不透水面是评估城市生态环境的重要指标，利用多源高分数据对其进行精细分类是新型城镇化建设的必然需求。但是城市地表环境复杂，高分光学影像阴影问题突出，传统不透水面提取算法应用于多源高分影像时，在两个关键问题上存在不足：其一是缺乏对不透水面成像机理的研究，难以充分对其进行建模表达；其二是分类器设计独立于表达模型，难以融合多源特征。本项目通过对高分光学影像阴影特征的研究，建立阴影分层模型；然后针对不同的分层，在深入分析不透水面成像机理的基础上，引入引力智能优化（GSA）算法，构建不透水面特征提取和优化方法，建立表达模型；最后基于表达模型构建分类器优化和融合算法。本项目将充分利用GSA算法优越的全局寻优性能，建立一套完整的分层优化提取理论体系，通过层层优化控制误差，解决不透水面提取面临的问题。本项目将突破目前高精度不透水面提取理论体系缺乏的难题，推动遥感数据智能处理的发展，具有重要的理论和现实意义。

Impervious surface is a significant indicator for evaluation of urban ecological environmental quality. Multi-source remotely sensed high-resolution images based fine impervious surface classification is an inevitable demand of new-type urbanization construction. However, the complex urban land types and significant shadow issues make traditional impervious surface extraction algorithms challenging in multi-source high-resolution remote-sensed images. For one thing, existing imperious surface extraction algorithms are short of research on impervious surface imaging mechanism, leading to the insufficient expression of modeling. For another, multi-source features fusion is difficult for traditional impervious surface extraction algorithms due to the fact that their classifier design is independent of the expression model. In this project, we firstly establish the stratified model for shadow by analyzing the shadow feature in high-resolution optical images. Then, in each layer, by deeply studying the impervious surface imaging mechanism and introducing the gravitational search algorithm (GSA), we construct the feature extraction and optimization methods of impervious surface and establish expression model. Finally, the optimization and fusion algorithm will be constructed on the basis of the expression model. In this project, the global optimization ability of GSA is efficiently utilized to establish a complete stratified optimization extraction theory system, in which the performance of the method is optimized in each layer. Therefore, the proposed method in this project can meet the challenges of impervious surface extraction in high-resolution remotely sensed mages. This project will make breakthroughs in the lack of theory system of current precise impervious surface extraction researches, promoting the development of intelligent remote sensing data processing, which is of great theoretical significance and practical value.