusions. Compared with traditional data association algorithms, this algorithm is more stable and real-time. The experimental results show that this algorithm can realize robust multi-target tracking in different states, and resolve the occlusion problem effectively.