作者：沈然鑫1, 朱培逸1,2
作者单位：1. 盐城工学院电气工程学院, 江苏 盐城 224002;
2. 常熟理工学院电气与自动化工程学院, 江苏 苏州 215500
关键词：全自动血型检测系统;微柱凝胶法;特征提取;alexnet模型;图像分类
摘要：
针对微柱凝胶卡血型分析仪判读准确率较低的检测问题，设计一种基于机器视觉的全自动血型检测系统。该系统通过相关的特征提取算法和图像识别算法来提高系统的判读准确性。在特征提取算法中，采用颜色通道分离合并和熵处理结合的图像处理算法，提取血型图像中的颜色特征和灰度特征，从而提高图像分析的能力。在图像识别算法中，以alexnet网络模型为基础，设计一种改进的分类神经网络模型进行训练预测，在网络结构中加入一个通道注意力机制，赋予权重，扩大特征对结果的影响，在训练参数中采用adam优化器，并优化多分类交叉熵损失函数，调整学习率固定步长衰减策略的参数，有利于加快模型收敛速度，提高模型的准确率。之后设计判读界面进行系统调试。实验结果表明：使用所设计的全自动血型检测系统，绝大多数的血型图像都可准确检测出结果，相较于原来的93.789%，准确率提高，可达到97.516 %；并且结合所做的血型检测试验，可以有效地判读出凝胶卡的最终试验结果。所采用的血型判读技术对微柱凝胶卡检测是行之有效的，可大大提高检测系统的判读准确性。
design of fully automated blood group detection system based on microcolumn gel cards
shen ranxin, zhu peiyi
1. school of electrical engineering, yancheng institute of technology, yancheng 224002, china;
2. school of electrical and automation engineering, changshu institute of technology, suzhou 215500, china
abstract: to address the detection problem of low accuracy of micro-column gel card blood group analyzer interpretation, this paper designs a fully automated blood group detection system based on machine vision. the system is designed to improve the reading accuracy of the system through relevant feature extraction algorithms and image recognition algorithms. in the feature extraction algorithm, this paper adopts an image processing algorithm combining color channel separation and merging and entropy processing to extract color features and grayscale features in blood type images, so as to improve the image analysis. in the image recognition algorithm, this paper designs an improved classification neural network model based on alexnet network model for training and prediction. in which, a channel attention mechanism is added to the network structure to assign weights to expand the influence of features on the results. in the training parameters, adam optimizer, multi-classification cross-entropy loss function optimization and learning rate fixed-step decay strategy for tuning the parameters are used, which helps to speed up the convergence of the model and improve the accuracy of the model. afterwards, this paper designs the interpretation interface for system debugging. the experimental results show that using the designed fully automated blood group detection system, most of the blood group images can be accurately detected with an increased accuracy of up to 97.516 % compared to the original 93.789%. and the final test results of the gel cards can be effectively interpreted in conjunction with the blood group testing tests performed. the blood group interpretation technique used in this paper is effective for microcolumn gel card testing and can greatly improve the interpretation accuracy of the detection system.
keywords: fully automated blood group detection system;microcolumn gel method;feature extraction;alexnet model;image classification
2023, 49(11):23-29,44 收稿日期: 2023-01-05;收到修改稿日期: 2023-02-28
基金项目: 国家自然科学基金（61903050）
作者简介: 沈然鑫(1998-),男,江苏常州市人,硕士研究生,专业方向为图像处理。
参考文献
[1] 段傲, 李莉, 杨旭. 基于alexnet的图像识别与分类算法[j]. 天津职业技术师范大学学报, 2022, 32(1): 63-66.
[2] 李露. 基于支持向量机的微柱血型卡图像识别研究与应用[d]. 天津: 河北工业大学, 2016.
[3] song w q, huang p, wang j, et al. red blood cell classification based on attention residual feature pyramid network[j]. frontiers in medicine, 2021, 8: 1-12.
[4] aristoy a a, rozenbaum y a, evtushenko g s. an automated method for blood type determination by red blood cell agglutination assay[j]. biomedical engineering, 2022, 55: 328-332.
[5] parab m a, mehendale n d. red blood cell classification using image processing and cnn[j]. sn computer science, 2021, 2: 1-10.
[6] 凝胶卡法基础操作手册[g]. 江苏贝索生物工程有限公司, 2021.
[7] 姜伟平, 孟子晗, 嵇志康, 等. 基于视觉伺服的机器人辅助医用显示器可视角检测系统设计[j]. 中国测试, 2022, 48(5): 128-133,141.
[8] a arndt p, garratty g. a critical review of published methods for analysis of red cell antigen-antibody reactions by flow cytometry, and approaches for resolving problems with red cell agglutination[j]. transfusion medicine reviews, 2010, 24(3): 172-194.
[9] 段宇秀, 杜文华. 影像测量仪自动定位方法研究[j]. 电子测量技术, 2019, 42(10): 95-98.
[10] 罗刚银, 王弼陡, 孙海旋, 等. 基于dsp的血型图像判读系统[j]. 液晶与显示, 2015, 30(2): 283-289.
[11] krizhevsky a, sutskever i, e hintion g. imagenet classification with deep convolutional neural networks[j]. communications of the acm, 2017, 60(6): 84-90.
[12] 刘怀宾. 基于微流血型检测卡的全自动血型分析系统研究[d]. 北京: 中国科学院大学, 2021.
[13] 刘玉红, 陈满银, 刘晓燕. 基于通道注意力的多尺度全卷积压缩感知重构[j]. 计算机工程, 2022, 48(12): 189-195.
[14] sivaranjini s, sujatha c m. deep learning based diagnosis of parkinson’s disease using convolutional neural network[j]. multimedia tools and applications, 2020, 79: 15467-15479.
[15] li l, doroslovacki m, h loew m. approximating the gradient of cross-entropy loss function[j]. ieee access, 2020, 8: 111626-111635.
[16] dawud a m, yurtkan k, oztoprak h. application of deep learning in neuroradiology: brain haemorrhage classification using transfer learning[j]. computational intelligence and neuroscience, 2019, 2019(6): 1-12.
[17] sun g, cholakkal h, khan s, et al. fine-grained recognition: accounting for subtle differences between similar classes[c]//proceedings of the aaai conference on artificial intelligence, 2019.
[18] gyoung s n a. efficient learning rate adaptation based on hierarchical optimization approach[j]. neural networks, 2022, 150(2): 326-335.
[19] zhang j l, khavatnezhad m, ghadimi n. optimal model evaluation of the proton-exchange membrane fuel cells based on deep learning and modified african vulture optimization algorithm[j]. energy sources, part a: recovery, utilization, and environmental effects, 2022, 44(1): 287-305.
[20] toptas b, hanbay d. retinal blood vessed segmentation using pixel-based feature vector[j]. biomedical signal processing and control, 2021, 70: 1-12.
[21] iqbal m a, wang z, ali z a, et al. automatic fish species classification using deep convolutional neural networks[j]. wireless personal communications, 2021, 116(1): 1043-1053.
