Optical flow-based observation models for particle filter tracking

摘要: This paper presents three observation models suitable for particle filter tracking, based on the optical flow of sequence. Modern computation techniques can obtain in real time very accurate estimates, so we use it as a source evidence higher level image processing. Our motion-based are based, respectively, on: previously computed field, brightness constraint, and similarity measures. They take into account not only consistency measured with motion predicted by model, but also presence discontinuities object boundary. Experimental results show that resulting trackers comparable to other, state-of-the-art tracking methods. While model measures provides better performance, flow-field-based is option when field available.