基于形状的遥感图像检索系统

研究了基于内容的遥感图像库的快速检索系统．系统采用基于形状全局特征与局部特征相结合的两步检索策略,由简单的全局特征来过滤掉大部分的非相关形状,利用矢量近似方法来对所属类别进行快速定位,缩小搜索空间;局部特征相似性是通过形状小波变换系数之间的距离直接计算的．仿真试验表明,该查询结构能够高效准确地对高分辨遥感图像和航片进行检索．

A retrieval system for remote sensing image based on shape

A fast retrieval system for remote sensing image based on shape was introduced. Unlike common indexing tree, it was a kind of two steps searching strategy based on the combination of global features and local features. In the first step, through filtering dissimilated shapes by using simple global features, the searching space was reduced .In the next step, a scanning method was applied to find the best matching feature. The matching feature was a shape descriptor based on wavelet transform coefficients of contour points. This wavelet descriptor was scaling, transition and rotation invariant. A multi scale shape matching can be processed by it. For more effective performing, scale of query object as a constraint to limit the range of images needing reviewed was introduced. Theoretical analysis and experiments showed it was a high efficient indexing structure for remote sensing image database on web.