采用最优生成树的正射影像镶嵌线快速智能检测  

Fast and Intelligent Seamline Detection for Orthoimage Mosaicking Based on Minimum Spanning Tree

作　　者：陈继溢 许彪[1] 张力[1] 艾海滨[1] 杜全叶[1] 

机构地区：[1]中国测绘科学研究院,北京100830 [2]国家测绘地理信息局卫星测绘应用中心,北京100830

出　　处：《测绘学报》2015年第10期1125-1131,共7页Acta Geodaetica et Cartographica Sinica

基　　金：国家自然科学基金(41271374);国家科技合作专项(2012DFA20930)~~

摘　　要：针对高分辨率遥感影像提出了一种基于最优生成树的镶嵌线快速智能检测方法。该方法利用影像亮度差异和梯度构建重叠区域差分影像并视其为带权无向图,以Bottleneck模型为基础,采用最优生成树生成方法智能检测最佳镶嵌线。该方法回避了迭代搜索过程,具有速度快、效率高的优点。试验结果表明,所提出的方法能够在保证镶嵌线质量的同时具有较高的效率,可有效解决正射影像镶嵌过程中镶嵌线的自动选择问题。A method of fast and intelligent seamline detection is presented that based on minimum spanning tree for high resolution orthoimage mosaicking.The image gradient and difference of homonymy pixels in the overlap area are calculated to build the differential image,which is deemed as a weighted undirected graph.According to the Bottleneck model,the optimal seamline is detected on the differential image by finding the minimum spanning tree of the weighted undirected graph.This method discards the conventional iterative process,thus achieves high speed.Experiment results illustrate the value of the proposed method which achieves great efficiency and guarantees the quality of the seamlines at the same time.

关 键 词：最优生成树 Bottleneck模型 正射影像镶嵌 镶嵌线智能检测 

分 类 号：P237[天文地球—摄影测量与遥感]