天津职业技术师范大学学报

利用平行双目相机模拟人的双眼获得真实世界物体距离信息的思想,在PC机上搭建一个基于MATLAB和OpenCV的双目立体视觉测距系统。讨论了相机校正,消除左右图像畸变,实现了左、右图像共面行对准;应用BM和SGBM算法获得深度图并最终获得了目标到摄像机的距离。该系统只需普通的平行双目相机采集左右图像,降低了成本。为满足短距离应用且获得较快速度匹配的要求,采用了BM和SGBM匹配方法,并对2种匹配方法进行了对比。结果表明:在物距1 m以内,2种匹配算法获得的距离相对误差均小于3%。

In this research,parallel binocular cameras were used to simulate human eyes to obtain information about distance in the real world.A binocular vision distance measurement system was established based on MATLAB and OpenCV on PC.The paper discussed camera correction to eliminate image distortion and make the left and right images coplanar line alignment.The paper also calculated the depth map through BM and SGBM algorithms and ultimately obtained the distance between the object and the camera.The system only need the ordinary parallel binocular cameras to collect the left and right images,greatly reducing the cost.In order to meet the requirement of the short distance application and the fast speed matching,this paper adopted BM and SGBM matching methods and compared these two methods.Results from the experiments show that the relative errors of the two matching algorithms are less than 3%,if the object distance is within one meter.