Motion capture and estimation of dynamic properties for realistic tree animation

Abstract

The realistic animation of real-world trees is a challenging task because natural trees have various morphology and internal dynamic properties. In this paper, we present an approach to model and animate a specific tree by capturing the motion of its branches. We chose Kinect V2 to record both the RGB and depth of motion of branches with markers. To obtain the three-dimensional (3D) trajectory of branches, we used the mean-shift algorithm to track the markers from color images generated by projecting a textured point cloud onto the image plane, and then inversely mapped the tracking results in the image to 3D coordinates. Next, we performed a fast Fourier transform on the tracked 3D positions to estimate the dynamic properties (i.e., the natural frequency) of the branches. We constructed static tree models using a space colonization algorithm. Given the dynamic properties and static tree models, we demonstrated that our approach can produce realistic animation of trees in wind fields.