Based on the requirement of high precision map construction for unmanned vehicle, the relative pose calibration between three-dimensional lidar and GPS/INS integrated navigation system is necessary. The coordinate systems of the three-dimensional lidar and the GPS/INS integrated navigation system are established according to the charac-teristics of each type, respectively. Aiming at the problem that the relative position between the two coordinate sys-tems can not be measured accurately and the relative rotation angles between them are difficult to be measured, the calibration of three-dimensional (3D) lidar external parameters is studied. Based on the point cloud registration, the hand-eye calibration model of 3D lidar and GPS/INS integrated navigation system is established. The point cloud pose is obtained after registration by ICP algorithm. Using the multi-pairs point cloud pose, the extrinsic parameters are calculated by the least square method.The 3D lidar data are resolved into continuous single frame point clouds, and the longitude and latitude coordi-nates of GPS are converted into plane coordinates. At the same time, the data of the two kinds of sensors are syn-chronized in time, and the pose tag is attached to each point cloud frame. Aiming at the phenomenon of point cloud distortion in vehicle moving process, a method of coordinate correction for laser scanning point in single frame point cloud is proposed. The point cloud with the attitude angle changed is chosen as the point cloud pair. A point cloud registration algorithm is used to register the corresponding point cloud pairs, and the new pose after registration is obtained. The relative pose of each pair of cloud points before and after registration can satisfy the model of hand-eye calibration. Finally, a set of over determined equations is constructed by multi-pairs registration data satis-fying the “non-parallel condition”, and the transfer matrix is solved by decoupling. The rotation matrix is trans-formed into quaternion form, and the rotation parameters are solved by the least squares method. After the rotation parameters are obtained, the translation parameters are computed using the point cloud positions before and after registration.Monte Carlo simulation results show that the algorithm is accurate and effective. The calibration experiment is carried out on the unmanned vehicle laboratory JJUV-6, and the 3D relative pose relations are obtained. It shows that after calibration the effect of 3D point clouds reconstruction is better, and the overlapping degree of point clouds is close to the registration effect. For the purpose of applications, it can meet the needs of 2D map construc-tion and 3D environment reconstruction.