:研究了一种多目标识别算法,该算法用SUSAN角点形成SIFT特征点,采用阶梯图像金字塔结构实现尺度不变,为所有匹配点建立统一的超定线性方程组并对该方程组系数矩阵进行简

化使其维数降低一半,得到增广矩阵.对增广矩阵进行列变换,依据坐标转换的特性可从中提取多目标的稳定正常点,实现了快速分离多目标的匹配点.

-: Study of a multi-target recognition algorithm using SUSAN corner formed SIFT feature points, image pyramid structure by step to achieve scale-invariant, for all the match points to establish a unified super-set of linear equations and coefficients of the equations dimension matrix can be simplified to reduce by half, by the augmented matrix. on the augmented matrix column transformation, coordinate transformation based on the characteristics of multi-objective can be to extract the stability of the normal points, for a rapid separation of multi-target match point.