摘要
针对国内对显微镜下致病菌种识别以及细胞活性检验等问题,提出以darknet-53网络模型与Yolov3算法取出疑似目标,再配以传统形态学算法进行筛选,从而达到准确、快速、智能化的诊断方式。Yolo算法检测速度较传统检测算法提高上百倍,可满足医学诊断的高效需求。再加入形态学算法,对目标做形状椭圆拟合,对色值、轮廓和大小等维度做判断,从而进一步提升检测正确率与检出率。经过自建样本数据训练以及大量实验表明,加入传统形态学处理后,对孢子的识别算法准确率高达94%以上,检出率达82%以上。本文算法将传统图像处理方法与深度学习方法相结合,应用于实际检测中,提高了检测准确率,检测速度是专业医务人员检测速度的两倍以上。
Aiming at domestic problems such as identification of pathogenic bacteria under microscope and cell viability test,it is proposed to use darknet-53 network model and Yolov3 algorithm to extract suspected targets,and then use traditional morphological algorithms for screening to achieve accuracy,speed and intelligence diagnostic methods.The detection speed of the Yolo algorithm is hundreds of times higher than that of traditional detection algorithms to meet the high efficiency of medical diagnosis.then add the morphological algorithm to shape the target and judge the color value,contour,size and other dimensions to further improve the detection accuracy and detection rate.After self-built sample data training and a large number of experiments,the accuracy rate of the spore recognition algorithm after adding traditional morphological processing is as high as 94%and the detection rate is more than 82%.The algorithm in this paper combines traditional image processing methods and deep learning methods.When applied to actual detection,it improves the detection accuracy and is more than twice the detection speed of professional medical personnel.
作者
李鑫铭
赵磊
邵宝民
王栋
LI Xinming;ZHAO Lei;SHAO Baomin;WANG Dong(School of Computer,Shandong University of Technology,Zibo Shangdong 255000,China)
出处
《智能计算机与应用》
2020年第5期30-34,38,共6页
Intelligent Computer and Applications
