Research On Surface Quality Inspection Of Carbon Fiber Prepreg Based On Machine Vision

With the development of economy,composite are widely used in war industry,transportation,energy,chemical industry and other fields due to their light weight and corrosion resistance.Carbon fiber prepreg is widely used in industrial manufacturing as an intermediate material in the manufacturing process of composite materials.Due to poor quality of raw materials during production,irregular production operations,and improper post-production transportation,the occurrence of defects is unavoidable,which affects the mechanical properties of the product and leads to a decline in product quality.At present,the defect detection of carbon fiber prepreg is still mainly based on manual visual inspection,which has low efficiency and strong subjectivity.In recent years,machine vision inspection technology has been widely used in defect detection due to its advantages such as fast detection speed,high accuracy,and continuous work.It not only saves labor,but also improves the efficiency of automated detection,and has good application prospects.In this paper,machine vision is used to identify and classify the four common defects of carbon fiber prepreg folds,scratches,cracks,and holes.According to the characteristics of carbon fiber prepreg defects and inspection requirements,the overall plan was designed,and a carbon fiber prepreg surface quality inspection system was built.The detection scheme of fixed-focus lens equipped with industrial area array camera was designed,based on the light source configuration of low-angle illumination combined with backlight illumination,a carbon fiber prepreg image with high-resolution and high-contrast is obtained to realize identification and classification of defects.Aiming at the problem that environmental interference and other factors affect the imaging quality during the imaging process of carbon fiber prepreg,in this paper,the wavelet transform filtering method is used to process noisyimages.The processed image has a higher signal-to-noise ratio,and it has a better denoising effect while retaining the edge contours of the image.An entropy-weighted method of maximum inter-class variance image binarization is proposed,which is based on the method of maximum inter-class variance,and the entropy of grayscale pictures is used as the weight value to calculate the optimal threshold of image binarization,which solved the problem of low recognition precision of scratch defects.Aiming at the problem of low accuracy of defect classification,a feature extraction method based on weighted fusion of HOG features and geometric features of defects in prepreg strip images is proposed.Extract the HOG features of the prepreg image and use the PCA method to reduce the feature dimension to describe the texture gradient features of the prepreg strip image.Calculating the roundness,rectangularity,length-width ratio of the circumscribed rectangle,density and moment features of the defect area as geometric features,and performing feature fusion through weighted fusion.According to the characteristics of the experimental samples,through the comparative analysis of the experiment,a support vector machine is selected as the classifier.The experimental verification shows that the accuracy of classification of four defects of wrinkles,scratches,cracks and holes is high by using two feature weighted fusion methods.The designed inspection system can replace manual inspection while improving the efficiency and accuracy of defect detection,which has certain practical significance and reference value.