震害影像分类在震害评估中起着关键作用。受地震复杂环境的影响，影像分类在地震场景描述和分类器设计及优化两个关键问题上面临巨大挑战。目前的分类算法在场景描述方面，信息利用不足；在分类器方面，参数依靠经验确定，智能化程度不高，且容易陷入局部最优。引力智能优化算法（GSA）优越的智能全局寻优性能，结合多源遥感影像，可望解决上述问题。针对这一现状，本项目利用多源遥感影像，在深入分析GSA的基础上，对上述问题建立优化模型，提出以下研究内容：（1）研究震害目标多源特征提取、组合和优化理论，构建特征空间，建立场景描述模型；（2）进一步研究震害目标在特征空间的聚类规律，设计并优化分类器。最终建立一套复杂地震场景描述及分类的智能优化模型和算法，解决震害影像分类面临的问题，提高分类精度。作为遥感信息领域一个重要的基础性研究，本项目不仅对自然灾害评估具有重要价值，对遥感信息地学认知和智能处理也具有重要的科学意义。

Classification of remotely sensed image of earthquake disasters plays a critical role in disaster evaluation. The classification is facing big challenges in two key issues because of complex seismic environment. One is the description of earthquake scene and the other is the design and optimization of classifier. Some serious problems appear in existing classification algorithms. In the description of earthquake scene, information use is too simplicity. In the design and optimization of classifier, parameters highly depend on experience, intelligent degree is low, and optimum fall into local easily. Gravitational Search Algorithm (GSA) is known with strong intelligence ability of global optimization in feature space, combined GSA with multi-source remote sensing image is expected to solve the above problem. Therefore, based on in-depth analysis of GSA and multi-source remote sensing image, our project establishes the optimization model for the above issues and proposes the following research objectives: (1) Study the theory of feature detection, combination and optimization with GSA to provide the feature space in seismic multi-source remote sensing image, and establish the earthquake scene description model. (2) Study the land-surface objects clustering rules in the feature space, and design or optimize classifiers in the feature space with GSA. An intelligent optimization model and algorithm will be finally established for complex scene description and classification, all of which will fundamentally improve the accuracy of classification of the remotely sensed image of earthquake disasters. As an important basic research in remote sensing, this project not only provides strong theoretical and technical support for natural disaster evaluation, but also offers important scientific significance to the cognition of Geo-science and intelligent processing based on remote sensing information.