基于行为建模的人工鱼动画技术研究

【作者】 吴学龙；

【导师】 潘志庚；

【作者基本信息】 浙江大学 ， 计算机软件与理论， 2004， 硕士

【摘要】 智能虚拟环境的研究内容是在虚拟环境中加入有生命的对象—即虚拟主体，通过对他们活动的模拟来进一步增强虚拟环境的真实感和沉浸感，以更逼真地模拟真实世界。在计算机图形学中，大多数动画的创作是采用传统的、花费大量劳动的“关键帧”技术。近年来，越来越多的研究者开始研究虚拟主体的自主行为模型。 本文包括两大部分内容：虚拟海洋环境的建立和“人工鱼”(artificial fish)的行为建模。我们对虚拟海洋环境的建立、人工鱼的感知系统、人工鱼的自主行为模型、行为的层次结构进行了深入的研究，对传统算法做了改进。在此基础上，我们实现了一个虚拟的海洋环境，其中栖息着我们的“人工鱼”，每条人工鱼都是一个自主智能体。 在虚拟环境的建立中，我们采用了粒子系统模拟水；用基于图像相位变化的动画算法模拟水草在水中随水流的流动而摆动；用建立人工鱼的方法建立浮游生物；用粒子系统的思想模拟海中的水泡。 本文采用合成原始流的方法模拟水流。该方法定义一组原始流如均匀流、源点流、汇点流、漩涡流等，采用合成的方法合成复杂的水流。我们采用基于图像相位变化的动画算法，结合光流场两幅图像之间实现动画，针对算法对图像质量的损害，对图像后期处理，提高了动画的图像质量。用lαβ颜色空间把这种算法扩展到彩色图像。最后，把这种动画应用到水草模拟。 本文实现了虚拟视觉和虚拟记忆。虚拟视觉采用查询图形数据库的方法。我们的算法能够真实地模拟人工鱼视觉的视场范围，并具有遮挡判断等视觉功能，它同样也适合用到其他的虚拟主体中。我们还用队列的方法设计了一个虚拟记忆模型，能够对场景进行记忆。 在行为模型上，本文设计了一个“人工鱼”的自主行为模型，设计了人工鱼的基本行为集合，并根据行为分层理论，把行为组织成层次结构，在此基础上通过组合、包容等手段有效地支持了人工鱼的特定任务的完成。我们采用了抑制(inhibition)和疲劳(fatigue)模型进行行为选择。提出用多项式函数代替传统模型中用线性函数来更新内部状态，修正了传统模型中不符合动物习性的部分，使动物的行为更符合实际情况。 最后，本文设计了一个两条鱼相互寻找的具有童话色彩的情节。我们采用SmartObject技术建立交互模型。人工鱼根据内部状态和感知信息，以及从周围基于行为建模的人工鱼动画技术研究浙江大学硕士毕业论文环境的物体询问到的信息，能寻找目标鱼的位置，并具有碰撞检测及其他减少饥饿、减少疲劳等基本行为等。更多还原

【Abstract】 The content of research on Intelligent Environment (IVE) is to integrate virtual life, such as virtual agent into Virtual Environment (VE), and to simulate their activities to improve the reality and interactivity of the VE. In the area of Computer Graphics, most of the computer animations are made by traditional, time-consuming "Key Frame" technique. Recently, many researchers begin to study autonomous behavior model. In this model, the virtual agent chooses its behavior by itself according to its own inner states and the condition of the surrounding environment.This thesis focuses on two main topics: modeling of the Virtual Environment of the sea and Behavior Modeling of "artificial fish". We focus on the simulation of water and aquatic, the virtual perception of artificial fish, autonomous behavior model and the hierarchy of behaviors. We improve the traditional algorithm and based on this we develop a virtual sea environment enriched with our artificial fishes, and every artificial fish is an autonomous virtual agent.On the modeling of virtual environment, we use particle system to simulate water; algorithm of motion without movement to simulate swaying of aquatic in water; the method of modeling artificial fish to model plankton object and modeling virtual bleb with the thought of particle system.We propose the method of synthesizing flow primitives to simulate water. The idea is to define a group of flow primitives such as uniform flow, sinking flow, source flow, vertex flow and then synthesize these flow primitives into a complex flow. We present an algorithm of motion without movement to simulate swaying of aquatic in water, and improve the image quality of the motion aiming at the damage to the image quality. In addition, we extend the arithmetic to color image with lap color space. We use this algorithm to simulate aquatic by texture-mapping.We implement virtual vision and memory for artificial fishes. The virtual vision is obtained by the method of querying graphics database. Our algorithm can simulate the range and visibility of fishes’ vision, and it can be applied to other virtual agents. Also, we implement virtual memory by the method of queue.On behavior modeling, we implement an autonomous behavior model for artificial fish. We design a set of basic behaviors of artificial fish and arrange the behavior into hierarchy according to behavioral hierarchy theory. We achieve high-level tasks through the combination of the basic behaviors. We adopt inhibition and fatigue model to choose behaviors and propose to use nonlinear changing function to update animal’s inner state instead of traditional linear changing function. It modifies the part that does not fit the habit of animals, thus makes animal’s behaviors more natural.Finally, we design a fairy story about two fishes playing with each other. We construct an interacting model with Smart Object technology and use virtual memory to remember what happened. The artificial fish in the VE looks for the target according to current inner states, the information observed by virtual perception and the information got from objects around itself. In the looking course, it utilizes basic behavior such as obstacle avoiding, reducing hungry, reducing fatigue and so on.更多还原