摘要
自主车上的立体视觉系统一般由两台固定在平台上的定焦摄像机组成,因此摄像机内参数经一次标定后不再变化,只需要考虑外参数的标定.本文针对自主车视觉系统的特殊应用情况,提出一种基于多尺度几何分析思想的摄像机对弱定标算法,该算法采用Contourlet变换对左右图像中的角点进行检测,利用Hartley规范化8点法估计摄像机对的基础矩阵.依托现有的摄像机内部参数标定工具箱,在摄像机对弱定标的基础上还可以快速地获得摄像机对之间的外参数矩阵.实验结果表明该方法具有较好的精度.
In practice, the stereo system is usually constructed by two fixed cameras on top of mobile vehicle view the scene from a left and right point of views. Due to lens distortion and camera displacements, both cameras need to be calibrated. The internal camera parameters and the orientation of the cameras to each other are assumed to be settled. For the special application of mobile vehicle, a new multiscale geometric method for binocular weak calibration is proposed. The aim of multiscale geometric analysis is to find a kind of optimal representation of such type of image in the sense of nonlinear approximation. In this paper, a novel explanation is given for contourlet transform and detecting corners of images by the multiscale geometric method. Finally, the eight-point algorithm of Hartley has been used to determine the fundamental matrix. It is important that the fundamental matrix predigests the following problems, such as, the intrinsic parameters matrix, the motion parameters, and structure recovery from uncalibrated stereo rig. Experimental results show the efficiency of this algorithm.
出处
《小型微型计算机系统》
CSCD
北大核心
2007年第6期1115-1118,共4页
Journal of Chinese Computer Systems
基金
国家自然科学基金项目(60453001)资助.
