无地面控制点航空图像无缝拼接算法  

Airphoto Seamless Mosaic Algorithm Without Ground Control Point

作　　者：贾西茜[1] 崔希民[1] 邱亚辉[1] 金晶晶[1] 范冬丽[1] 张绪郡[1] 

机构地区：[1]中国矿业大学(北京)地球科学与测绘工程学院,北京100083

出　　处：《影像技术》2011年第4期34-39,共6页Image Technology

摘　　要：针对航拍图像拼接中,因为航带中航片拼接误差积累导致拼接后期图像发生扭曲问题,本文提出一种利用捆绑调整技术削弱航片拼接过程中误差累积。该算法采用SIFT(Scale Invariant Feature Transform)进行特征点提取和匹配;结合改进的RANSAC(Random Sample Consensus)对特征点进行提纯,剔除外点;由过滤后的特征点通过最小二乘法计算图像间的单应性矩阵,在此基础上运用捆绑调整法整体优化单应性矩阵进行图像间的拼接,解决了拼接后期图像扭曲问题。最后,通过动态加权的融合方法实现图像接缝处平滑过渡。为验证该算法的有效性,选用真实无人机航拍序列图像进行拼接实验,取得良好的拼接效果。For aerial image mosaic,because airphoto splicing error accumulates in the aerial strips so as to cause the problem of later image distortion,in this paper a bundle adjustment technique was used to reduce error accumulation in the process of stitching.The algorithm used SIFT （scale invariant feature transform） to extract and match feature points;and Improved RANSAC （random sample consensus） purifies feature points,excluding outliers;then eligible points were used to calculate value of homography matrix between images by the least square method,on this basis,bundle adjustment method was used to entirely optimize homography matrix in order to solve image distortion in the period of post-splicing,thus mosaic image was accomplished.Finally,the dynamic weighted fusion method was used to better achieve seamless transition in the image seam.To verify the validity of the algorithm,splicing experiments using UAV aerial image sequence achieved good results.

关 键 词：尺度不变特征变换 动态加权 捆绑调整法 误差积累 

分 类 号：V19[航空宇航科学与技术—人机与环境工程] P715.6[天文地球—海洋科学]