摘要
针对非共线多CCD遥感图像匹配点的分布特点,该文提出一种基于聚类的误匹配点去除方法。首先,根据匹配点的沿轨方向偏移量曲线,获取匹配点的多维特征向量。然后,对匹配点集进行聚类处理,将所有点聚为一个类簇,最后根据簇半径序列曲线的变化趋势区分正确点和误匹配点。通过天绘1号02星全色遥感图像的实验和处理,结果表明在误匹配点去除和正确匹配点保留方面所提算法与其它方法相比具有更好性能。
Considering the distribution characteristic of the matching points of non-collinear multiple Charge- Coupled Device (CCD) remote sensing images, a new method based on clustering to eliminate the mismatching points is proposed. First, the multi-dimensionality feature vector of matching points is obtained on the basis of the disparity curve in along-track direction. Second, all points are clustered to one cluster. Finally, the points are marked off according to the variation trend of the semi-diameter of the cluster. The experiment results running on the panchromatic image of mapping satellite 1-02 show that the method has better performance on eliminating the mismatched points and keepin~ the matched points.
出处
《电子与信息学报》
EI
CSCD
北大核心
2017年第10期2382-2389,共8页
Journal of Electronics & Information Technology
基金
国家自然科学基金(41001214)~~
关键词
遥感图像
非共线
电荷耦合器件
误匹配点
聚类
Remote sensing images
Non-collinear
Charge-Coupled Device (CCD)
Mismatching points
Clustering