Biography
Enrollment Date: 2011
Graduation Date：2014
Degree：M.S.
Defense Date：2014.05.27
Advisors：Xiang Xie
Department：Institute of Microelectronics，Tsinghua University
Title of Dissertation/Thesis：Accurate Intrinsic and Extrinsic Parameters Calibration Method for Pro-Cam System
Abstract:
The common projector-camera system, as a structured light system, is widely used in reverse engineering, culture heritage and quality control etc. Recently, with the development of the projector technology and computer vision, the projector-camera extended its application into human-computer interaction and augmented reality etc. Generally, in order to obtain desirable results, the parameters of the projector-camera system need to be calibrated in advance.
In this thesis, a flexible method with high accuracy is proposed to calibrate the projector-camera system. For this issue, the major challenge is the accurate acquisition of coordinates’ correspondences between the image plane of the projector and the calibration board. However, the pattern of the calibration board would interfere the extraction of the projected pattern, decreasing the final calibration accuracy To solve the problem, a quad-saddle based pattern generation and extraction method is proposed to compensate the detection error. Moreover, a Baysian classifier based filter is designed to further improve the accuracy. Compared with the traditional calibration method, the proposed method improves the calibration accuracy greatly. Finally, the experiment results show that the maximum value of reprojection error for projector is 0.3102 pixels, and the RMS reprojection error is 0.1028 pixels. The proposed calibration method achieved the most accurate results among the published papers.
In addition, a triangulation method for the projector-camera system is proposed, after a detailed analysis of traditional triangulation methods. A projector-camera based human-computer system is also modeled in this thesis. The triangulation method for this system is proposed and the theoretical accuracy of the system is analyzed.
Finally, this thesis presents a real-time and adaptive projection surface calibration method, which can be used in keystone correction and human-computer interaction. The keystone effect would occur when the axis of the projector is non-perpendicular to the projection surface, decreasing the visual comfort. So the information of the projection surface needs to be calibrated to correct the keystone effect. Moreover, the position and the pose of the projection surface are needed sometimes, in order to recognize the human-computer interaction in the projector-camera based human-computer system. Thus the parameters of the projector plane needs to be calibrated first. The proposed method can calibrate the projection surface automatically and does not require any tag information on the projection surface.