Research And Implementation Of Group Animation Based On Evolutionary Computation

With the rapid development of computer graphics, 3D animation has already been applied into every aspect of our lives, especially in the films, games, entertainment industry and etc. As the wide use of animation technology, it is very hard for the ordinary 3D animation to satisfy the audience nowadays, whereas the group animation as a new idea attracts more and more attention and becomes a hot question in the computer graphics. Recently the simulation of group behavior is used in films, games, virtual reality, disaster prevention, transportation planning, etc. This paper focuses on the human model and group animation.As model uniqueness and action intelligence become significant in the animation research, along with the coupling degree between each module of animation system improves, and the difficulty of group animation increases, the limit of traditional group animation method is obvious: the first drawback is animators need design every model from sketch with great imagination and creativity, which cost long time for single modeling with low efficiency. The second one is, in traditional animation techniques (i.e. key frame), all motion details are controlled by animators. With the longer time of animation, the larger number of model and the higher degree of complexity, animators have to pay more effort. The last one is the character built by key frame method is lack of autonomy. If the circumstance is changed, the whole animation needs to remake. The character cannot adapt with the surroundings which reduces the modification and interactive of animation.This thesis proposes a new approach to improve the traditional group animation with evolutionary computation. First genetic algorithm is applied to character modeling in group animation. Secondly, a character walking parameter plug-in is designed. Finally an improved PSO algorithm is proposed and used in group path planning to simulate group behavior, which reduces the workload and increases efficiency. The main work of this thesis is as follows:1. Genetic algorithm is applied to character modeling in group animation.Usually there are two modeling methods: one is 3D modeling software like Maya, the other is using 3D- scanner, both of which have high price but low efficiency. This thesis takes Maya model as a prototype to segment the model surface. Then each component is evolved by GA with the fitness function designed in advance. Diverse character modeling is generated by selection, crossover and mutation operators, therefore designers get inspired.2. A character walking parameter plug-in is designed.This paper analyzes the key frame of character walking and designs a plug-in using parameter such as walk distance, and lift high. Users can adjust the size and location of bones, rotation of joints and etc. This plug-in is general to different character models.3. An improved PSO algorithm is proposed.PSO is a swarm intelligence algorithm by the research of group behavior. The methods of neighbor choose influences algorithm performance. This thesis proposes an improved PSO algorithm by adding vision range parameters for each particle and taking the particles in its vision as neighbors. The experiment indicates that the improved PSO can enhance the accuracy of convergence and the ability of global search.4. A group path planning plug-in is implemented.PSO is swarm intelligence based on the birds'movement in nature. By using PSO algorithm in path planning, the path has intelligence and can simulate the group behavior in nature. The group path planning plug-in can simulate crowd gathering, leader following and obstacle avoidance. The experiment shows good simulation effect.