摘要
为解决遥感图像配准易受到环境变化等影响,导致匹配鲁棒性低、配准精度差的问题,提出一种基于网格运动约束的遥感图像配准方法。利用AKAZE (accelerated-KAZE)算法提取图像特征点,采用二进制描述子BRISK (binary robust invariant scalable keypoints)对特征点进行有效描述,利用GMS (grid-based motion statistics)基于网格的运动估计方法实现快速且鲁棒性较强的特征点匹配,利用随机采样一致性算法对匹配点对进一步提纯。实验结果表明,该方法提高了匹配速度和匹配正确率,在遇到光照变化、尺度变化及旋转变化的情况下能够保持稳定的匹配效果。
To solve the problem that remote sensing image registration is susceptible to environmental changes,resulting in low matching robustness and poor registration accuracy,a remote sensing image registration method based on grid motion constraints was proposed.The AKAZE(accelerated-KAZE)algorithm was used to extract image feature points,and the binary descriptor BRISK(binary robust invariant scalable keypoints)was used to validate the feature points.The GMS(grid-based motion statistics)estimation method was used to implement fast and robust feature point matching.The random consistency algorithm was utilized to further improve the matching point pairs.Experimental results show that the matching speed and matching accuracy are improved,and the stable matching effect can be maintained in the case of illumination changes,scale changes and rotation changes.
作者
李丹
徐倩南
LI Dan;XU Qian-nan(School of Electrical and Information Engineering,Anhui University of Technology,Ma’anshan 243032,China)
出处
《计算机工程与设计》
北大核心
2020年第7期1947-1951,共5页
Computer Engineering and Design
基金
国家自然科学基金项目(61601004、61472282)。
