Vision based object tracking under indoor environment

By integrating mean shift and particle filter, a new method for robotics visual tracking is proposed based on similarity distance. The method mainly depends on particle filter, and the particle number is proportional to the similarity distance. When the particle number exceeds the threshold, a small number of particles integrated with mean shift replace the traditional particle filter. After the tracking algorithm of each frame is finished, the similarity distance between the target and the model is calculated, which will adjust the number of the particles and determine whether the mean shift algorithm is called for the next frame. Moreover, the correlation coefficient between the particle number and the similarity distance is set through experiments. A frame tracking consumes about 5-10 ms. The experiment shows that our method can achieve accurate and real time tracking.