海上丝路沿岸多云多雨地区实现大比例尺制图的遥感分类方法研究

The core of the Marine Silk Road is the channel value and strategic security. The basic geographic information along the silk road is vital for assuring safe shipping lanes. Remote sensing technology has become a fast, efficient method to gain basic geographic information of large scale in coastal zone. But the areas along the silk road mostly locate in the tropics, its cloudy and rainy weather makes the acquisition of high resolution optical image difficult..This project intends to use high resolution full-polarimetric SAR and medium resolution optical image together instead of high resolution optical image to carry out the classification of this area. Main contents are included as follows: 1)Analyzing the original polarization characteristics, textural features, polarization component characteristics of ground object via full polarization SAR, and then evaluating the feasibility of replacing high optical image; 2)Introducing spectral signature of medium resolution optical image to discuss the feasibility of joint classification according to the ground object of lacking SAR response characteristics; 3)analyzing the supplement ability to the classification by single phase image based on with multi-temporal features; 4)Building classifier facing characteristics of multi-source heterogeneous based on multinomial logistic regression algorithm and carrying out the classification experiment and validation for the best classification scheme of large scale mapping in this area. The project result can serve navigation channel security of the silk road, and offer technical support for economic development and preventing disasters and reducing damages of the countries along the marine silk road.

海上丝绸之路的核心是通道价值和战略安全，掌握丝路沿岸的基础地理信息对于保障运输通道安全具有重要意义。遥感技术已成为快速、高效获取海岸带大比例尺基础地理信息的重要手段，但丝路沿岸大多区域处于热带，其多云多雨天气导致高分光学影像获取困难。本项目拟利用高分全极化SAR，并结合中分光学影像开展该地区的地物分类，从而代替高分光学影像，主要内容：开展高分全极化SAR的地物原始极化特征、纹理特征和极化分量特征分析，评价其替代高分光学影像的可行性；引入中分光学影像的光谱特征，针对SAR特征响应不足的地物探讨开展联合分类的可行性；基于不同物候的影像时相特征，分析其对单时相影像地物分类的补充能力；基于多项式逻辑回归算法构建面向多源异构特征的分类器，并开展分类实验与验证，遴选出面向该区域大比例尺制图的最优地物分类方案。项目成果可服务于丝路航行通道安全保障，并为沿岸国家海岸带经济发展与防灾减灾提供技术支持。

海上丝绸之路的核心是通道价值和战略安全，掌握丝路沿岸的基础地理信息对于保障运输通道安全具有重要意义。遥感技术已成为快速、高效获取海岸带大比例尺基础地理信息的重要手段。本项目利用高分全极化SAR，并结合中分光学影像开展该地区的地物分类，从而代替高分光学影像，主要内容：开展高分全极化SAR的地物原始极化特征、纹理特征和极化分量特征分析，发展了一种适用于海岸带全极化SAR多特征的两层分类方法，构建了面向高分SAR地物分类的卷积神经网络模型，评价其替代高分光学影像的可行性；引入中分光学影像的光谱特征，提出了顾及极化特征的全极化SAR与中分光学影像融合方法，发展了基于特征组合的全极化SAR与中分光学联合分类方法，其替代高分光学影像开展地物分类是可行的；分别开展了光学和SAR多时相数据的特征显著性分析，证实了其对单时相影像地物分类的补充能力及有效性，构建了基于多元核逻辑回归的多源多时相遥感分类方法，发展了基于多维特征时间序列相似性度量的分类方法。本项目发展的相关技术为多云多雨地区的高分辨率遥感监测提供了保障，可为沿岸国家海岸带经济发展与防灾减灾提供技术支持。