The resumption of characteristic points'' deepness information is completed based on stereo vision measurement principle, also presented is a image based three-dimensional modeling method. All interior and exterior parameters of the camera are calculated. After the camera parameters are obtained, the corner points'' coordinates in 3D space are calculated based on the corresponding image points'' coordinates in two or more images. Therefore, the deepness information of the object can be obtained form several images. Finally, 3D model of the object is constructed in Rhino by using the discrete and sparse points cloud of the corner points.