Research And Application Of Medical Natural Biofilm Defect Detection Based On Machine Vision

Millions of people suffer skin damage each year from accidents or diseases of their own.The application of biofilm and other dressings has greatly improved the cure rate of skin injury,which preserves the pore structure conducive to the survival and reproduction of skin cells,and has been marketed and applied clinically.With the development of modern medical technology,the treatment of skin injury is increasingly dependent on advanced biological materials.Biofilm has gradually become a commonly used material for skin transplantation,and the corresponding demand for medical biological materials is also increasing.The medical natural biofilm(hereinafter referred to as biofilm)used in this study is a dry feed material,which is a kind of acellular dermal matrix derived from cow skin.The pores of the acellular matrix are naturally existing and similar to animal tissues,which is conducive to skin wound healing.Biofilm has good skin repair properties,and its product quality requirements and production scale are increasing year by year.At present,the production of biofilm requires preliminary treatment,cleaning,decellulation,washing and disinfection,drying,cutting and other processes,each of which may cause appearance defects of biofilm.The size of one side of the biofilm ranges from 20 to100 mm.Different sizes of the biofilm have certain dimensional accuracy requirements,so it is necessary to carry out dimensional detection of the finished product of the biofilm.At present,the detection of biofilm still relies on manual visual inspection,and the detection workers are easy to fatigue,and easy to produce subjective assumptions,resulting in inaccurate detection results.Machine vision detection can completely avoid this situation,it has the advantages of fast and simple,small error range,free from subjective factors and can store data.Machine vision detection has been gradually popularized in the application of paper and leather appearance defect detection,and the detection results are accurate and stable,more reliable,and higher application value.The surface characteristics of biofilm are similar to paper and leather.Therefore,this topic uses machine vision to detect the size and appearance defects of biofilm.The main research contents are as follows:(1)Select the detection system according to the size and appearance of the biofilm.The software system of this project uses Vision Pro of Cognex Company,which has powerful visual detection function and many built-in detection cases,so as to facilitate users to detect various defects.The hardware system selected the German Basler ac A2440-20 gm Gig E camera,which is a black and white industrial camera with up to 5million pixels.The lighting system used LED lights,the biofilm size and hole perforation defects were detected by backlight,and other defects were detected by coaxial light source.The calibration board used homemade 5×5mm checkerboard.In order to help enterprises develop appropriate detection methods,this study analyzed the appearance characteristics of biofilm,designed a reasonable detection process and specific detection methods.(2)The biofilm image should be preprocessed before the defect and size detection,and the average value of PSNR,LPIPS and SSIM was compared to evaluate the effect of mean filtering,median filtering and Gaussian filtering on the biofilm image;Image enhancement improves the contrast of biofilm defects and makes them more intuitive and obvious.After image binarization processing,the amount of biofilm image data is reduced,so that the outline of the biofilm is more clearly visible,convenient for size measurement,and improve the detection efficiency.(3)Biofilm size detection results showed that 30 biofilms were randomly selected from 300 biofilms for length and width measurement,and the variation ranges of length and width size errors were as follows: [-0.147 mm,0.290mm],[-0.21 mm,0.188mm],12 of the 120 circular biofilms were randomly selected to measure the radius size,and the radius variation range was as follows: [-0.169 mm,0.253mm],the dimensional accuracy obtained by using machine vision dimensional measurement method is within 0.3mm,within the allowable error range.(4)The detection of internal defects of biofilm adopts the difference imaging method,which uses the difference between the preprocessed image and the fuzzy image to display the location of defects.The area of black spots is between 0.1～0.2mm2,and the area of flocs is basically between 0.2～0.65mm2.The defects can be classified according to the defect area.The defect is larger and easier to detect.The hole penetration detection using a coaxial light source is easy to be confused with black spots.The backlight detection will show bright spots with an area between 0.1 and 0.2mm2.The characters on the back image are easier to detect than those on the front image.Template matching method can find the characters quickly and accurately.The testing time and accuracy of the biofilm system were tested,and the results showed that the biofilm size and defect detection speed were 7 sheets per second.The accuracy rate of biofilm defect,pore penetration and size detection can achieve 100%,and the black spot and floc can reach more than 98%.The biofilm detection system can meet the requirements of biofilm size measurement and defect detection.This topic provides a basis for proving that machine vision detection of appearance defects of biofilms can replace manual work.In industrial production,timely discovery and effective treatment of unqualified biofilms is a key measure to ensure product quality and enhance enterprise competitiveness.