基于极线校正的亚像素相位立体匹配方法

Sub-pixel matching method based on epipolar line rectification

针对双目结构投影光栅相位法三维测量系统中立体匹配的效率与精度问题,提出了基于极线校正的亚像素相位立体匹配方法。首先,通过外极线校正算法将双目立体视觉几何结构校正为平视双目标准几何结构,使匹配点对在左右相机上的成像点位于同一水平线上。然后,采用双线性内插法获取校正后新相位图像对,并提出一种基于相位区域匹配的立体匹配算法,获得初始匹配点。最后,提出采用最小二乘拟合将初始匹配点附近3×3邻域内的相似度量拟合为一个二次曲面,然后求解得到该曲面的局部最小值,得到与待匹配点相位相同的右摄像机匹配点坐标。通过对标准平板平面度进行测量与精度分析,表明该方法能够快速、准确的实现立体匹配。

In order to improve the efficiency and accuracy of the projected grating three-dimensional measurement system, a sub-pixel matching method based on epipolar line rectification was proposed.Firstly, the epipolar line rectification algorithm was used to transform the bi-stereo geometry into the epipolar standard geometry, so the certain point on the left image had the same horizontal epipolar line with the same row coordinate on the right image. Then, the new phase images of left and right camera were achieved by bilinear interpolation algorithm, and the initial matching points were obtained using the proposed method of stereo matching algorithm based on phase region matching. Thirdly, the local minimum of the surface was achieved by the least squares fitting of 3 ×3 neighborhood similarity metric of the initial matching points to be a quadratic surface. Then the right camera matching point coordinate was acquired which had the same phase with the left camera matching point. The planeness of a standard planar was measured to evaluate the accuracy of the proposed method. The results show that the proposed method can achieve the stereo matching fastly and accurately.