基于简化SIFT算法的无人机影像重叠度分析

Overlapping degree analysis of images from an unmanned aerial vehicle based on a reduced scale-invariant feature transform(SIFT) algorithm

利用无人机进行低空摄影获取地面高分辨率影像,具有成本低、方便、快捷等优点,但由于无人机飞行姿态不稳定,会导致影像自动匹配效率及准确程度降低.SIFT算子具有良好的尺度、旋转、光照等不变特性,但传统SIFT算法复杂度较高,处理影像时间较长,使数据处理工作效率降低.介绍了一种简化的SIFT算法,并与原SIFT算法做出了对比.说明了影像重叠度分析的主要步骤,将简化的SIFT算法RANSAC算法相结合应用到序列影像重叠度分析中,通过实验证明了可行性.

Compared with traditional images, high-resolution images obtained by unmanned aerial vehicles （UAV） are superior in many aspects such as low cost, rapidness and convenience. However, the attitude of unmanned aer- ial vehicles cannot be determined accurately, which leads to a decrease in accuracy and efficiency of automaticmatching. There are several ascendant characteristics of a SIFT operator such as invariance to scale, rotation, and lightness, and these characteristics are very useful in increasing the matching accuracy of UAV images; however,the traditional SIFT algorithm is complex and requires a long time for image processing. In this paper, a reduced SIFT method was described and some comparisons were made. The overlap analysis of UAV images were processedbased on a reduced SIFT and the random sample consensus （RANSAC） algorithm, and the steps of overlap analysis were described. Finally, satisfactory results were obtained.