基于改进SPHP算法的无人机遥感图像拼接

Remote sensing image stitching of UAV based on improved SPHP algorithm

针对无人机遥感图像拼接的稳健性问题,对SPHP(Shape-preserving half-projective,SPHP)图像拼接算法进行了改进。利用SIFT(Scale invariant feature transform,SIFT)算法完成特征粗匹配,按粗匹配点对之间的距离对所有匹配点对进行排序,以最优匹配点对作为随机采样一致算法的初始样本集进行特征精匹配。利用投影变换和相似变换相结合的变换模型进行图像配准,采用线性加权图像融合方法进行无人机遥感图像的无缝拼接。实验结果表明:本文算法能有效剔除错误匹配对,匹配正确率平均提高约7%,明显改善图像拼接的视觉效果和图像质量,可以满足实际需求。

In view of the robustness of UAV remote sensing image stitching,the basis of SPHP image stitching algorithm is improved.The SIFT algorithm is used to complete the rough feature matching,then all the matching point pairs are sorted according to the distance between the rough matching point pairs,and feature matching is performed with the optimal matching point pair as the initial sample set of the random sample consensus algorithm.The transformation model combined with projection transformation and similar transformation is used to perform image registration.The linear weighted image fusion method is used to seamlessly splicing the remote sensing image of the UAV.The experimental results show that the proposed algorithm can effectively eliminate the mismatch matching,and the matching accuracy rate increases by about 7%on average,which significantly improves the visual effect and image quality of image stitching.It can meet actual needs.