提出一种基于月球影像显著区域的内容特征(LIFBS),同时面向多核处理器架构,提出并行优化的LIF-BS特征生成算法.该算法考虑显著子区域之间的方位、尺度、强度和距离关系,为每一个显著区域生成一个LIFBS局部特征,以此描述月球影像的内容.理论分析与实验结果表明,LIFBS特征具有良好的不变性与相似性表达,同时特征生成算法具有较高的并行效率.
A feature descriptor called LIBFS is proposed for lunar remote sensing images based on the saliency regions and a parallel LIBFS generating algorithm is proposed facing the multi-cores processor architecture. The algorithm considers the location, scale, slxength and distance relationship among different saliency regions in a lunar image. Several local LIBFS features are generated to describe the content of the image together. Theoretical analysis and experiments show that the LIBFS works well in invariance and similarity presentation,and the feature generating algorithm is of high parallel efficiency.