Robust vision system for micro-object manipulation

Abstract

Semi-autonomous and fully autonomous assembly and manipulation of micro- objects is a complex process. As part of the MINIMAN micro-robot project, a vision subsystem is required which can recognise and track objects both under an optical microscope and within a scanning electron microscope. This subsystem is required as part of a more complex system for assembling and manipulating micro-objects. The two different operating environments provide many challenges when building a generic vision system due to the vast differences in the quality of the images. This paper provides a detailed description of this new vision system, together with a discussion and analysis of its flexibility and extensibility. Recent results of the system's ability to recognise rigid objects robust to camera noise and object occlusions are given, and the adaptability of the system to recognising biological objects is discussed. Finally, an overview is provided of the communication strategy between the vision subsystem and the micro-robot control system.