Research On Calibration And Reconstruct Of Stereo Vision For The Lunar Rover

To ensure the safety of cruising on the lunar surface for lunar rover, a stereo vision system, which is used to provide the terrain information of surrounding, is installed on the lunar rover. Camera calibration and 3D reconstruction are important parts of stereo vision, and their precision is critical to the performance of the whole system. In this paper, we deal with these two steps for application on lunar rover.In the stereo vision system of a lunar rover, the parameters of the cameras are fixed in the working and the requirement of precision is relatively low. Therefore conventional calibration technique is selected after analyzing the merits and defects of various kinds of calibration methods. In order to improve calibration precision without increasing computation complexity, Tsai's two-stage calibration technique with calibration points in multiple planes is refined. This improved algorithm is proved to be computationally simple and easy to implement.After camera calibration is completed, the camera parameters will be used to compute the 3D coordinates of scene points together with stereo matching results based on triangulation. To date only discrete points are obtained, but not the surface of the scene. Hence a least-square method is used to fit terrain surface. Besides providing the surface of the scene, the fitting procedure helps to filter out some large reconstruction errors.The calibration and reconstruction device with fairly high accuracy is designed and constructed. The experiments with our calibration and reconstruction algorithm on the device produce satisfactory results. Experimental data indicates that the calibration and reconstruction method presented in this paper are feasible for application on a lunar rover. The research possesses important reference value to the design of stereo vision system for lunar rovers and other outdoor mobile robots.