基于无人机航拍图像拼接算法的优化  

An Optimization Algorithm for Stitching of Aerial Images Taken by UAV

作　　者：吴成东[1,2] 张润[1] 刘宝德 房卓群 

机构地区：[1]沈阳建筑大学信息与控制工程学院,辽宁沈阳110168 [2]东北大学信息科学与工程学院,辽宁沈阳110819 [3]辽宁省城市建筑设计院,辽宁沈阳110031

出　　处：《沈阳建筑大学学报（自然科学版）》2015年第1期182-192,共11页Journal of Shenyang Jianzhu University：Natural Science

基　　金：国家自然科学基金项目(61273078)

摘　　要：目的为了提高无人机航拍图像拼接的精度,深入研究了航拍图像拼接中提取特征点的算法,并对原算法加以优化.方法在图像特征点提取的SIFT算法中,设计了一种将Harris角点检测算子融入SIFT特征点提取的优化算法,优化后可以突显获取到的特征点的独特性.结果利用优化算法获取图像特征角点,可以降低实验过程中所消耗的检测时间,有效地改善了SIFT算法中匹配数据量大及过程繁琐的弱点,同时优化算法简化了图像特征点匹配的计算过程,降低了计算量,提高了实验效率.结论优化算法可以去除大量的类匹配点,使图像特征点的独特性更加明显;也提高了图像的配准精度,增强了关键点的稳定性,在关键点的匹配速度和准确率上有积极的影响.In order to improve the mosaic accuracy of aerial images taken by the UAV, algorithms for extracting feature points of image stitching and optimization of the original algorithm were studied in this paper. In the SIFT algorithm for extraction feature points, the authors designed an operator optimization algorithm that mixes a Harris corner detection together with SIFT feature point extraction, so that the unique nature of the acquired feature points were highlighted after opti mization. It is found that the optimization algorithm for capturing image feature corners can short the test time in experiment, effectively improve the weakness for great amount of data and process cumbersome in SIFT. Conclusion is that the optimization algorithm can remove a lot of class matchpoints and make the unique of image feature points more obviously. The image registration accura cy can be improved and the stability of key points can also be enhanced. Meanwhile, the proposed algorithm simplifies the calculation process, reduces the amount of computation and enhances ex periment efficiency. The optimization algorithm has made a positive impact on the speed and accu racy of matching key points.

关 键 词：无人机 图像拼接 优化SIFT算法 小波变换 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]