Smooth 3D Trajectory Segmentation for Flying Insects

We focus on the computation methods for analyzing the bumblebee behavior. In this paper, we recover the smooth 3D trajectory for flying bumblebee by stereo vision by employing EKF. We also tackle the matching problem when more than one targets appeared in the views. A variable length sliding window method is proposed to segment the 3D trajectory into three basic types. In our experiment, we show our method has more accuracy result than the state art of work.