Research And Application Of Fabric Defect Detection And Recognition Based On Deep Learning

In the actual production process of fabric,due to the influence of factors such as production equipment failure and improper operation,it is easy to cause creases,pinholes and other defects.In order to ensure the quality of fabric products,it is necessary to detect and recognize the fabric defects,which has become an important step in the production and quality management of textile enterprises.However,at present,most textile enterprises rely on manual to detect whether there are defects in the fabric,which not only has the problems of low detection efficiency and lack of consistency,but also easily restricted by physical ability and influenced by subjective factors,resulting in missing detection and wrong detection.In view of the above reasons,it is of great significance for textile enterprises to design and develop a system that can quickly and accurately detect and recognize the fabric defects.For the above problems,this paper proposes the SE-Inception v4 fabric defect detection and recognition model combined with deep learning and image classification technology,and develops a web-based fabric defect detection and recognition system based on this improved model,so as to help textile enterprises improve the efficiency and accuracy of fabric quality detection.The research contents and achievements of this paper are as follows:(1)Research on the technology and theory of fabric defect detection and recognition.The theory and technology of fabric defect detection and recognition based on traditional machine learning and deep learning are studied in this paper,and some representative research results in fabric defect detection and recognition are listed.LBP algorithm,classification network of Res Net,Dense Net and Inception,deep learning development framework Tensor Flow are mainly introduced.(2)Collection and augmentation of the fabric image data set.In this paper,geometric transformation,scale transformation and LBP algorithm are used to enhance the data set to meet the basic needs of model training.(3)Construction and improvement of the fabric defect detection and recognition model.In this paper,two traditional machine learning classification methods and five classic deep learning networks are used to build fabric defect detection and recognition models respectively,and the Inception v4 model with the best performance in each evaluation index is selected for optimization and improvement.The SE module is added to the Inception v4 model to realize the attention mechanism.Compared with the experimental results,the accuracy,precision and other performance indexes of the SE-Inception v4 model have been improved.(4)Design and development of fabric defect detection and recognition system.Based on the SE-Inception v4 model,this paper designs and develops a web-based fabric defect detection and recognition system.According to the method of software engineering,the feasibility analysis,requirement analysis and architecture design of the system are carried out,and it is designed and implemented in detail.The system is divided into two modules: client and management.The client includes registration and login,image recognition,image detection record and personal center.The management includes user management and model update.