利用属性图的遥感数据与GIS数据松弛标号匹配

Relaxing labeling matching for remote sensing image and GIS data based on attributed graph

本文提出一种基于图形匹配来配准影像和地图的方法。该方法将道路网表示成图形,通过对两个图形进行最优匹配来实现影像和地图的配准;同时采用一种整体相容函数来量测两个图形的整体对应优度,并通过连续松弛标号来优化此过程。该方法对平移、旋转变化和尺度变化具有不变性(通过利用合适的属性),同时网络结构的利用降低了非精确匹配过程中的搜索空间的不确定性;该方法不需要用户自定义阈值来调整局部匹配。实验结果表明该方法具有较高的可靠性。

This paper presented a method for image-to-map registration through graph matching. The road networks were considered as graphs, and the registration was implemented through the optimal matching between the two road network graphs. Meanwhile a global compatibility function was used to measure the overall goodness of correspondence of the two graphs, through continuous relaxation labeling to optimize the process. This method would be invariant to translation, rotation and scale differences. Moreover, it would reduce the ambiguity of the search space in inexact matching by network structure. Furthermore, it would not need to define threshold to justify local matches. The experiment results proved its reliability.