一种基于边缘线的三目立体匹配方法

Trinocular stereo matching method based on edge segment

为实现准确的三维场景匹配,提出了一种基于边缘线的三目立体匹配方法。Canny边缘是一种常用的视觉特征,通过对Canny算子加以改进,提高了边缘线条匹配的可靠性。匹配过程分层次进行,首先通过边缘上少量点的成功匹配确定边缘线条间的对应关系,然后以此来约束该边缘上其他点的匹配。详细介绍了对应特征匹配所用的约束条件。首先,使用三目系统中第三个摄像机提供的额外极线约束,有效地减少了误匹配。由于常规所用的三目极线约束条件给出的匹配效果并不理想,提出了另外一种更加有效的三目极线约束应用方法。此外还介绍了边缘点的梯度方向约束,给出了基于以上约束条件的边缘线匹配算法。实验结果表明,该算法具有较高的匹配正确率,是一种有效的立体匹配算法。

To realize accurate 3D scene matching, a trinocular stereo matching scheme based on edge segment was developed. Canny edge was one of most popular vision features in image, and improvement of canny operator made the matching of line segments more reliable. Matching process was conducted hierarchically. Correspondence between line segments was firstly validated by successful matching of a few points on edges, which in turn instructed matching of other edge points. Constraints employed in matching of corresponding feature were discussed in detail. Firstly, the extra epipolar constraint given by the 3rd camera in trinocular vision was applied to reduce mismatch effectively. Since conventional matching methods utilizing trinocular epipolar constraint performed faultily, another practical trinocular constraint treatment method was presented. In addition; gradient direction constraint of edge point was introduced. Matching algorithm based on aforementioned constraints was proposed. Experimental results show that the proposed algorithm has the high probability of correct matching.