平面测量问题中单应矩阵的算法研究

在平面测量问题中,单应矩阵扮演着十分重要的角色。同计算机视觉中的其他问题一样,单应矩阵的计算对数据噪声极其敏感。本文系统地研究了求解单应矩阵的三种算法。首先分别给出了各种算法的理论推导;其次对每种算法都做了数值模拟实验,并分析和比较了实验结果。结果表明基于直线对应的算法比点对应算法更具有鲁棒性。最后我们还给出了真实图象的实验结果。

Research on Algorithms of Homography Matrix in Plane Measuring

The homography takes an important part in plane measuring. As same as other problems in computer vision, the homography matrix computation is very sensitive to noise. In this paper, the homography matrix algorithms are studied thoroughly. First, we derive the algorithms and list the steps. Then we present the simulative experiments to validate every algorithm. We also compare and analyze the results of these experiments. The results indicate that the algorithm based on line homography is more robust than the algorithm based on point homography. In the end, we display the results of the experiments using true images.