Research And Application Of Machine Vision Inspection Of Fiber State In Braiding Process

With the wide application of braiding products in various fields,various problems arising from braiding have become the focus of research by many scholars.At present,the abnormal state of fibers in the braiding process is mostly identified by manual observation,and a large number of flying flocs that are harmful to workers’ bodies will be generated during the braiding process,and the labor cost of labor is high.By using machine vision technology in the braiding process,the harm to workers’ health can be reduced and the quality of knitted preforms can be improved.Based on this,this paper mainly uses machine vision technology to detect the fiber state on the braiding spindle and the preform during the braiding process,thereby improving the braiding quality.Firstly,by consulting related braiding literature,understanding the current development status of domestic and foreign braiding technology and the main problems to be solved,the development of machine vision technology and the current status of using it to solve fiber and fabric defects are studied and analyzed.Secondly,the current production demand of the braiding factory was investigated on the spot,and the overall design plan of the fiber condition detection system was determined.The selection of the camera,light source and other hardware used in the subject was introduced and purchased.Design and install fixed fixtures of camera and other hardware devices by 3d software Solidworks.Thirdly,the algorithm flow of fiber abnormal state detection on braided spindle is described,and several filtering algorithms in the image processing process are introduced at the same time,several existing image filtering algorithms are analyzed and the best filter is selected by calculating the index of image quality.The normalized cross-correlation matching algorithm,the secondary positioning based on image segmentation,the geometric feature detection algorithm and the integral projection algorithm are used to identify and detect the three abnormal states of fiber breakage,fiber entanglement,and fiber ring that appear on the braiding spindle,and use the built platform for experimental verifications.Next,A multi-layer perceptron model based on LAWS texture filter and gray cooccurrence matrix feature vector was used to detect the abnormal fiber state of braided preform.The training and experimental verification of different algorithms are accomplished,and it is concluded that the multi-layer perceptron model has a high accuracy rate and realizes the detection of abnormal fiber state of the braided preform.Finally,by integrating different image processing modules through requirement analysis,a host computer system is developed,which can directly and accurately reflect the detection results.