基于亚像素级角点检测的双目视觉距离测量

Binocular range measuring based on sub-pixel corner detecting

要提高双目测距的精度,需要高精度的视差,为此引入亚像素检测方法。首先,采用Bouguet立体校正算法将左右图像进行极线校正,以减少匹配计算代价,提高匹配效率;然后,利用Harris角点探测器在一矩形领域内寻找一个最强特征点作为待匹配点,再采用最大互相关法对左图像进行匹配搜索确定匹配点。最后,对已匹配好的点对进行亚像素级检测,即可得到亚像素级视差。实验结果表明,该方法能将角点检测精度提高到0.001像素,测距精度也相应提高,距离越远效果越明显。

The improvement of the range-measuring accuracy depends on the high precise parallax,so this article introduces a sub-pixel detecting method. At first,it uses bouguet algorithm to rectify the left and right images to make their epipolar line forward-parallel so that decrease the computation and improve the matching efficiency. Then it uses Harris corner detector to find a most characteristic corner as candidate matching point. After that,it uses most cross correlation matching rule to match left image. Finally,it detects the matched points at sub-pixel level to get the sub-pixel parallax. the result shows that,the method used in this article can improve the corner detecting precision to 0. 001 pixel,and the range-measuring accuracy be improved accordingly,the farther the targets＇ range,the more evidently the improvement of precision.