水土保持生产建设项目扰动遥感分析  

Studies on remote-sensing disturbance analysis for soil and water conservation projects

作　　者：孙涛[1,2] 范福新 梁藉 闫宝伟[3] 付文博 

机构地区：[1]中国水利水电科学研究院,北京100038 [2]水利部防洪抗旱减灾工程技术研究中心,北京100038 [3]华中科技大学水电与数字化工程学院,武汉430070

出　　处：《中国水利》2017年第20期38-40,12,共4页China Water Resources

基　　金：北京市科技计划课题"北京市水土保持遥感技术应用研究与示范"(Z161100001116102);水体污染控制与治理科技重大专项-北京市智慧水务综合管理平台研究与建设项目(2017ZX07104)

摘　　要：生产建设项目遥感监督管理往往需要通过大量遥感图像人工标注训练样本,进行深度学习模型的训练。为此,通过对大量遥感影像进行统计观察,结合实际建立了包括裸地、水泥地、森林等14种基本地类解译标志库,并以此为基准制作训练样本。在此基础上,详细分析生产建设项目的要素组成,探究了机场工程、涉水交通工程生产建设项目的地类组成规律,为后期深度学习模型的构建提供可靠的约束条件与决策支撑。In -depth learning process is necessary for management of remote -sensing monitoring that normally requires intensive training with a great deal of remote -sensing images and sampling of manually annotation. Therefore, statistical observation should be conducted for a large number of remote-sensing images, together with interpretation symbol library including 14 land types of bare land, cement floor and forest etc., which may be used as baseline for training samples. On the basis of the above, key components of production and construction project are evaluated in detail, so as to explore the formation rule of land type for airport or water-related transport. It can provide a reliable base for decision making and understanding of restrict conditions in the later establishment of study model.

关 键 词：生产建设项目 扰动土提取 人工标注 遥感解译 组合规律 

分 类 号：S127[农业科学—农业基础科学] S157