Considering the effect on the image measurement system, a distortion calibration method was designed after analyzing the model of lens distortion. The method consists of two steps, namely calibration on the non-linear distortion and calibration on the perspective distortion. In the first step, the non-linear distortion coefficients of the lens were obtained form drawing the multi feature straight lines and straightening the feature lines. With these coefficients, the non-linear of the image was calibrated, and an image without any non-linear distortion was achieved. In the second step, the distortion parameters were obtained through fitting the parallel-line feature of the image. With these parameters, the perspective distortion was calibrated, and an image without perspective was achieved.The experiments and simulations show that the radial and the perspective-distortion relatively error can be reduced to 5% with this method, which can be used in many similar engineering systems based on image measurement and target recognition in non-fixed locations.