Research On Yarn Detection And Tracking System

With the continuous advancement of Industry 4.0 technology and the"Made in China 2025",the "machine substitution" intelligent production has become the actual demand of the textile industry,in order to enable industrial robots to accurately capture different colors in motion during online operation.The yarn tube,this subject is based on the yarn tube sorting and feeding process in the textile industry.Based on the machine vision technology,the yarn tube detection and tracking system is studied.The main tasks are as follows:(1)A monocular vision system experimental platform is built.Considering the requirements of the subsequent yarn tube detection and tracking system,image preprocessing and camera calibration techniques are studied in the early preparation stage.The industrial camera is calibrated by Zhang Zhengyou’s chessboard calibration method at the same time,so as to obtain the camera’s internal and external parameters,establish the transformation relationship between the image coordinate system and the spatial coordinate system.Fourteen images were collected by the moving calibration board to calibrate the camera.Finally,the validity of Zhang Zhengyou’s camera calibration method was verified by the analysis of the position error of the re-projection point and the accuracy requirement of the system was satisfied.(2)In the moving object detection module,aiming at the problem that the traditional moving object detection method based on Mixture Gauss Modeling takes a long time to build the background and can not recognize the motion information of different color yarns,this thesis combines the three-frame difference method with the Mixture Gauss Modeling method to build the Gauss model by dividing the average of the background area into blocks,and uses different update factors to speed up the background updating rate.The experimental results show that the running time of the improved algorithm only accounts for 74.95% of the traditional algorithm on the premise of the same number of frames.Combining the algorithm with the color recognition module,the moving object information of different color yarns can be extracted.(3)In the moving target tracking module,for the problem that the Camshift target tracking algorithm has poor tracking performance when the target bobbin is occluded,similar color interference and illumination changes,this thesis adds color recognition motion based on the traditional Camshift target tracking algorithm.The target detection algorithm and Kalman prediction mechanism enable the improved algorithm to implement the automatic initial tracking window,avoid manually setting the error generated by the tracking window,and predict the position of the next frame of the target object to improve the stability of the target tracking.Through experimental tests,it is verified that the improved algorithm can effectively track the tube of a specific color under the condition of target occlusion,similar color interference and illumination changes.