High-Precision Calibration Approaches to Robot Vision Systems

Abstract

Motivated by the practical issues in applying the vision systems to the industrial robot vision applications, the dissertation has made great efforts on camera calibration, camera recalibration, vision systems calibration and pose estimation.

Firstly, the calibration methods from Tsai are analyzed and improved by solving the degenerate and extent situations. Since the image origin is not calibrated in Tsai methods, the direct linear method and the calibration with vanishing points are referred. Theycalibrate the image origin but neglect the lens distortion. The situation in practice is that the lens distortion is more sensitive than the image origin to pose estimation and it is difficult to give an initial guess to implement the distortion alignment. Therefore, a direct search algorithm for the image origin is introduced by use of the other camera parameters. Finally, the refinement with nonlinear minimization for all camera parameters comes into the discussing sight.

After the settle down of the mathematical issues in camera calibration, some approaches to online calibration are proposed according to the application environments. The calibration with a robot tool and with a calibration body are the alternative solutions for the robot vision applications. Taking further the application procedure into account, an approach to camera pose calibration with an external measurement system is introduced.

When the applications in industries are given more concerns, the camera recalibration needs to be considered. Since the camera is disturbed by its pose in most of cases, the recalibration is simplified to determine the changes happened to the camera pose. Three recalibration approaches are proposed for checking the changes and determining the corrections of the camera pose in real time.

Eventually, some contributions on vision systems calibration and pose estimation are made according to the applications. Although the application with a mono-camera system and the calibration to a stereo sensor are discussed in details, the dominating target is the multi-camera system. Some valuable approaches, including pattern weight,zero measurement, pattern compensation and security control, to improve the system performances in industrial applications are therefore brought forward.

All the methods and approaches referred in the issertation aim at applying the vision systems accurately and efficiently to the robot vision applications. They relate fruitfully the techniques in laboratory to the industrial applications.