一种无人机航拍影像快速特征提取与匹配算法  

A Fast Feature Extraction and Matching Algorithm for Unmanned Aerial Vehicle Images

作　　者：余淮[1] 杨文[1] 

机构地区：[1]武汉大学电子信息学院,武汉430072

出　　处：《电子与信息学报》2016年第3期509-516,共8页Journal of Electronics & Information Technology

基　　金：国家自然科学基金(61271401;91338113)~~

摘　　要：无人机影像具有非常高的分辨率,边缘和纹理信息更加丰富,基于经典SURF特征的影像拼接算法在处理无人机影像时面临着新的挑战。为提高无人机航拍影像拼接效率,该文提出一种快速特征提取与匹配算法。在特征提取环节,提出采用局部差分二进制算法描述特征,在不降低特征区分性的同时,较SURF描述子而言降低了特征维度。在特征匹配环节,提出采用局部敏感哈希搜索算法代替kd树搜索算法,提高了最近邻特征匹配效率。实验结果表明,与基于SURF描述子和kd树搜索算法的最近邻匹配拼接算法相比,该文算法特征匹配效率有明显提升,匹配精度也有所改善,更适合应用于基于特征的无人机航拍影像快速制图。Unmanned Aerial Vehicle（UAV） images are characterized by a very high spatial resolution, and consequently by more abundant information of the edge and the texture. The conventional stitching methods, which use Speeded Up Robust Features（SURF） and kd-tree based nearest neighbor matching, are facing with new challenges for processing UAV images. In this paper, a fast feature extraction and matching algorithm is proposed for more efficient stitching of UAV images. Firstly, the Local Difference Binary（LDB） algorithm is used to describe the feature, which could reduce the dimension of feature without sacrificing its discrimination. Then, the Local Sensitive Hash（LSH） is used to replace kd-tree search structure, which achieves nearest neighbor matching more efficiently. Compared with the conventional stitching method, experimental results demonstrate that the proposed method achieves a higher accuracy and greater efficiency, which is more applicable to rapid mapping of UAV images.

关 键 词：无人机 影像拼接 特征提取 特征匹配 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]