中文摘要：

在介绍了CognitiveAgent（CogAg）来源和特点的基础上，阐述CogAg图像分割的原理和方法。这种方法中供CogAg搜索的环境是数字图像，这个环境（又称Artifact）中每个像元为CogAg提供用于分割的4个特征信息：局部平均灰度、局部标准差、局部能量和局部分形维，这些信息提供CogAg思考、推理、决策选择同质像元。经实际航空影像分割的实验和对比，CogAg图像分割方法是很有应用前景的方法。

英文摘要：

A new method, Cognitive Agent for image segmentation, is proposed. Then, the basic principle and the main idea of this method are introduced. There are four kinds of pheromones,i, e. pixel＇s gray local mean value, pixel＇s local standard deviation, pixel＇s local energy and pixel＇s local fractal value; which are extracted in neighbour pixels of each pixel and then regarded as the criterion of thinking, reasoning and decision-making of Cognitive Agent for selecting homogeneous pixels. The method is tested on three aerial images; and the experiments show that this proposed method is reasonable and promising in the future.