To improve the measurement accuracy of a binocular vision system for target points in remote 3D coordinates, the system structure was optimized, and a high precision coordinate measurement method for the binocular vision system was given. The influence of system structure parameters on the measurement accuracy was analyzed , then the feature target was placed in the monitoring area to calibrate two cameras. In measurement, the image information of obtained feature target was induced into the measuring model to do a calculation and to obtain the space 3D coordinate. The effects of the measuring accuracy of a leveling sensor and the system layout on the 3D coordinate measurement accuracy were analyzed and the error change trends were obtained. It suggests the accuracy of the leveling sensor should be ±0.1° and the system layout should be more reasonable, which will provide a reference for constructing a excellent binocular vision system. In application to monitoring the area with a diameter of 200 m, the results show that the relative error of the target points is less than 0.33%, which satisfies the system's requirement, and is convenient for system's layout as well as avoids the design blindness.