In the automated manufacturing system robots, the robot system should have the ability to interact with a flexibility environment with intelligence.
In this thesis, an intelligent robot system which matches randomly placed jig-saw puzzle pieces of arbitrary geometric patterns is studied.
To be specific, arbitrary geometric patterns in two-dimensional image at low resolution are extracted using the proposed boundary tracking algorithm.
As features, critical points of each puzzle piece are found using the discontinuity of curvature, and at each point convexity or concavity is determined to characterize puzzle pieces.
To fit the puzzle pieces, a heuristic puzzle matching algorithm based on a priori information of the boundary of the matched puzzle is proposed.
The performance of puzzle matching is evaluated using a newly defined accuracy measure.