Calibration and rectification for the computer vision sensor are the key links of the application of machine vision system which will make the image more accurate so as to impove detection precision of the machine vision system. It is a key step that can not be ignored. An improved planar template method is presented, which extends the tangential distortion model based on the traditional method and the Heikkila model. The imaging model increases the high order nonlinear distortion and has some simplification.Through the model calibration parameters optimized and characteristics of the pixel space rectification value are obtained. The pixel grayscale is corrected by the linear interpolation method, the calibration results and the corrected image can be obtained. The method is implemented through LabVIEW tools in the machine vision detection platform built by the authors. Experiment verifies that images collected by the visual system are corrected well and the rectification accuracy reaches 0.17%.