Aiming at the calibration optimization problem of the multi-vision system in the process of motion capture experiment, this paper explores the influence of the selection of reference camera on the iterative calculation of bundle adjustment through the error analysis of camera parameter estimation process, and proposes a Q parameter evaluation method based on it, which uses the entitled undirected graph to obtain the shortest connected path between camera nodes. The back projection error and the weighted mean of mean square error are used to evaluate the rationality of the reference camera selection. In the experiment, multiple cameras have been used to form a multi-view stereo vision system to verify the reference camera selection method proposed in this paper. The experimental results show that the selection of reference camera is necessary and the Q parameter evaluation method designed is effective, which can further improve the accuracy of the system calibration and has important reference value for the application of optical motion measurement.