Fast copying of the profile of sample object is a requirement of vision-based reverse engineering system. However, there is a problem that the size of sample object is limited to a certain range. In this study, this obstacle is solved by a novel self-calibration method. Based on inverse perspective transformation approach, using a flat marked plane, the 2D profile of a sample object can be obtained with a freely moving camera, such as the camera of smart phone. This allows quickly collecting of the 2D shape of an object with any size. Experimental results show that the proposed method achieves high accuracy. This helps to improve the 2D vision-based reverse engineering system more efficient and flexible.