Research And Implementation Of Automatic Cutting And Packaging System For Blood Glucose Test Strips Based On Machine Vision

According to the Global Diabetes Overview released by the International Diabetes Federation,the number of diabetic patients in the world has reached nearly 425 million by 2017.However,China's diabetes patients account for 1/4 of the total number of patients worldwide,and it is the country with the most diabetes.Blood glucose test strips,as an important tool for blood sugar content testing,are used in large quantities.Therefore,it is of great practical significance to study and improve the production efficiency of blood glucose test strips.In the production process of blood glucose test strips,it is necessary to carry out cutting test strips,rejecting unqualified test strips,and bottling test strips.Accordingly,this paper analyzes the characteristics of blood glucose test strip production lines,and develops a set of blood glucose test strip automatic cutting and packaging system software based on machine vision technology.The main contents of this paper are outlined as follows.(1)A solution for the automatic cutting and packaging system of blood glucose test strips is designed.The hardware system includes camera,light source selection and lighting scheme design,and the software system includes software framework design,software flow design,etc.(2)The principle of low contrast surface defect detection for uniform surface images is studied.The detection process is mainly divided into three steps.In step 1,the image is firstly converted into frequency domain by Fourier transform,then the image is filtered by Gauss filter in frequency domain,and the image is reconstructed back into the space domain.Finally,the reconstructed image can make a good estimation of the image background.In step 2,the original image is subtracted from the reconstructed image,and then the subtracted image is transformed linearly to enhance the defect features significantly.In step 3,the defect is identified by an unbiased curve structure extraction algorithm.(3)The solution of strip alignment with accuracy of 0.05 mm is given.After sub-pixel edge extraction,contour analysis,contour sorting,the contour is fitted using a weighted least squares method based on Tukey weight function,and then the alignment reference line is determined.By limiting the direction of the edge and the direction of contour fitting,a robust reference line recognition algorithm is proposed.Finally,the detection process of the strip alignment algorithm is designed based on the found reference lines.(4)The image classification algorithm is studied for the requirements of red marker recognition.Furthermore,the advantages and disadvantages of traditional image classification algorithms and image classification algorithms based on convolutional neural networks are compared.In the neural network structure design,this paper replaces the full connection layer with the global average pooling layer,which reduces the model parameters.Compared with the traditional image classification algorithm,the classification algorithm based on convolutional neural network does not need to manually set the threshold,and the classification accuracy rate is up to 100%,which improves the detection accuracy of the classification algorithm.(5)The human-computer interaction interface of the software is designed.It includes 6 camera real-time test result display interface,camera parameter setting interface,visual algorithm parameter setting interface,etc.Finally,the software is implemented in C++ code in the Windows development environment.It has been successfully used on the automatic cutting and packaging equipment of blood glucose test strips,which greatly improves the production efficiency of blood glucose test strips.