Manufacturing flexibility can be significantly improved with the application of machine vision for 3D object recognition. Majority of previous works in this field, however, have been devoted to 3D static- or 2D moving-object recognition.

This thesis presents a novel approach to the 3D moving-object recognition problem. Two principles, active vision, via a robot-mounted camera, and object premarking, via a circular-marker-based surface-identification scheme, are utilized. As a natural extension of the (static) active object-recognition system (ACTOR), previously developed in the CTMLab. the moving-object recognition system (MORE) integrates many of the components that were used in ACTOR. These include elliptical parameter estimation. 3D pose estimation of circular features. 2D boundary-representation scheme, and standard object-view matching.

In MORE, a Kalman-Filter-based motion-estimation method is utilized for tracking the moving circular features. The predicted poses of the markers are subsequently used to guide a mobile camera to appropriate object-viewing positions.

One of the major difficulties in applying the circular-feature pose-estimation method to moving objects is the orientation-duality problem. To solve this problem, the thesis formulated several methods, based on the analysis of consecutive images, by using constrained object motion and additional surface features. Ill-conditions which lead to the failure ofthe methods were also investigated analytically. Experiments were conducted to verify the feasibility ofthese methods.

In conclusion of the work, an overall robotic experimental system was utilized to test the feasibility of integrating all the involved components of the proposed movingobject recognition method.