Video tracking of human-faces with occlusion using enhanced Camshift approach

An enhanced Camshift approach in conjunction with Kalman filter is presented aiming at tracking the multi-human-faces with occlusion in the moving course. Starting with detection of facial region of interest, the proposed tactic induces overlap coefficient into determining whether or not the target occlusion occurs in the course of motion. The matching matrix is subsequently set up based on the target's dynamic model, solving the likely mismatching problem arising from posteriority of occlusion. The algorithm is put to test suggesting adequately promising occlusion-resistance capacity and practicality.