摘要

An improved particle filter algorithm was presented to track a moving target under natural environment, in which the color histogram was integrated with the measurement model, the latest observations was taken into account and the system model is the second order autoregressive process. The proposed method can deal with rotation and scale transformation, variations in the light source and partial occlusions, and the tracked target can be rigid or non-rigid. The demonstration examples showed that the method can track the target with robustness in real-time.