摘要
针对肺炎影像检测问题,提出一种基于SSD的肺炎影像检测模型,解决了小发病区域检测准确度较低的问题。首先,将原始SSD网络模型的第3网络层的输出结果通过池化降维等操作与网络的第6卷积层的输入特征进行融合;然后,将网络模型的第5池化层的输出结果通过反池化、反卷积等操作后与SSD网络模型的第7特征进行特征融合。为了检验基于SSD的肺炎影像检测模型对肺炎的检测识别性能,针对常见检测模型进行了对比实验,结果显示基于SSD的肺炎影像检测模型对肺炎疾病的检测准确率为91.24%,优于其他对比实验的方法。
To address the issue of pneumonia detection from medical image,this paper proposes a pneumonia image detection model based on SSD,which solves the problem of low detection accuracy angle in small disease area. Firstly,the output results of the third network layer of the original SSD network model are fused with the input features of the sixth feature layer of the network through pooling dimensionality reduction,and then the output results of the fifth pooling layer of the network model are fused with the seventh feature layer of the SSD network model through anti pooling and deconvolution. In order to test the detection and recognition performance of pneumonia image detection network model proposed in the paper,comparative experiments were carried out based on common detection models. The accuracy of pneumonia image detection model proposed is 91.24%,which is better than other comparative experimental methods.
作者
马驰
胡辉
路生亮
布安旭
金海滨
MA Chi;HU Hui;LU Shengliang;BU Anxu;JIN Haibin(School of Computer Science and Engineering,Huizhou University,Huizhou 516007,Guangdong,China;School of Computer and Software Engineering,Liaoning University of Science and Technology,Anshan 114051,Liaoning,China)
出处
《惠州学院学报》
2022年第3期75-80,共6页
Journal of Huizhou University
基金
广东省教育厅重点建设学科科研能力提升项目(2021ZDJS082)
福建省自然科学基金项目(2019J01751)。
