基于定量反演与云模型的山区森林主要建群种多光谱遥感识别研究

项目名称： 基于定量反演与云模型的山区森林主要建群种多光谱遥感识别研究

项目编号： No.41301482

项目类型： 青年科学基金项目

立项/批准年度： 2014

项目学科： 天文学、地球科学

项目作者： 王凌

作者单位： 山东农业大学

项目金额： 25万元

中文摘要： 树种遥感识别对森林高效管理具有重要意义，但仍是一个难点。本项目以泰山森林为例，通过分析主要建群种不同物候期的冠层实测光谱差异，确定多光谱遥感识别的最佳时相；以高分辨率多光谱影像为数据基础，充分考虑树种遥感识别中的影响因素（地形、大气、冠层结构、像元组分等），将改进的地形辐射校正算法、混合像元分解方法和植被冠层反射模型用于冠层反射率定量反演，提出针对不同冠层结构的反演模型；基于反演反射率与重采样的实测反射率构建并筛选各树种的敏感光谱指数，利用云模型方法建立基于光谱特征的树种识别模型；对于易混淆区，引入图像纹理特征参数，建立基于纹理特征的树种识别云模型。提出6-8个建群树种的遥感识别方法，并整合构建主要树种综合分析识别模式。该研究可为森林分类信息的及时获取提供可行方法，也为"精准林业"发展提供科学依据。

中文关键词： 树种识别；定量反演；云模型；多光谱遥感；山区森林

英文摘要： Tree species recognition with remote sensing data is very important for forest efficient management, but it still has some difficulty with methods. Taking forests in Mountain Tai as a case, firstly, the optimum periods for extracting tree species information would be chosen by contrasting measured canopy reflectance among different constructive species. Secondly, with a overall consideration of effecting factors on canopy reflectance including topography, atomophere, pixel components and conopy construction, the ground surface reflectance would be retrieved from high-resolution multispectral remote sensing images through radiometric correction using DEM and atmospheric radiation transfer model, and the reflectance of tree canopy with different canopy constructions would be further retrieved by using pixel unmixing method and canopy reflection models. The retrieval accuracy of each species would be assessed by the comparison of retrieval reflectance with measured reflectance of samples. Thirdly, the most sensitive spectral indices would be constructed and chosen by mathematical transformation of band reflectance, then they would be used as independent variables to construct the recognition models for each species with the methods of cloud model, as well as the texture feature indices for tree species in confusion

英文关键词： Tree species recognition；Quantitative retrieval；Cloud model；Multispectral remote sensing；Forests in mountain areas