Multi-feature target tracking based on lifting scheme

Both color feature and edge feature have been researched in target tracking field, but these two features have been studied separately in target tracking field. Use single feature make the tracking system instability sometimes. In order to enhance robustness of real-time target tracking system under complex conditions such as scenes changes, occlusion and background confusion, this paper use two mutual complementary feature track target. This paper studies extract the edge feature (high-frequency parts) and the color feature (low-frequency parts) of the video images in a certain moment by using lifting scheme. Then, to the edge orientation histogram (EOH) of edge feature and the color feature of low-frequency part use mean-shift (MS) algorithm combining with Kalman Filter (KF) algorithm parallel track the target. Finally, gets the ultimate state of the tracking target by fusing the two tracking results. This method has a good balance between real-time and robustness. After experimentation this method has such advantages as not sensitive to target´s luminance, occlusion, shape and scale, stabilization and efficiency.