Research On Leaves Based On Graphics And Image

With the improvement of computing power and graphics processing,various research directions in the field of computer graphics have been fully developed.Plant modeling,as one of the important research directions in the field of computer graphics,has been extensively involved in agricultural research,film and television animation,game development,garden design,cloud tourism and other fields,and people have higher requirements for the visual sense of virtual plants.According to the different types of target plants,different plant modeling methods are proposed for virtual plant modeling,but most of the methods are mainly aimed at the overall shape modeling of plants.Leaves are an important part of plants,and most of the existing modeling methods are based on triangular mesh and texture mapping.This method needs to pay a lot of cost in time and space,but also needs to be improved in the sense of reality and universality.Therefore,this thesis combined the front and back leaf images and traditional graphic methods to simulate the three-dimensional morphology of leaves in real environment and the structure of netted veins.The research content of this thesis is as follows:(1)The 3D point cloud is obtained based on image matching.Due to the difference in color and spots of front and back leaves,which are easily affected by the environment.The existing traditional image registration algorithms have not achieved satisfactory results in the front and back leaf images.To solve the above problems,this thesis firstly preprocesses the acquired images such as background removal and contour rendering.Secondly,the corresponding feature points are extracted from the pre-processed leaf image according to its contour features,and the optimal parameter matrix is calculated based on the corresponding feature points.The image of the front leaf is taken as reference,and the back leaf is transformed to complete the rough registration of the leaf.Then,in order to further accurately obtain the registration results of leaves,interpolation processing is carried out on the edge of leaves,so that the pixels of the front and back leaf images are consistent.Finally,the checkerboard Mosaic method and Root of Mean Square Error method are used to evaluate the transformed front and back leaf images,and the matched images are converted into 3D point clouds to obtain the ordered point cloud sequence of leaves.(2)The leaf deformation based on 3D point cloud.According to the above point cloud data,the leaf morphology changes in 3D under the guidance of hand-painting.Firstly,the target deformation curves of different two-dimensional planes are drawn with the mouse and fitted into the 3D space.Secondly,the region of leaf deformation is selected,and the deformation curve is one-to-one corresponding to the leaves in 3D point cloud space.Then,the adjacent regions of the deformation region are fitted with functions to ensure the integrity and authenticity of the leaves.Finally,the initial point cloud sequence of the leaves is moved to the target direction based on the criterion of Euclidean distance invariant to complete the leaf deformation in 3D.In this thesis,the color,texture and other details of leaves under 3D point cloud are all from the original leaf image,and the initial point cloud data obtained by this method is an ordered 3D point cloud sequence from the image,which provides great convenience for subsequent data processing.Therefore,this thesis has improved the realism and speed of 3D leaf morphology simulation.(3)The simulation of reticular vein based on water transport.With the distribution of veins conforming to the golden ratio,LS grammar satisfying the golden section is designed to draw the primary and secondary veins of leaves.Secondly,combined with the characteristics of vein transportation and water transportation,the branch points of tertiary vein and higher veins are randomly selected within a certain step length,and the growth direction of vein is initialized.The subsequent growth path drawing is based on certain randomness principles,so that it is more consistent with the characteristics of random vein growth.Finally,the corresponding solutions to the problem of meeting and edge overflow in the subsequent growth process are proposed.The experimental results show that this method can simulate the structure of reticular veins in natural.(4)Based on the above research content,the prototype system of leaf modeling based on graphics and images is designed and implemented.The prototype system is based on the thought criterion of modularization,which is divided into four functional modules: image preprocessing,image registration,3D leaf deformation and reticulated vein simulation.With simple and intuitive interface design,the three-dimensional modeling of plant leaves can be realized.