Research On Real-time Detection Of Fabric Defects Based On Machine Vision

Textile testing is the core link of quality control in textile enterprises,which directly affects product quality and economic benefits of enterprises.For a long time,there are some problems such as poor accuracy,low efficiency and easy fatigue.In order to overcome many disadvantages of manual detection,the real-time,automatic and intelligent detection system is the inevitable trend of future development.Nowadays,the real-time testing equipment is mainly imported from abroad,the procurement and maintenance cost is high,and the applicability to China’s textiles is poor,so it is very necessary to develop the real-time fabric fault detection system in China.This paper for the above problems,starting from the enterprise practical use,using machine vision and image processing technology,design a fabric fault real-time detection system,complete the hardware platform construction and software operation,fault detection algorithm,detection real-time integration function of the fabric automatic real-time detection system,realize the fabric fault real-time,automation and intelligent detection.The specific research work of this paper is as follows:(1)Based on the actual demand of real-time detection and production of fabric defects,a hardware platform system suitable for real-time detection of fabric defects is built first.According to the requirements of the real-time detection system,the hardware architecture is built,select the appropriate hardware components,mainly including the camera,light source,lens,flat mirror and photoelectric sensor components.In order to collect good fabric images,and then design the fabric image acquisition area according to the selected components,the collection area provides different lighting methods to meet the shooting needs of different thickness of fabric,multiple cameras are designed side by side to meet most of the fabric detection width requirements.Ststatic and real-time detection software interface is designed.(2)Study the real-time detection algorithm of fabric image,and propose an improved ITTI significant model.The acquired fabric images were first preprocessed and the image noise was eliminated using a Gaussian smoothing method.Then,the fabric image pyramid is extracted by the downsampling method,and the brightness features are extracted by the central peripheral difference operation,and the image direction features are effectively extracted by the Gabor filter operation.Finally,the fault point significant map is obtained through feature fusion,and when the adaptive threshold image is used to segment the fault point image,the experiments show that this method can effectively detect the oil stain,meridian fracture and hole breaking defects.(3)Integrate and debug the real-time detection system of machine vision.Different functional modules are designed according to the system functions to ensure the rationality of the detection system operation.Block processing of the images collected by the system,use the fault detection algorithm to compare the time optimization effect of small blocks,verify that the multi-threaded image block parallel detection method can improve the real-time detection speed of fabric detection;integrate the operation software and system hardware,debugging and realize the real-time detection of fabric defects under machine vision;For fabric fault real-time detection research,this paper proposed machine vision fabric fault real-time detection system,in the hardware construction and software operation task,complete the fabric fault real-time detection by embedded fast fabric fault identification algorithm,with high detection effect and real-time degree,to meet the actual needs of enterprise production detection.