用于单目视觉的路径失真矫正系统设计与实现

Design and implementation of path distortion correction system for monocular vision

针对单目视觉机器人产生的路径图像失真问题,分析了产生失真的原因及常用的矫正方式,提出了一种可以通过快速标定来解决路径图像畸变失真问题的方法。将待矫正图像的纵向与横向像素序列分别与标定板进行分析对比,纵向上采用非均匀行采样原理,通过单张标定照找到特征采样行,使采样后图像与实际道路相对无畸变;横向上采用线性伸缩矫正算法,结合适当的插值运算,达到横向矫正。该方法避免了要多次测量外部标定数据及内部传感器参数,在运算量上与实用性上取得了较好的平衡。在此基础上,介绍了如何利用Matlab工具箱快速设计实现图像矫正系统,并且通过数据对比证明了该系统具有一定的准确与便捷性,在工程上具有较高的价值。

In order to solve the path image distortion problem in monocular vision robot, it analyzes the causes and com-mon methods of correcting distortion, and presents a quick method to solve the problem of image distortion through cali-brating image coordinate. It analyzes and compares separately the vertical and horizontal pixel sequence in the image to the calibration plate. And in the vertical, it captures a digital image and processes in the principle of non-uniform row sam-pling. Then it searches for the characteristic sample rows of original image, by which the new sample image is formed. Thus the new image is relatively undistorted with the original one. In the horizontal, it uses the algorithms of linear scal-ing and appropriate interpolation to make the correction. This method avoids measuring some extern calibration data and inter parameters in sensors repeatedly, and has a great balance in calculating amount and practicality. On these basis, the paper introduces in detail how to use Matlab toolboxes to design and realize a image correction system quickly, and dem-onstrates that the designed system is accurate, simple and highly valuable in engineering.