In order to improve the accuracy of the recovered object

this paper proposed a new algorithm. This algorithm tries to express the machine vision model by parameter equation. Based on this parameter equation

this paper proposed a kind of weighted least square method. It is well know that the recovered accuracy will be changed by the camera’s position. But the original algorithm neglects this

and uses the least square method to fit the point with the minimal error. Aim at this defect

this paper weighted the least square method according to different position

and use parameter equation to express the machine vision model. This algorithm improved the recovered accuracy. The crucial step of the weighted least square method is to evaluate the weight. At the end of this paper

the experiment verified the efficiency of the algorithm.