一种基于结构张量的MAS边缘检测算法  

Novel MAS Algorithm Based on Tensor for Edge Detection

作　　者：陈亮[1] 郭雷[1] 王雅萍[1] 杜亚勤[1] 

机构地区：[1]西北工业大学自动化学院,西安710072

出　　处：《计算机科学》2009年第1期131-133,152,共4页Computer Science

基　　金：国家自然科学基金项目(60675015)资助

摘　　要：MAS(Multi-Agent System,多智能体系统)是一种群体智能系统,具备并行计算的能力,由多个运行机制简单易于实现的个体Agent构成。在边缘检测过程中,根据结构张量的特征值属性表示边缘特征,提出一种可行的MAS实现架构,扩展了MAS的应用。该算法根据个体Agent具有的简单状态机制,结合边缘位置的结构张量的特征值的外在表现,让每个Agent采用指定的策略进行边缘探测和标记,其中包括扩大范围、位置平移和突变,最终通过多Agent的共同努力,检测出图像中的边缘。实验表明,得到的边缘位置更为精确,并对噪声的影响不敏感。With the parallel and distributed mechanism, as well as the simple and easy implement of each agent, MAS was employed in edge detection with the feature being defined by the eigenvalue of the strucutre tensor. A novel reactive MAS framework was proposed which is a extension to its application. A finite status machine was presented. Based on it,the agent explores the interesting feature and employs some strategies to seeking and propagating which includes scope enlarging, shifting and mutation of the position. The edge was detected through working together of all agents. The experiment was designed to improve the effect of the proposed appraoch. It shows that the position of the result edge is more accurate and the approach is insensitivity to the noise.

关 键 词：多智能体系统 结构张量 有限状态机 边缘检测 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]