Each sensor in a large-scale machine vision system has its own local coordinate system to which its measurements are referred. To use the measurement results, the data need to be transformed into a global coordination system's. This dictates that all of the local sensor systems relative to the global coordinate system need to be calibrated by a global calibration method. In this paper, a global calibration method for determining the relationship between each sensor coordinate system and the global coordinate system in a large-scale machine vision system is introduced. Two coordinate transformation methods corresponding to the global calibration method, direct and indirect transformation methods, are developed. The global calibration method is very useful and has been successfully used in many practical applications, and very good results have been achieved, as observed from the experimental results in the end of the paper.