The Method Of Tree Reconstruction From A Single Image Based On Gaussian Mixture Model And Iteration Function System

Tree modeling is a hotspot in the research of virtual reality. Gene structure and environment of tree growth is rather complex, and this will lead to trivialities and complexity in the simulation of tree. Natural scene modeling is high dependent on real-time and fidelity. The traditional tree modeling is subject to some bottle-necks including low precision and slow rendering rate, which limits application of the tree modeling in the virtual scene. From this view, the method was proposed to generate the three-dimensional tree model based on single image and simulate the swaying tree in the wind.Firstly, the traditional image segmentations were subject to some defects including being disturbed easily, instability and low precision. So the method of image segmentation based on Gaussian Mixture Model was improved. The image pixels were classified from the view of the statistics, and then average value of image pixels was matched with Gaussian Mixture Model. Pixel type was confirmed depend on matching rule, and the image pixels were distributed to corresponding model. The low precision problem was solved from the angle of pixel field.Secondly, for solving the problem involving low-fidelity and interacting less in the adjustment of branches shape, the method based on Iteration Function System combined with image segmentation was presented. 3D trunk closed with the tree in the image was produced by classifying pixels. Using the self similarity of tree growth, the sub branches and leaves were derived from the trunk.Thirdly, swaying tree in the wind was simulated by using dynamics theory. For small amplitude of non-twig branches, branches distortion was calculated by state press analysis. For big amplitude of twig branches, branches distortion was calculated based on wave theory. Extent of swaying was controlled by panel to achieve the interaction.Finally, corresponding software system was designed by OpenGL graphics editor. 3D model of tree was constructed and feasibility of theories upwards was validated. Optimization was reached in fidelity and real-time. This method had been used in the Hi-Tech Research and Development Program (863 Program).