Three-dimensional Dynamic Data Capturing And Data-driven Animation Of Trees

Realistic rendering and animating vegetation still pose a significant problem due to the inherent complexity of plants.Recently,Data-driven tree animation has been developed to simulate plant motion with data captured from nature tree motion.However,withexpensive motion capture equipment,complex process,virtual tree model,and limited form of simulated tree movement,the promotion of Data-driven based plant animation was restricted.This thesis presents a new scheme to capture tree motion in nature scene and implements Data-driven tree animation based on modal analysis.The main contributions of this thesis are as follows:(1)An experimentplatform to capture tree data with KinectV2 which connected to a PC workstation and erected in a tripodwas presented.Compared to Passive Optical System,this method can capture the image stream and depth stream of outdoor trees in 30 frames per second.(2)To analyze the vibration frequency and damping ratio of branch and stem from captured data of tree with Kinect V2,we firstly tracked the single colored mark to extract the trajectory of each mark in plane.According to the tracking result,a local relative deflection angle criterion which calculates the local relative deflection angle between parent and child branches was proposed.Then transformingdeflection angle to frequency domain by Fast Fourier Transformation,the tree vibration frequecy was setas frequency coresponding to the first-order vibration peak.The damping-ratio determined by half-power bandwith method after extracting thefirst-order tree vibration frequency.The experiment result showed that the vibration frequency of the tree was the inherent characteristic of the tree and does not change with the change of the external force.The damping of the tree during the movement was obviously related to the external force and the tree branch structure.(3)To provide a realistic tree model for Data-driven tree animation,anew application ofSpace Colonization Algorithm in reconstructing the tree model from point cloud was implemented.By combining the Space Colonization Algorithm with a proposed skeleton decimation algorithm,a realistic tree model was reconstructed,and itpreserves the features of main trunk and canopy of original point clouds.(4)Based on the extracted dynamic parameters and the static tree model reconstructed from point cloud,a modal analysis based Data-driven tree animation engine was implemented.First of all,the wind velocity field was constructed by the fBm noise and the wind force was calculated based on the aerodynamic.After that,we solved equation of motion of the three dominant modal shapes by separating it into space-time equations with vibration and damping ratiointegrated in it.Finally,by integrating the movement of the curved branches,an efficient and realistic tree animation was achieved.Based on the abovementioned Data-driven animation schema,we can produce visually plausible and scalable animations of tree under wind at interactive frame rates.