一种基于改进ORB算法的双目视觉测距方法

Binocular vision ranging method based on improved ORB algorithm

针对传统基于特征点图像匹配算法误匹配率高和双目视觉系统测量精度低问题,提出一种基于改进ORB算法的双目视觉测距方法。首先采用引导滤波对图像进行预处理,然后将FAST检测得到的特征点作为中心,得到围绕该特征点的区域,将该区域进行分割,分别求出左右区域的质心坐标,根据两质心坐标进行主方向的确定;再采用Sobel算子计算出像素点水平和垂直梯度,比较每个像素点对之间水平与垂直梯度中的较大值,把比较结果串成一个二值位字符串的形式,从而形成描述符,采用汉明距离进行匹配;最后在测量阶段用二维二次函数拟合的方法获得特征点的亚像素坐标,通过三角测量原理获得对应特征点的空间三维坐标,从而得到被测物体的尺寸。实验结果表明,本文改进算法的匹配精度较传统ORB算法提高35.25%,同时测量的最低相对误差达到0.428 4%,满足测量要求。

In the traditional image matching algorithm based on feature points cause a high matching error rate,and the low measurement accuracy of binocular vision system,so binocular vision ranging method based on improved ORB algorithm was proposed.First,guided filtering was used to preprocess the image,then the feature point detected by FAST was used as the center to get the area surrounding the feature point,the area was divided,the centroid coordinates of the left and right areas was obtained respectively,the main direction was determined according to the two centroid coordinates;then the Sobel operator was used to calculate the horizontal and vertical gradients of the pixels,the larger value of the horizontal and vertical gradient between each pixel pair was compared,finally,the comparison results were stringed into a binary string form to form descriptor and the Hamming distance was used to match;Finally,in the measurement process,the two-dimensional quadratic function fitting method was used to get the sub-pixel coordinates of the feature points,and the three-dimensional coordinates of the feature points were obtained through the principle of triangulation,so the size of the measured object was obtained.Experimental results show that the matching accuracy of the improved algorithm in this paper is 35.25%higher than that of the traditional ORB,and the minimum relative error of measuring reaches 0.4284%,which meets the measurement requirements.