Research On Woven Fabric Defect Real-time Detection Platform Based On Machine Vision

Chinese weaving factory a biggest producer of fabrics in the world.The work on quality control of fabrics should not be underestimated.The final part of fabric weaving is usually the inspection of fabric defects,and then evaluates its grade.In general,fabric defects are closely related to raw materials,weaving process and loom status.Therefore,strengthening the inspection of fabric defects,improving the detection efficiency of fabric defects,and using advanced fabric defect detection equipment are important means for improving the core competitiveness of weaving companies.However,most textile companies in China still use human eyes for inspection.Therefore,it is imperative to develop a new generation of intelligent cloth inspection machines.At present,there are two kinds of automatic inspection platform based on machine vision: The on-line inspection platform installed on the top of the loom roll cloth roller and the off-line cloth inspection platform after the fabric weaving is finished.However,the online inspection needs to overcome the vibration of loom,off-line inspection because of its high cost of work instability and other reasons are not widely used in textile industry inspection.After summarizing the research results of the inspection equipment developer and scholars in recent years,this paper puts a fabric new defect detection platform,it is from machine vision technique.The platform is mainly composed of fabric transmission system,light source and imaging system,image data acquisition,transmission system,image processing and user interface system.The illumination system and image acquisition system are installed on the manual inspection machine.After finishing the fabric image acquisition process,the image is transmitted to the image processing system by using Gigabit Ethernet.After the system receives image information,it processes and detects it in real time.After the fabric defect image was found,the test results were displayed on the human-machine interface.On the basis of reference to previous research work,this paper puts forward the whole design scheme of the real time testing platform,including the design requirements and the principle structure of the test platform.The task and function of the sub module of the test platform are divided.In the platform hardware system construction,the two-motor synchronous control of the fabric transmission system and photoelectric automatic control to the edge ensure the smooth transmission of the fabric.The high resolution CCD camera is combined with the light source and lens to realize the acquisition of the fabric high definition image.Image processing module mainly adopts PC-based CPU + GPU universal heterogeneous parallel processing architecture to complete the high-speed image processing.In the platform software system design,this project completed the image getting program design,image processing program design and UI design.First,an image acquisition program was developed based on the API provided by the camera driver.Secondly,a defect discrimination method based on feature parameter matching is proposed,and the image processing process is realized by using OPENCV.After analysis,it is found that the image filtering and calculation of the mean and variance of sub-images is time-consuming,which is not conducive to the real-time performance of the system.Then CUDA-based image processing technology was proposed to complete the parallel processing of images.The experiment results show that the parallel processing speed is increased 10 times.Finally,the Man-machine interface is developed based on MFC.It includes functions such as camera operation,defect detection,real-time image and defect image display,and fabric defect detection result display.On the basis of the design and implementation of platform hardware and software,the debugging of hardware platform with transmission,illumination,camera and image acquisition and processing module is completed,and the software debugging based on image acquisition,image processing and Man-machine interface is finished sequentially.In order to verify the real time detection platform,a defect identification method based on feature parameter matching is adopted in the algorithm,and the recognition process of image defects is realized based on OpenCV.After analyzing the process of image serial processing,it is found that the conversion of grayscale image to color image,image filtering,especially the calculation of mean and variance of sub images is time-consuming,which is not conducive to the realization of real-time detection platform.Then,the algorithm of fabric image defect recognition based on Cuda is proposed,and the parallel processing of image is completed.After the experiment,it is found that the Cuda parallel processing of image Gray is improved by about 10 times times,and the speed of image filtering is raised to 2-3 times,and the processing speed of the gray mean and variance of the fabric chunking is increased 30 times times.Then using this detection algorithm to complete the actual fabric defect detection experiment of 100 m white fabric,the platform basically meets the real-time detection requirements and detects 8 defective fabric images.In this paper,a new type of surface array CCD camera is developed to carry out the real-time machine cloth inspection platform.This is a good foundation for the design and realization of other fabric defect image real-time processing algorithms,and provides a design idea for the development of commercial automatic inspection system.