Tracking Trajectory of 3D Trees Moving Based on Video Data Driven

Abstract

Tracking trajectory of three-dimensional trees is a difficult problem in computer animation and virtual reality. It requires not only high sense of reality for the morphology of trees and tree moving, but also adequate real-time. In this paper, we present a simulation method based on video data driven. Firstly, split out the main branches and leaves of trees from video images by using hybrid method of edge detection and K-means clustering based on color. And then, construct wind-speed sequence of trees in the video by using feature matching algorithm through analyzing the model of trees trunks and set of cloud point of leaves. Finally, by classifying and tracking video sequences with statistical models, we can get trajectory of trees swaying. Experimental results show that only a small part of the motion tracking in the algorithm is calculated in real-time way, the rest are calculated in pretreatment. So this model greatly improves the real-time effect of computing the trajectory of trees moving while keep a good sense of reality at the same time.