Abstract: SIFT algorithm is widely used in Unmanned Aerial Vehicle（UAV） image sequence mosaic.Since the traditional SIFT-based image mosaic takes highly computational cost，through analyzing the homography relations between the original images and the down-sampling images，a method for improving mosaic speed is proposed.The role of various parameters of SIFT feature detector is dealt with，and a mechanism of mosaic parameter selection is derived.Experimental results show that the mosaic method can select suitable mosaic parameter value according to the image characteristics and applications，perform fast and steady.

Key words: image mosaic, Unmanned Aerial Vehicle（UAV）, Scale Invariant Feature Transform（SIFT）, homography, mosaic parameter

摘要： SIFT算子在无人机影像拼接中有着广泛的应用，针对传统的基于SIFT的影像拼接处理速度慢的问题，通过分析原始影像与缩小影像之间单应矩阵的关系，提出了一种改进拼接速度的方法。分析了SIFT特征检测时各种参数的作用，设计了一种选择拼接参数的机制。实验结果表明，该方法可以根据影像数据的特点及应用需求设计出合适的拼接参数，有效地提高拼接速度，保证较好的拼接效果。

关键词: 影像拼接, 无人机, SIFT, 单应矩阵, 拼接参数