针对合成孔径雷达图像配准对精度和时间的需求,文中对基于相似三角形的配准算法进行了改进:通过角度判断相似三角形,计算简单且可阻止病态三角形影响配准结果；鉴于用于视频追踪的强角点的优良性能,首次引入它作为配准点特征;基于开源计算机视觉库开发了图像自动配准软件,文中简述软件实现过程,分析了代码优化问题；与传统判断相似三角形的配准算法以及常用配准算法相比,实验数据表明,该算法能够精确求解几何形变参数,且运算时间降至毫秒级,对3种几何形变的配准率均高于97％.%For the need in time and precision of synthetic aperture radar image registration, the article advanced an improved algorithm based on similar triangles: judged the similar triangles based on angles> and the simple computation could prevent the ill-condition triangles from influencing the registration results. Introduced the strong corner used in video track for its good performance for the first time. Developed the auto image registration software based on open source computer vision library, the article summarize the realization process of the software and analyzed the code optimization. Compared with registration based on area and length of side to judge similar triangles and common registration methods, the experiment demonstrates that the improved algorithm can calculate the geometric distortion parameters of the registration, the time used decreased to milliseconds and the registration ratios of all geometric distortions reach above 97%.
展开▼
