基于空间推理的专题影像解译

THEMATIG IMAGE INTERPRETATION BASED ONSPATIAL REASONING

遥感影像的计算机解译已广泛地应用于许多领域，但要达到完全自动地解译的目标仍然存在不少问题。本文研究的问题是，在统计分类的基础上，引入空间推理方法，对遥感影像进行专题解译，直至最后自动地输出专题图。据此，文中提出了按专题影像分析需要所形成的由原始影像、分割影像、等质区影像、分类解译影像和功能区影像等构成的表达法层次。在低层次影像分析中，影像按照统计属性被分割；在中间层次和高层次的影像分析中，与地物和专题有关的结构和空间知识被利用。本文说明了四叉树生成、复合形标记、复合形归并和匹配等在沟通不同的表达法中的作用。设计了应用于专题影像解译的差别图。介绍了与差别图和空间分析有关的规则。最后，上述技术已被有效地应用于中国吉林省双阳县的植被调查。

: Computer representation and interpretation of remotely sensed imagery have been widely used in many applications. But there exist some problems to be solved for a ful ly automatic solution. The suggested approach in this paper deals with primary image, segmentation image, homogeneous region image, class interpretation image and functional area image according to the needs for thematic image analysis. The image is segmented in accordance with statistical attribute in low level image analysis. The structure and spatial knowledge about the scenes and themes are under consideration in intermediate level image analysis and high level image analysis. The effects of quadtree generation, complex labeling and complex incorporating,and matching in linking up all representations are described The use of discrimination graph algorithms for thematic image interpreta tion has been designed. A set of rules for a particular kind of discrimination graph in combination with some spatial constraints have been provided by our interactive knowl edge acquisition subsystem. Finally, the proposed techniques and algorithms have effec tively been applied to and area of Shuang Yang county in Jilin province of China for Vegetation classification.