摘要
针对传统尺度不变特征转换(SIFT)算法在高分辨率影像中检测特征点数量多、匹配效率低、无法快速对无人机低空遥感影像进行特征匹配的问题,该文优化SIFT-OCT算法的特征检测、特征匹配方法,主动放弃第一组尺度空间进行特征检测,并采用影像分块的方法加快检测过程;在特征匹配阶段,提出相似性系数进行匹配点对二次筛选,利用随机抽样一致性(RANSAC)算法计算透视变换模型参数进行精匹配。选取同一无人机序列影像中的4组不同地物类型的影像进行对比验证实验,结果表明,优化SIFT-OCT算法极大地限制特征提取数量,提高影像匹配效率,适合无人机低空遥感影像匹配。
Aiming at the problem that the traditional scale invariant feature transform(SIFT) algorithm detects a large number of feature points in high-resolution images, and the matching efficiency is low, so it is impossible to quickly match the features of the unmanned aerial vehicle(UAV) low-altitude remote sensing images. The feature detection and feature matching methods of the SIFT-OCT algorithm were optimized, actively abandoned the first set of scale spaces for feature detection, and used the image block method to speed up the detection process. In the feature matching stage, a similarity coefficient was proposed for secondary screening of matching points, and the perspective transformation model parameters were calculated by random sample consensus(RANSAC) algorithm for precise matching. Four sets of images of different feature types in the same UAV sequence were selected for comparative verification experiments. The results showed that the optimized SIFT-OCT algorithm greatly limited the number of feature extractions and improved the image matching efficiency. It is a method suitable for UAV low-altitude remote sensing image matching.
作者
王亮亮
胡海峰
WANG Liangliang;HU Haifeng(College of Mining Engineering,Taiyuan University of Technology,Taiyuan 030024,China)
出处
《测绘科学》
CSCD
北大核心
2021年第6期102-108,共7页
Science of Surveying and Mapping
基金
国家自然科学基金项目(51574132)。
