The Research On Extraction From Leaf Image And Visual Realization Of Virtual Plant

The virtual plant is a computer simulation of the structural growth and development of plants in 3-D spaces. The study on visual techniques of the virtual plant is a main part of the researches of the virtual agriculture, which has potential applications ranging from computer-assisted landscape, education, games, and virtual garden to digital agriculture such as the study and management of plant diseases and plant bleeding as well. In contrast to the extensive studies on modeling plant architecture, the researches on modeling of plant organs still remain unexplored. So the study of image-based reconstruction of plants organs has its a great practical value. Leaf is one of the most important visual features and offers source information for constructing the picture warehouse of plant organs. So the extraction from leaf image and its reconstruction are a key step of rule-based modeling of virtual plants.The dissertation carries out the research on the extraction from leaf image and visual implement of virtual plant, but focus mainly on extraction from leaf image, image-based leaf reconstruction, L-system and image-based visual plant development. The main study and contributions in this dissertation are given as followings:Firstly, it states the basic concepts of plant models and compares different methods of modeling virtual plants; then it provides the way to achieve the visual development of image-based virtual plants. This idea is based on image to extract the visual features of plants and further use L-system to"reorganize"the growth of plants.Secondly, it analyzes the rule-based reconstruction of virtual plant organs which only focuses on using simple geometric figures to approximately denote plant organs. Although this model has a controllable quality through some parameter adjustings and computer graphics techniques, it lacks high visual effect. Leaf is one of the most typical plant organs and offers the important features of plants. Extraction from leaf image is the basic step to reconstruct the leaf surface and automatically identify plants. This dissertation designs the extraction of leaf contour. However it is not sufficient to reconstruct leaf only using leaf contour. The leaf vein, as one of the most important visual characteristics of leaf, should be considered. So the dissertation brings forward the extraction algorithm of leaf vein combining snakes technique with cellular neural networks (CNN). Due to its high flexibility and control of the contour dynamics of the snakes, this algorithm can improve the precise and robustness of the segmentation. The...