1. Patent Examination Cooperation Center of the Patent Office, State Intellectual Property Office of the P. R. C, Beijing 100190，China；
2. College of Information and Electrical Engineering, China Agricultural University, Beijing 100083, China；
3. College of Food Science and Nutritional Engineering, China Agricultural University, Beijing 100083, China
Abstract：This paper presents a method for building 3D models of pork carcass characteristics to improve the
accuracy in the automatic grading systems based on 2D images. The 3D models were established from carcass images
taken from different viewpoints through image processing, camera calibration, stereo matching based on epipolar
constraint and 3D coordinates calculation. For the data of half carcass length, the results of 3D reconstruction had a
total relative error of 5%, while 10% obtained from a single image. For fat between the 6th and the 7th ribs, the results
of 3D reconstruction had a total relative error smaller than 8%, whereas the relative error obtained from a single image
was as high as
2013, Vol. 27 
