Human-robot collaborative handover refers to the collaboration between humans and robots so as to handover objects. In the process of human-robot collaborative handover, it is difficult to grasp targets well because of the inability to identify the objects’ pose accurately. Therefore, a 6D pose recognition network based on the PnP algorithm is introduced to achieve accurate identification of the object posture and to achieve a precise grasp. Further, an improved method for making dataset of the passed objects is proposed to realize accurate recognition of arbitrary objects. Additionally, the precise pose positioning and accurate grasping of objects are realized by the vision system calibration, coordinate transformation, and improved grasping scheme. In order to verify the feasibility of the proposed human-robot collaborative handover system, 4 volunteers were invited to perform 80 times of human-robot collaborative handovers. The experimental results show that the average deviation distance of the proposed human-robot handover system is 1.97cm, the average handover success rate is 76%, and the average handover time is 30s. In case of ignoring the grasping posture, the success rate can reach 89%, which is provided with good robustness and good application prospect.