A new algorithm for a three-axis auto-alignment system using vision inspection

Abstract

In this study, a new algorithm was developed for a vision-aided auto-alignment system. The proposed algorithm makes use of the correlation between the input signals and the output feedback. Through matrix transformation, the algorithm establishes the characteristic matrix of the system function between input and output signals. The search for this characteristic matrix involves inputs of real-time movement commands. With movement of fiducial images captured by CCD, the characteristic matrix can be obtained after image processing and matrix transformation. Alignment operation using the proposed algorithm will only make use of the coordinate system within the field of view of CCD. With the difference in fiducial mark as well as positional variation obtained, the alignment commands can be fed to the three-axis motion control mechanism to compensate for the difference in position and orientation, thus, achieving accurate alignment. Results from over hundreds of experimental runs have proved that rapid automatic and precise alignment can be achieved using a vision-aided system integrated with the proposed algorithm.