Modeling and Animation of Real-world Trees (Introductory Talk)

Trees play a significant role in enriching the beauty of nature. In order to generate realistic models of specific trees, we often start from a set of images or point clouds. Modeling trees from images has the advantage of easily data acquiring and lower cost. Previous work has been done to reconstruct outdoor trees from more than 10 images with the coverage of wide angles around the tree. We present an interactive method to reconstruct trees from sparse images with narrow sampling angles.

The other popular way is to model real-world trees from point clouds directly captured by laser scanning device. Airborne LiDAR scanner has the advantage of covering the big picture of a large-scale scene. However, it is difficult to capture the details of tree branches than ground LiDAR. We propose an efficient method to create natural-looking tree branch structures by a constraint-based greedy algorithm that approximate the input sparse point clouds in a voxel space.

There has been much research on animation of trees, but limited work on considering the movements of real-world trees. Based on the observation of Lissajous curve from the pull-and-release test of trees, we choose three basic mode shapes from the modal analysis of a curved beam, and combine them with a driven harmonic oscillator to approximate the phenomenon. Furthermore, we measure the natural frequencies and damping ratios of real-world trees from videos by object tracking, and use these parameters to guide physically-based tree model, and hence more realistic animation could be achieved.