Flower Modeling Based On Single Image And Dynamic Simulation

In recent years, with the development and application of computer graphics and visual simulation technology, modeling and simulating plants has become one of the important research fields over the world. As an important branch of plant modeling,flower reconstruction is necessary for building virtual natural landscape. In addition,movement of flower adds realism and beauty to the natural scene. However, due to the characteristics of the structure of flower, flower reconstruction is more difficult and needs more professional knowledge. The complex model can also affect the real-time performance of flower dynamic simulation. In order to make more users to carry out flower modeling and meet the realistic and real-time requirements of natural scenes, this paper uses image-based method to finish 3D reconstruction of the flowers, and carries on the dynamic simulation to the constructed flower model. Based on the image-based modeling method and the plant behavior animation, we finish a flower modeling system and animation.Firstly, we analyze the demand of flower modeling and research the method of feature extraction of image. Aiming at the problem of background information interference, we utilize probability of areas we marked to optimize the result. When extract the edge of the flower,we only need to mark target area,then establish the multiresolution copy of the image and apply Canny edge detection on them. After combining the test results, the edge points in the target area are selected according to the probability of occurrence. By the smooth connection of these edge points, we obtain a closed contour with a width of 1 pixel, which is the characteristic information needed for the flower modeling system. Because flowers often occupy a small part of the image, only the target area marked by the user will be extracted, to avoid the calculation of the complete image and improve the computational efficiency. According to the multi-resolution edge extraction results to determine the final edge can filter the noise well to retain the main profile.Secondly, with regard to the problem that the single flower image is difficult to be calibrated, we transform the two-dimensional contour point to the three-dimensional point cloud by using the cone fitting. Considering the geometric shape features, simplify structure and dynamic calculation of the model, determining the structure of flower and generating a mesh model. Then, based on the conic surface and reference to the 2D contour to deform the model, which make the model similar to the image. On the basis of the traditional Laplacian mesh deformation, a restoration coefficient is introduced to overcome the problem of the scale of the grid model is obviously reduced after deformation. Based on the existing flower modeling process, this paper analyzes the demand targets of each stage, selects the relevant methods and improve them to get vivid visual effect. Finally, we obtain the 3D model with simple structure and floral morphological features.Finally, the flower model is applied to simulation of the natural scene. In this paper,we use the physical analysis method to simulate the motion, combined with the movement characteristics of the flower in reality. In addition to the simulation of the conventional flower shaking, we analysis and dynamically simulate the situation of petals falling to improve the realism of the scene. As a result of the structure of flower is simple and supports LOD expression, greatly simplifying the calculation, on the basis of ensuring the reality of the scene to improve the rendering efficiency of the scene.