Aiming at the disadvantages of other calibration methods for 3 D image reconstruction,a new self-calibration algorithm is proposed.Firstly,the camera calibration model is established.Corresponding image points in homogeneous coordinates are generated by perspective projection in the imaging plane.Then self-calibration matrix is used to make corresponding points in different images.Finally,using the projection plane constraint equations to determine whether the pixel falls within the coordinates of camera positioning module.Simulation experiments get the space information from six sides and reconstruction accuracy is high.