• •    

基于立体视觉的高精度标定与测量方法研究

孔颖乔,赵健康,夏轩   

  1. 上海交通大学
  • 收稿日期:2016-11-07 修回日期:2017-01-10 发布日期:2017-01-10
  • 通讯作者: 孔颖乔

Research on High-precision Calibration and Measurement Method Based on Stereo Vision

  • Received:2016-11-07 Revised:2017-01-10 Online:2017-01-10
  • Contact: Ying-Qiao KONG

摘要: 立体视觉测量系统中,光学系统产生的畸变使目标的成像偏离了理论成像点,导致系统产生测量误差。为了提高系统的测量精度,根据标定板上各角点的像素分辨率,拟合整个成像平面的四次多项式,且多项式的系数与物体到相机的距离成比例。应用双目测距原理,测量被测物体的纵向距离;基于所得的多项式,应用单目相机测量待测物体的横向尺寸。实验结果表明,用所提方法,当物体距离相机10m以内时,纵向距离误差可以减小到5%以内;当物体距离相机1m时,横向宽度测量误差在0.5mm内,逼近理论最高分辨率。

关键词: 标定, 像素分辨率, 畸变方程, 拟合, 双目测距, 单目测宽

Abstract: In stereo vision measurement system, distortion caused by the optical system makes images of targets deviate from theoretical imaging points, which results in measurement error of the system. In order to improve the measurement accuracy of the system, a quartic polynomial on the whole image plane is obtained through fitting the pixel resolution of each angular point on calibration board, the coefficient of which is proportional to the distance from the object to the camera. Then, we utilize binocular vision technology to calculate the distance of the detected object, and monocular vision to measure its transverse dimension. Experimental results show that, when the distance between the object and the camera is within 10m, the distance error can be reduced to less than 5%. And when the object is 1 meter away from the camera, the measuring error of the transverse width of the object is within 0.5mm, which approaches to the theoretical highest resolution.

Key words: calibration, pixel resolution, distortion equation, fitting, measuring distance by binocular model, measuring size of monocular vision

中图分类号: