无人机影像拼接多分辨率自适应无缝融合方法  

Multi-resolution adaptive seamless fusion method for UAV image stitching

作　　者：石耀榕 肖敬达 SHI Yaorong;XIAO Jingda(School of Remote Sensing and Information Engineering,Wuhan University,Wuhan 430079,China;School of Future Technologies,Beijing Institute of Technology,Beijing 102488,China)

机构地区：[1]武汉大学遥感信息工程学院,湖北武汉430079 [2]北京理工大学未来精工技术学院,北京102488

出　　处：《测绘通报》2024年第6期36-40,共5页Bulletin of Surveying and Mapping

摘　　要：本文针对无人机影像拼接中存在的几何错位和拼接缝问题,提出了一种联合应用最佳拼接缝算法和图像融合的方法,对影像几何错位和拼接缝区域进行修复和无缝融合。首先利用图切割算法搜索帧间拼接最小差异线,修复帧间几何错位;然后基于差异线建立条带融合兴趣区,且融合兴趣区范围大小随帧间色调曝光差异大小自适应变化;最后引入融合兴趣区掩膜改进拉普拉斯多分辨率图像融合算法,对帧间色调曝光差异进行平滑处理,实现无人机影像拼接的无缝融合。试验结果表明,相较于传统拉普拉斯融合算法,本文影像融合成果互信息平均提升了12.79%,峰值信噪比平均提升了21.48%,取得了自然无缝的目视效果,具有一定应用价值。Aiming at the problems of geometric misalignment and stitching seams in UAV image stitching,a joint application of the optimal stitching seam algorithm and image fusion is proposed to repair and seamlessly fuse image geometric misalignment and stitching seam regions.Firstly,the graph cut algorithm is used to search for the minimum difference line of interframe splicing to repair the geometric misalignment between frames.Then,the region of interest(ROI)of strip fusion is established based on the difference line,and the size of the fusion ROI range changes adaptively with the size of the tonal exposure difference between frames.Finally,the fusion ROI mask is introduced to improve the Laplace multi-resolution image fusion algorithm,and the tonal exposure difference between frames is smoothed,so as to achieve seamless fusion of UAV.And the seamless fusion of image stitching.The experimental results show that compared with the traditional Laplace fusion algorithm,the image fusion results in this paper improve the mutual information by an average of 12.79%,the peak signal-to-noise ratio by an average of 21.48%,and achieve a natural and seamless visual effect,which is of certain application value.

关 键 词：无人机 几何错位 拼接缝 图切割 多分辨率融合 

分 类 号：P237[天文地球—摄影测量与遥感]