Real-time 3D Plant Structure Modeling by L-System with Actual Measurement Parameters Rawin Viruchpintu and Noppadon Khiripet National Electronics and Computer Technology Center, Pathumthani 12120, Thailand. rawin.viruchpintu@nectec.or.th, khirin@nectec.or.th Abstract Plant modeling is the process combining biological knowledge, mathematical formalisms and computer graphic to generate virtual plants. Recently, the most popular technique used to model plants is the Lindenmayer system or L-system. This technique presents self-similarity patterns of a tree by its 3D turtle grammar. While the resulting rendering of the traditional L-system may look like a real tree, no parameters of the modeled tree are taken into consideration. In this study, the correct parameters such as the angle between branches, the length ratio of branches and phyllotactic patterns were integrated into our customized L-system interpreter. The 3D interaction and simulation of this virtual tree were developed using java3D technology. The final 3D model can also be exported to other sophisticated 3D modeling software via the Wave-front OBJ file format. Our approach can be easily extended and applied to other 3D modeling problems requiring more faithful results. 1. Introduction Modeling objects with mathematic and computational tools is always a crucial step in understanding the real process and situation. In biological domain, such as plant modeling, this requires biological knowledge of plant physiology along with mathematical formalism and computer graphic. There has been an active research area in modeling plant using Lindenmayer System [1], which simplify the plant shape as a 3D object. The object is defined by a self-similarity grammar. Many existing tools nowadays can facilitate the generating of grammar to represent any particular plants. For example, L-System Generation Program (LSystem4) [2] is a 3D visualization tool which offer the enhance options for viewing 3D model such as rotation and movement control, texture mapping per layers and visible assignment of textures to layers. J. Scott Cameron has shown the wonderful 3D L-Systems application [3] to model the growth of trees. His L- system was nicely implemented by the use of Microsoft's Direct3D API to show the skeleton of the tree structure. LYNDYHOP [4] is a simple Java-Application for generating self-similar fractals in 2D graphics in step by step. The program also provides a graphical representation of the rules, which helps to understand the growth of the pattern. Another application is the Floradig software [5], which can record 3D coordinates and then convert them into geometric properties such as the internode length, and the angle between the main stem and branches. It also uses linear and non linear regressions to fit the data and provides geometrical attributes for the plant model. However, we believe that what is missing from traditional L-System modeling approach is the sense of what is really going on in nature. In stead of doing trials and errors in adjusting L-System parameters and inventing grammars until the resulting 3D plant object look like the actual plant, can this actual parameters taken from the plant and integrated into a simple grammar? In doing so, the final output object should represent not only the plant but also how much we learn from the nature in creating the plant. Furthermore, the tool should allow real-time user interaction to help visualize the 3D object. The proposed FractaL-Tree system consists of four modules as depicted in Fig.1. Starting with L-system generator, the axiom and rules were conducted to new string as defined by L-systems concept. Next, the coordinates and directions of model will be generated by Turtle interpreter module follow by decorating the rough model by