In the study of robotics, human motion imitation and learning is an active area of research. To learn motion from imitation, robots allow realistic movement without complex teaching and it's also thought that cooperation tasks between people and robots become easy. In those researches, image feedback is useful. Features of the image feedback by vision sensor are non-contact and possible to extract many environmental information. In this paper, I proposed control system which can concrete task with human motion imitation. We capture human motion by camera, and make robots imitate the motion. Grasping and manipulation are achieved by force control based on human motion. Specifically, this paper is focused on adjusting of unknown shaped objects and switching control system for contact and non-contact. With this in mind, proposed control system is evaluated by simulation.