Abstract: For the design and development of a vision sensor for the ground test, a method of 3D reconstruction of feature point on large scale object surface from a single image is proposed. By establishing the math model of the 3D coordinate of feature point on the object surface and using the spatial ray through feature point, the 3D coordinate of the feature point can be determined using sigle image. According to the characteristics, the object surface can be classified into three types, high order surface type, block plane typeand block surface type, while the corresponding location methods are introduced . Simulation experiment is conducted to compare the accuracy of three different 3D reconstruction methods, and the proposed method performs the best. By the measurement precision of 1 mm in the range of 8000mm8000mm, it is proved that the proposed method is suitable for 3D reconstruction of feature point on large scale object surface.