基于双目立体视觉的纱线运动参数识别

Motion parameter identification of yarn based on binocular stereo vision

针对目前运用传统接触方法难以准确、无干扰测量纱线运动参数的问题,提出基于双目立体视觉的纱线运动参数识别方法。采用双目高速相机采集运动纱线图像,开发双目高速同步抓帧存图程序;利用高精度标定板进行双目标定;经滑动窗口和线面交点方法定位纱线测试点;对极线校正后严格行对准的像对使用块匹配、原始和改进半全局块匹配等立体匹配算法计算深度图,并采用视差原理求得纱线测试点空间位置坐标;结合存图时刻求得纱线运动参数。实验表明,本文方法得到的纱线运动参数中位移误差均值为0.132 mm,与实际情况吻合良好。

Aiming at the problems of low accuracy and no interference of measuring yarn motion parameters for traditional contact method,a binocular stereo vision method was proposed:Images were collected by bin-ocular high-speed camera,stored by the developed software of binocular high-speed synchronous frame cap-ture;The binocular camera was calibrated by high precision calibration board;The yarn test point was de-tected by sliding window and intersection;After polar correction,the depth map was calculated by stereo matching algorithms of Block Matching,Semi-Global Block Matching and improved SGBM algorithm.The spatial location information of yarn test point was obtained by parallax principle.The yarn motion parameters were obtained with the three-dimensional coordinate and image storage time,the average error of displace-ment was 0.132 mm.The experimental results showed that the yarn motion parameters were in good agree-ment with the actual situation.