Occlusion Handling Based on Particle Filter in Surveillance System

Object tracking with occlusion handling is a challenging problem in intelligent video surveillance system. Among various tracking algorithms, particle filter (PF) is a robust and accurate one for different applications. In this paper, a new approach based on particle filter is presented for tracking object accurately and steadily when the target encountering occlusion in video sequences. First, the object pixels are classified as foreground and background for each frame using background subtraction. Our approach combines the foreground region with the particle initialization and similarity measure step to lower the background distraction. Second, a set of cues including a motion estimation model, an elliptical shape model, a spatial-color mixture of Gaussians appearance model, and an edge orientation histogram (EOH) model is fused and modeled by a data likelihood function. Then, a particle filter algorithm is used for tracking and the particles are weighted and re-sampled based on the fusion of the cues. Results from simulations and experiments with real video sequences show the effectiveness and robustness of our approach for tracking people under occlusion conditions.