可变阈值SIFT算法匹配无人机影像

Variable Threshold SIFT Algorithm for UAV Image Matching

针对无人机重量轻、稳定性差,所拍摄的数字影像存在重叠度不规则、旋转尺度大,造成影像匹配处理具有一定难度的问题,SIFT匹配算法具有尺度不变性、旋转不变性等特征,本文扩大了检测性范围、采用自适性阈值对算法进行改进。改进前后的算法对两组重叠影像进行试验研究,在影像特征点提取数量、匹配效果之间进行对比分析,实验结果表明,对于旋转不稳定性的无人机数字影像,扩大检测范围并采用自适性阈值的SIFT算法更具有优势,可获得更优的影像匹配效果。

It is difficult for traditional image processing algorithm to deal with the Unmanned Aerial Vehicles (UAV) image, because the UAV have light weight, poor stability and the captured digital images are overlapped irregularly. Aiming at the problems, this paper adopts the SIFT algorithm with an adaptive threshold and expanded detection range, which has the characteristic of scale invariance, rotation invariance. Two methods are adopted to process two pairs of overlapping images. The comparison is made between the number of image feature points extracted and the matching effect. The experimental results show that the SIFT algorithm with variable threshold and expanded detection range has more advantages than the SIFT algorithm as to process the rotating and unstable UAV images, thus better matching effect is achieved.