Data Fusion for Robust Head Tracking by Particles
Yonggang Jin, Farzin Mokhtarian · 2006
The paper presents a data fusion panicle filter for robust head tracking in video surveillance applications. With head detection based on moving region contour analysis, we propose a data fusion particle filter to fuse head detection results with colour and edge cues for robust head tracking. Connections of the proposed particle filter with previous work are also discussed where proposal distributions of M. Isard and A. Blake (1998) and P.Perez et al., (2004) are shown to be an approximation with fixed ratio of importance and prior samples. Experimental results demonstrate the robustness of head tracking using proposed data fusion particle filter.