Multiple-view multiple-scale navigation for micro-assembly

Abstract

The problems of operator interaction in micro-assembly tasks are addressed. The solution is a visualisation and tracking system based on techniques from computer vision and robotics. The man-machine-interface leaves the operator free to concentrate on the task whilst the machine handles the problems of precision and scale. The benefits of this approach are increased productivity and precision. An example of an implementation is presented that concentrates on tracking and navigation from a perspective view. This illustrates principles of robustness in feature tracking, parameter estimation and model reconstruction. Estimation is then followed by bundle adjustment , which enforces a parameterisation consistent with a motion model. The mathematical framework reduces dependency on explicit calibration of system parameters, using the constraints between motion kinematics and multiple view geometry.