Binocular vision is a new kind of non-contact three-dimensional (3D) measurement technology, the calibration result will directly affect the precision of the 3D object measurement. Based on the iterative closest point (ICP) algorithm principle to obtain translation and rotation parameters between two point sets, a method is proposed by compensating binocular stereo calibration matrix to improve the precision on the basis of the results of traditional binocular pose calibration. The camera model, binocular vision measurement model and the basic steps of the ICP algorithm are introduced. The external parameters of binocular vision and the same target plane are used for obtaining binocular vision pose matrix, and a method is proposed by using the rotation and translation matrix of the two groups of point set to compensate binocular pose matrix based on ICP algorithm. The analysis model of corresponding target point coordinates projection error is established. Nine sets of calibration images including 5×7 points are collected, and the binocular vision calibration parameters are obtained, and the pose matrix using the ICP algorithm is compensated and nine sets of calibration error by using points coordinates projection error model are analyzed. Experimental results show that the application of ICP algorithm used to compensate the binocular calibration model could significantly improve the accuracy of binocular calibration.