论文部分内容阅读
双目立体视觉是光学被动三维测量的主要方法。为了快速、准确地寻找对应点 ,通常要将立体图像对进行校正 ,消除垂直视差。提出了一种无需标定相机的立体图像对校正方法。该方法从基本矩阵计算初始透视投影、旋转和竖直平移变换矩阵 ,然后以对应点坐标为基础对这些变换矩阵进行优化计算 ,从而有效地避免了优化计算的局部最小值 ,而且不过分依赖基本矩阵的计算精度。通过对提出的方法和完全基于基本矩阵的图像校正方法以及无需基本矩阵计算的图像校正方法进行实验比较 ,结果表明提出的方法图像校正速度快 ,能有效地消除垂直视差 ,而产生的图像变形较小。
Binocular stereo vision is the main method of optical passive three-dimensional measurement. In order to quickly and accurately find the corresponding point, the stereoscopic image pair usually needs to be corrected to eliminate the vertical parallax. A calibration method of stereo images without calibration camera is proposed. In this method, the initial perspective projection, rotation and vertical translation transformation matrices are calculated from the basic matrix, and then the transformation matrices are optimized and calculated based on the coordinates of the corresponding points, so as to effectively avoid the local minimum value of the optimization calculation and not excessively rely on the basic Matrix calculation accuracy. The experimental comparison between the proposed method and the image correction method based on the basic matrix and the image correction method without the basic matrix show that the proposed method can quickly correct the image and correct the vertical disparity effectively. small.