Different graphic techniques and methods are usually presented to simulate the 3D plant shape structure and growth in the field of virtual plant and visualization. Research results had been widely used in agriculture, forestry, landscape design, game, VR and computer animation, especially in plant shape structure-function modeling and natural landscape reconstruction. At present, the shape, color, texture and other morphological structure characteristics of the plants are mainly simulated by using the algorithms of computer graphics and rendering technology. The L-system is used to describe the geometry shape of plant and has faced the challenge of simulating the growth for complicated plant structure such as flower development. In this study, we present a method to describe the flower plant shape modeling and growth deformation based-on the scanning 3D point cloud data to avoid the complex plant procedural modeling. At first, we obtained the flower plant point cloud data in different growth development stages by using the 3D scanning device, and the sample data required for the function fitting and deforming by segmenting and intercepting the point cloud model. The sample data was fitted to summarize the growth rules of flower plant by using the tools of MATLAB and specified functional model. Then we combined the traditional flower growth simulation with the free-form deformation algorithm (FFD) based on the point cloud model. The traditional method deformed the shape by dragging the control point on the bounding box. But we made the control point move automatically, where the outside control points were driven to deform according to the growth function. In order to realize the operation of spatial points in the free-form deformation algorithm, the control points were converted from the world coordinates to local coordinates. Regardless of the world coordinates of the control points how to change, there were fixed local coordinates of the points take part in the mathematic transform. To fit the growth model of flowers, a special sampling and fitting method was designed. First, the flower was divided into different growth patterns and sampled separately according to the shape structure, we could get the fitting function by using the sample data. Then the real-time dynamic process of plant growth was operated by combining the growth function and deformation algorithm. To simulate the dynamic growth of flowers, a method was proposed based on the real measurement data of point cloud data model, which combined the FFD algorithm with the true growth principles to simulate the series change of petal shape. The test results showed that our method could maintain the plant morphological characteristics during the deformation and reconstruct the real process of plant growth according to the initial and final point cloud data, which could get a realistic visual effects. At last, we compared our deformation method with the physics-based simulation method and evaluate the algorithm precision by fitting regression analysis and measuring the geometry error between the real measure data and the reconstructed data. The results showed that the geometry deformation error could be controlled within 2 mm and the coefficient of determination was above 0.75 during the whole period of plant growth. The final reconstructed plant model was similar with the real plant and a novel approach was provided to simulate plant growth. © 2019, Editorial Department of the Transactions of the Chinese Society of Agricultural Engineering. All right reserved.
