【作者】 王萍；

【导师】 林宗坚； 张继贤；

【作者基本信息】 山东科技大学 ， 大地测量学与测量工程， 2004， 博士

【摘要】 土地利用／土地覆盖变化(LUCC)信息提取方法的研究是LUCC研究的核心內容和关键技术，本文以中分辨率遥感数据为主要数据源，针对不同生态分区和目标特征，发展了分区分层的决策树LUCC信息提取方法，基本实现了变化信息自动发现-提取以及类型表示的自动化流程。 笔者主要进行了以下方面的研究： 1、建立了遥感及地面实测数据支撑的综合数据库； 2、基本实现了LUCC变化信息的自动化提取流程； 3、发展了基于中分辨率遥感数据的变化信息自动发现方法； 4、有效融合了非监督分类结果以及专家知识基础上的决策树分类技术，建立了试验区决策树分类模型； 5、将决策树技术应用于变化信息的提取方法中，以多源多时相遥感及其他辅助数据为基础，构造了变化转化类型的决策树模型； 6、以中国西部石羊河流域和河北内蒙局部地区为试验研究区，进行了分区分层技术和决策树的方法应用，完成试验研究区反映1980年以来2—3个时相遥感影像的土地利用／土地覆盖分类和LUCC专题信息层的结果提取，并用变化转换关系的矩阵表表示，精度评价表明，分类和变化类型的精度均达到80％以上。 实践证明，所采用的方法和技术提高了LUCC信息提取的自动化水平和精度，并可以在更大范围内推广使用。更多还原

【Abstract】 The study on the method of extracting change information from land use / cover is the kernel content and key technique, this article is based on remote sensing data of the middle resolution as main data source, aims at different zoology district and goal characteristic, develops :he decision-tree method of extracting of LUCC information based on del-imination in different district technology, realizes auto-find and extraction and automatic flow of expression types.The author mostly carries on the study as follows:1. Established the synthetic database which are based on remote-sensing data and survey in the field;2. Realized the extracting flow of LUCC information automatically;3. Developed methods of auto-find change information based on middle resolution remote-sensing data;4. Synchronized the outcome of unsupervised classification and decision-tree classification based on expert knowledge efficiently, built the model of decision-tree classification in test areas.5. The technology of decision-tree is applied to methods of the extracting change information, with the multi-source, multi-time remote-sensing data and other data as the foundation, the decision-tree model of the change transformation type is constructed;6. Taking the area of Shiyanghe drainage being in the west of China and local area of Hebei and Neimeng as research district, based on methods of multi-layer in different district technology and decis on-tree, has finished and extracted land use /land cover’ s classification and LUCC information from remotely sensed data of 2-3 times since 1980, and they are expressed by a matrix of the relationship of change transformation. The precision evaluation shows that the precision of classification and change types exceeds 80%. Examples prove that the adopted method and technology improve the level of autoimmunization and precision of LUCC information extraction, and can be generalized in a larger scope.更多还原