Error sensitive gaussian particle filter and its evaluation for contour tracking
Chen Peng, Hui Qian, Wang Wei, Zhu Miaoliang · 2009
We present a novel system for visual based contour tracking and apply it to track the movement of human's head. Currently available technology is used in the implementation to take care of low level video processing and acquisition. The implementation consists of two main parts. First, there is an error sensitive Gaussian particle filter. It estimates the posterior density of the shape being tracked by modeling a Gaussian distribution as a collection of weighted samples. The second part is the observer which takes an estimate of the current shape position, and uses image features to find a new spline. The new spline is projected onto a subspace of Euclidean similarities of a spline template. It successfully tracks the human's head in a video stream, despite the low-resolution and low frame rate of the video.