面向序列影像的三维场景重建方法  

A method of 3D scene reconstruction based on sequence images

作　　者：侯晓芬[1] 邢帅[1] 马东洋 耿迅[1] 李鹏程[1] 葛忠孝 

机构地区：[1]信息工程大学,郑州450052 [2]中国天绘卫星中心,北京102101

出　　处：《测绘科学》2016年第2期126-129,共4页Science of Surveying and Mapping

基　　金：国家自然科学基金项目(41371436)

摘　　要：针对序列影像三维场景重建过程中存在的自动化程度不高、模型精度不够等问题,该文提出了一种基于序列影像的三维场景重建方法。该方法利用尺度不变特征变换算子获得了大量稳定的特征点后,结合K-D树搜索策略以及随机抽样一致性误匹配剔除方法得到稳定的匹配点对,然后采用稀疏光束法平差获取序列影像的相对位置姿态参数,最后采用多视密集匹配方法实现场景目标的密集重建。实验结果表明:该文所提方法可以快速得到准确、完整的三维重建结果。To reconstruct the 3Dscene more automatically and reliably,a new method of 3Dscene reconstruction for sequence images was proposed.Firstly,the key points were extracted and matched with the SIFT algorithm and the search strategy of K-D tree,and the outliers were removed by RANSAC method;then the sparse bundle adjustment was used to estimate the relative orientation and position;in the last,a dense 3D reconstruction was gained by the multi-view stereo method.The results of the experiments demonstrated that this method could be used for acquiring dense 3Dreconstruction effectively.

关 键 词：序列影像 稀疏光束法平差 密集匹配 三维重建 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]