基于最佳几何约束的遥感影像点匹配算法

A point matching algorithm for remote sensing image based on the best geometric constraint

针对尺度不变特征变换(SIFT)点匹配算法中几何约束缺失问题,提出了一种基于最佳匹配几何约束的点匹配算法.该算法以SIFT匹配算法为基础,首先构建左右影像特征点集的转换模型,然后采用改进的量子粒子群算法对模型参数进行迭代寻优,每次粒子位置更新后,采用基于搜索圆的特征点匹配算法获取新位置下的特征点,并根据获取的特征点情况计算其适宜度与辅助适宜度来对粒子位置进行评价,经过多次迭代,最终获取匹配影像的最佳几何约束与该约束下相应的匹配点,实现了特征点的匹配.选取多幅遥感影像进行点匹配实验,结果表明:相比其它的点匹配算法,该算法在匹配点的数目与精度上都有显著提高,能够获得更好的点匹配结果.

To solve the lack of geometric constraint in scale-invariant feature transform（SIFT）,apoint matching algorithm based on the best geometric constraint was proposed.The algorithm was improved based on the SIFT matching algorithm.Firstly,the algorithm built a transformation model to describe the feature point sets in images.Secondly,a modified quantum particle swarm optimization（QPSO）was utilized to optimize the parameters of the model.Every time after the positions of particles were updated,a feature point matching algorithm on the basis of search circle was used to obtain the feature points in the new position.According to the feature points,the suitability degree and the auxiliary suitability degree of particles were calculated to evaluate the position of the particles.After several iterations,the optimal geometric constraint of the matching image and the corresponding matching points were acquired,which realized the matching of feature points.Several remote sensing images were adopted for point matching test,and the experiment results demonstrate that the algorithm in this paper is capable of obtaining better point matching results than other point matching algorithms both in quantity and accuracy.