仿射变换中点匹配的筛选基的对准算法

Selectively Basis Alignment Algorithm of Point Matching under Affine Transformation

点模式匹配对准算法的基本思想是在两个点集中各选择不共线的3个点,唯一确定一个仿射变换,该变换作用于第一个点集中的所有点,用变换后的点集与第二个点集进行比较,检验变换的正确性,但尝试所有3点组合的计算代价太大。文章基于仿射变换的最小方差的参数估计以及数值计算的稳定性,对两个点集中参与匹配的基进行筛选,来确定匹配点对,并可极大地减少计算量。

The basic idea of the alignment method on point pattern method is that triples of image points are matched with triples of model points and for each match,an affine transformation that brings them into alignment is determined.The model points are then transformed by this transformation and compared against the image points to verify the cor-rectness of the transformation.But the computation cost is very high.Based on the minimum variance of parameters esti-mation and computation stability,this paper researches the techniques to eliminate hypotheses from consideration in the alignment method,thus the computation cost is reduced greatly.