Research And Development For Automatic Fabric Inspection System Based On Machine Vision

The defect is the main factor affecting the quality level of the textile.The defect detection is an important part of the quality management of the textile.Although the volume of textile industry in China is large,the degree of automation is still not high.The current detection of fabric defects is mainly manual,which is inefficient and unfavorable to the physical and mental health of workers.Therefore,the use of automatic fabric inspection system is a necessary way for textile mills to increase production efficiency,save labor costs,and become an industrial transformation and upgrading.Nowadays,the automatic detection of fabric defects has always been a hot issue for researchers at home and abroad.This paper presents an automatic fabric inspection system based on machine vision and adopts the overall design plan of IPC and PLC.At the same time,the system can be divided into four subsystems: main control system,visual inspection system,transmission system and automatic marking system.In the structure of the article,the overall design and development of the system,the design and development of the visual inspection system,the defect detection algorithm,and the research and development of the defect classification algorithm are introduced in turn.This paper first introduces the overall design of the system and the visual inspection subsystem design.Among them,the overall design of the system gives the overall framework and workflow of the automatic cloth inspection system.The visual inspection subsystem details the soft and hardware structure of the visual part.Then,this paper focuses on the detection algorithm and classification algorithm of fabric defects.A defect detection algorithm based on Gabor wavelet and a defect classification algorithm based on convolutional neural network are proposed.Among them,the defect detection algorithm uses the Gabor wavelet to adaptively detect all kinds of flaws in the fabric,and improves the real-time performance of the algorithm under the premise of satisfying its accuracy and versatility.The defect classification algorithm uses more advanced convolutional nerves.The network conducts independent learning on various types of defects,avoids the influence of artificialinterference,and has good feature learning ability and stronger robustness.At last,it summarizes the work of the full text,at the same time it puts forward the existing problems and the future direction of work.The automatic cloth inspection system described in this paper is still under research and development.Although some results have been achieved,there are still many places that need to be improved in order to meet the needs of actual cloth inspection.