【摘要】 针对无人机航拍过程中因发生机体抖动或机身倾斜而造成的拍摄图像畸变会导致匹配效率低、错误率高的问题,提出一种改进的ASIFT和RANSAC算法相结合的航拍畸变图像匹配策略。对航拍畸变图像采用ASIFT进行预处理,在关键点周围建立的内、外两个八边形邻域内构建112维描述子代替传统SIFT算法的128维描述子,再采用RANSAC算法剔除误匹配点。仿真结果显示,该方法适用于无人机航拍图像的配准,具有仿射不变性,且与传统的ASIFT+RANSAC方法相比匹配速率提高了10%～20%,正确匹配率提高了5%～10%。更多还原

【Abstract】 Aiming at the problem of low image matching and high error rate caused by the occurrence of body shake or body tilt during the aerial photography of the drone, an improved ASIFT and RANSAC algorithm combined with aerial distortion image matching strategy is proposed. Aerial distortion image was preprocessed through ASIFT, 112-dimensional descriptor was constructed in the inner and outer octagon neighborhoods around the key points, to replace the 128-dimensional descriptor of the traditional SIFT algorithm, and RANSAC algorithm was used to eliminate the mismatched points. Simulation results show that the method is suitable for the registration of unmanned aerial image, with affine invariance, and matching rate is increased by 10%～20% compared with the traditional ASIFT+RANSAC method, and correct matching rate is increased by 5%～10%.更多还原