基于人工免疫克隆选择算法的遥感图像识别

Remote Sensing Image Classification Using Clonal Selection Algorithm Based on AIS

遥感图像分类受自然环境的复杂性和实际样本数据的分布情况约束,其结果将直接影响对土地利用、覆盖情况的认知。人工免疫系统具有自学习、自组织、记忆的能力,可解决非线性分类问题中的局部极值、鲁棒性等难点。该文将人工免疫系统引入遥感图像分类领域,设计了基于克隆选择算法的遥感图像监督分类方法,并将其应用于广州市遥感影像分类。实验结果表明:与最大似然法相比,该方法具有更高的精度;同时该方法对公路、桥梁等线状城市用地较敏感,适用于快速发展的中心城市的遥感影像分类。

Limited by complexity of environment and distribution of specimen data, remote sensing image classification is an important reference in recognition for land use and cover. By possessing such biological properties as self--study, self--organizing and memory, artificial immune system （AIS） can resolve problems in non--linear classification such as local optimum, robustness, and so on. Thus, AIS method is employed for remote sensing image classification in this paper, where clonal selection algorithm is used to build a supervised classification method. The application of this method in Guangzhou shows that AIS classi- fication has higher precision than conventional maximum likelihood classification, and AIS classification has good applicability to remote sensing image classification in central city because it is sensitive to such slender linear urban land as road, bridge, etc.