An optimal measurement pose number searching method was designed to improve the pose selection method. Several optimal robot measurement configurations were added to an initial pre-selected optimal configuration set to establish a new configuration set for robot calibration one by one. The root mean squares (RMS) of the errors of each end-effector poses after being calibrated by these configuration sets were calculated. The optimal number of the configuration set corresponding to the least RMS of pose error was then obtained. Calibration based on those poses selected by this algorithm can get higher end-effector accuracy,meanwhile consumes less time. An optimal pose set including optimal 25 measurement configurations is found during the simulation.Tracking errors after calibration by using these poses are 1.54, 1.61 and 0.86 mm, and better than those before calibration which are 7.79, 7.62 and 8.29mm, even better than those calibrated by the random method which are 2.22, 2.35 and 1.69 mm in directions X, Y and Z, respectively.