Study On Vision Servoing For Micro-Handling Oriented Miniature Robot

As a typical micro electro-mechanical system (MEMS) device, miniature robot has the characteristics of compactable configuration, flexible motion ability and can accomplish some detection and repairing micro-handling tasks in relative narrow working space in which traditional mechanical system cannot enter. So, it has wide application future in micro-handling domain and has become a new research hotspot of corporations and institutes around the world. The development objective of the research is setting up miniature robot based micro desktop factory. Sponsored by the National 863 Project Foundation, the dissertation takes the real-time visual servoing of the miniature robot as the research mainline and aims at establishing a visual guided miniature robot based automatic micro-handling system. After abundant investigation and research work, the miniature robot visual servoing key technologies are deeply studied.On the base of analyzing the architecture of typical micro-handling system and considering the motion characteristics of designed miniature robot, a miniature robot based automatic micro-handling system is established. The mechanical configuration of macro/micro dual driven miniature robot, external intelligent global visual system, microscopic visual system and one dimenisation micro force sensor are introduced. Then, the control system architecture of designed micro-handling system is described briefly.Global visual sensor supervises the whole workspace of the miniature robot, measures the posture parameters of the robot and guides the robot to move quickly in the workspace. To realize real-time global visual control, a marker based robot posture parameters measuring method is proposed, and an image based global visual controller with dynamic Look and Move mode is designed to guide the robot to move into micro-handling area. When the global visual control is accomplished, to ensure the end-effector of robot gripper be inside the view-field of microscopic visual sensor, a motion detection based end-effector searching strategy is designed.The major effort of microscopic visual system is measuring the relative position information of objects and tools and then to guide the tools close to objects. In the dissertation, a new SUSAN based algorithm is proposed to realize auto-focusing. To solve microscopic image segmentation problem, a Canny edge detection with modified Hough transform object parameters measuring method is designed. To track end-effector of robot gripper in real-time, the Normalized Cross-Correlation (NCC) template-matching algorithm is utilized. And eye-hand harmonizing strategy, template sub-sampling are introduced into NCC algorithm to reduce matching time cost. To realize microscopic visual servoing, a dynamic robot orientation measurement method is designed and a finite state machine style microscopic visual controller is introduced to guide the robot to move to object with micro motion mode. Then, the depth directional control strategy of robot gripper is designed based on Depth-From-Focus principle.At last, real-time visual servoing software architecture is discussed. And dual processes parallel visual calculating system based on Inter Process Communication (IPC) is designed. Then, some micro-handling experiments such as micro parts gripping, transportion and assembly are done to vertify the validity of miniature robot based automatic micro-handling system.