螨虫图像特征可视化卷积神经网络研究
Study on Visual Convolutional Neural Network of Image Features of Mites
田欢 1王鑫2
作者信息
- 1. 兰州职业技术学院现代服务学院,甘肃 兰州 730070
- 2. 兰州职业技术学院电子信息工程学院,甘肃 兰州 730070
- 折叠
摘要
螨虫是古老的微型寄生害虫,它与人们的健康息息相关.文章借助基于数据驱动的深度学习技术对大规模螨虫影像进行特征学习分析,通过引入人工建模、AlexNet和VGG16经典的卷积神经网络等算法,并结合特征图可视化阐释了卷积的深层工作机理,对于螨虫相关的生物学领域及先进计算机视觉方法都具有一定的研究价值.经实验证明,文章采用数据特征提取结构不仅适用于螨虫图像的可视化研究,而且可以将此方法推广至其他研究领域,具有一定的鲁棒性,同时也为其他学科的可视化实验、研究提供了普适的方法.
Abstract
Mites are ancient micro-parasitic pests,which are closely related to people′s health.Using data-driven deep learning technology to conduct feature learning analysis on large-scale mite images will greatly pro-mote the research progress of related disciplines.This paper introduced artificial modeling,classical convolutional neural network algorithms such as AlexNet and VGG16,and combined with feature map visualization to explain the deep working mechanism of convolutions,which has certain research value for the biological fields related to mites and advanced computer vision methods.The experiment proves that the data feature extraction structure adopted in this paper is not only suitable for the visualization research of mite images,but also can be extended to other fields with certain robustness,in order to provide a universal method for visualization experiments and research of other disciplines.
关键词
螨虫/卷积神经网络/VGG16/特征可视化Key words
mites/convolutional neural networks/VGG16/feature visualization引用本文复制引用
出版年
2024
NETL
NSTL
万方数据