Extrinsic calibration of Lidar is necessary for vehicle-mounted 3D reconstruction system using multiple Lidars. Concerned on the extrinsic calibration of Lidar, as the scanning point of Lidar is invisible and scanning mode is single-lined, according to the 3D frame transforming model based on space vectors, a new calibration method is proposed, in which a trihedron target is adopted. The proposed method, which contains reconstruction of the target, segmentation of plane feature and extraction of corresponding vectors using the Random Sample Consensus (RANSAC) algorithm, solving the extrinsic parameters using the transforming model of 3D frame based on space vectors, was applied to extrinsic calibration of Lidar in a 3D reconstruction system. The effectiveness and precision of the method are proved by simulation and 3D reconstruction importing the calibration results.