An Adaptative Multi-agent System Approach for Image Segmentation
An Adaptative Multi-agent System Approach for Image Segmentation
复制标题
图像分割的自适应多智能体系统方法
DOI：
--
复制
发表时间：
2012
期刊：
影响因子：
--
通讯作者：
Amri Said
中科院分区：
文献类型：
--
作者：
R. Mohammed;Amri Said
本文提出了一种多智能体图像分割方法。多代理系统（MAS）是一种分布式系统，由一组代理组成，这些代理在它们能够感知的环境中与自己交互，并对其进行操作。提出的解决方案包括切割图像的空间，将其处理为一组子空间（图像的分区），其中创建了几个代理来检测物体的轮廓，然后跟踪它们（这些代理称为检测器-跟随代理）。这些智能体采用了一种非常有效的检测算法，并根据它们进化的区域的特征跟随轮廓。收集到的信息被传送到各级监督代理（代理分区），它们负责收集代理检测器-追随者发出的信息，更新包含分割参数的表，并制定代理检测器-追随者管理的全局策略（创建、销毁、以备用模式设置或初始化代理检测器-追随者）。在这个代理层次结构的最高层，我们找到整个系统的主管代理。通过使用Madkit系统实现这种方法，我们可以观察到性能和精度的提高，这是由于并行、并发和协作执行任务而非常重要的。
This article presents a multi-agent approach for the segmentation of images. A multi-agent system (MAS) is a distributed system consisting of a set of agents that interact with themselves in an environment they are able to perceive and on which they can act. The proposed solution consists in cutting the space of the image to treat it in a set of sub-spaces (partitions of the image) in which several agents are created to detect the outlines of objects then to follow them (these agents are called detector – followers agents). These agents adapt a very efficient algorithm of detection and follow the outline according to the characteristics of the region that they evolve in. The information so collected is transmitted to levels of supervision agents (agents partitions) which take care they with collecting the information emitted by the agents detector - followers, to update tables containing the parameters of segmentation and to elaborate global strategies of management of the agents detector - followers (creation, destruction, setting in a stand-by mode or initialization of agents detector-followers) . At the highest level of this agent’s hierarchy, we find the supervisor agent of this whole system. An implementation of this approach by the use of Madkit system allowed us to observe a gain in performances and in precision very important due to parallel, concurrent and cooperating execution of tasks.