Research On Integrated Online Detection Technology Of Medical Plastic Blood Warmer

Injection molded products have the characteristics of easy molding and stable chemical properties,and are widely used in the medical field.However,injection products are affected by many factors such as mold structure,different materials,and molding process,and are prone to defects such as warpage,scratches,pitting,and stress marks.Therefore,it is necessary to detect defects in medical injection products during the production process.This article aims at the current situation of medical plastic warming devices relying on manual detection,with the title of "Research on the integrated online detection technology of medical plastic warming devices",to study the online detection technology based on machine vision,to improve the accuracy and positive rate of product detection The actual production and testing efficiency of the enterprise.The thesis researches the integrated visual online detection platform technology,from three aspects: integrated platform,visual inspection,and online inspection,and discusses the relevant content of the medical plastic warming device comprehensive test platform.The main contents of the thesis are as follows:(1)Develop task requirements according to the production process and testing standards of the warmer,design a visual inspection system module and frame that fits with the warmer nut installation and leak testing equipment,combined with the test platform reserved station size analysis visual inspection Each part of the module analyzes the key technologies of the visual inspection platform for the problem of manual exposure to ultraviolet light sources.(2)Research and design the visual detection hardware system of the blood warmer.According to the low reflectivity of the side surface of the blood warmer,the foreground lighting of the ultraviolet projection lamp after modification and the UV light bar are selected as the light source.The calibration point is introduced to distinguish the front and back sides of the blood warmer.Combined with the actual production testing standards of the enterprise,the visual detection module hardware is built The system cooperates with the leak test beat to acquire the original image.(3)Study the image processing flow of the appearance inspection of the warmer.According to the task of size detection and defect detection,the image processing flow is formulated,the different filter operators are used to highlight the target feature effect,and the convolution kernel highlight operator is used to enhance the image features.The threshold segmentation process was used to extract the target,and the maximum inter-class variance method was improved.The experimental results show that the improvement can improve the operation time of the threshold segmentation algorithm.Morphological processing is used to remove irrelevant pixels to obtain the contour features of the defects on the side of the warmer.The particle analysis method is used to divide the distribution area of connected areas,and the defect classification is initially completed.(4)Study the defect classifier model based on convolutional neural network.According to the traditional particle analysis method,it is easy to misjudge small defects such as black spots and bubbles.Two classic network structures,Google Net and Alex Net,are selected,combined with the image features of the warmer defect,and the number of layers and parameters of the classifier are rationally designed to further improve the defect.The correct rate of classification.After experimental comparison,the defect classifier based on the Google Net model works well,and the recognition rates of genuine products,black spots,bubbles,dirt,lack of glue,and cracked products are: 98.10%,97.84%,98.66%,97.40%,100 %,98.78%.(5)Study the software system of the appearance detection system of the blood warmer.Based on Lab VIEW and NI Vision Assistant vision module,Tensor Flow platform,develop a blood warmer appearance detection software system,store the measured data in the My SQL database in order,and display it on the Navicat for My SQL interface,which is convenient for subsequent query and recheck.Experiments show that the size detection accuracy of the integrated visual inspection system reaches 0.1mm,and the overall defect classification accuracy rate reaches 99.50%.