Research On Defect Detection Method Of Blood Lancet Based On Deep Learning

Disposable peripheral blood lancet(short blood lancet)is a necessary means for self testing and medical diagnosis of diabetic patients.At present,the quality inspection of blood lancet mainly depends on manual detection.In order to observe the obvious defect characteristics,the detection of blood lancet needs to be carried out under strong light.Repeated detection for a long time will cause visual fatigue and affect the stability of detection.In order to realize the automation of blood lancet detection and improve the speed and accuracy of blood lancet detection,this paper studies the defect detection algorithm of blood lancet based on deep learning.The main work is as follows:(1)The lancet objects studied in this subject have the characteristics of a large number of targets and a small proportion of defective products.In order to solve the problem of blood lancet data set,a single blood lancet data set and blood lancet for classification network defect detection were produced respectively.A multi-target blood lancet data set with a balanced number of targets.(2)Research on defect detection method of blood sampling needle based on classification network.Four convolutional neural network models vgg16,resnet50,densenet121 and mobilenetv2 are selected for the research of blood collection needle defect detection.The model is trained and tested by using a single blood collection needle data set.In the test results,the detection accuracy of the four models is more than 99.50%,but the average detection time is more than 1s,which can not meet the real-time requirements of production detection.(3)Research on defect detection method of blood sampling needle based on improved yolov3 algorithm.CBAM attention module is added to the backbone network of yolov3,and the activation function is optimized.The multi-objective blood collection needle data set is used for model training,and the optimal model is obtained by parameter adjustment.On the test set,the map value of the improved yolov3 algorithm reaches 98.28% and the detection time is 67 ms.(4)Design and implement the defect detection system of blood lancet.The image acquisition module of blood lancet and the transmission movement device of blood lancet are built,the display interface design is completed,and the experimental test of the system is carried out.The test results show that the detection accuracy of the system is 99.15%,which can realize the defect detection of blood lancet after injection molding,the detection time is 0.5s,and has good reliability and practical value.This paper studies the application of deep learning technology to lancet defect detection,and initially designs the detection system of blood lancet defect,which realizes the automatic detection of blood lancet after injection.The system has good effect.