摘要
视觉图像检测在机器视觉领域有着重要的研究意义和应用价值。近年来,卷积神经网络的发展带动了视觉图像检测领域的进步。大量新理论、新方法被应用于卷积神经网络,提高了网络对特征的表达能力,降低了网络的复杂性,增强了网络的性能。研究阐述了卷积神经网络的基本构成,从卷积层,池化层,激活函数,网络正则化和网络优化等方面总结了卷积神经网络近年来的改进方法,梳理了卷积神经网络在视觉图像检测领域的应用,总结了卷积神经网络在视觉图像检测领域的优点,并展望了未来的研究方向。
Visual image detection has great research significance and application value in the computer vision field. In recent years, the development of convolutional neural network(CNN) has led to the progress of visual image detection. A large number of new theories and new methods are applied to convolutional neural network, which improves the network feature expression ability, reduces the network complexity and improves the network performance. This paper presents the basic structure of Convolutional CNN, summarizes the improvements of CNN in recent years on different aspects, including convolutional layer, pooling layer, activation function, network regularization and network optimization, sorts various applications of CNN in visual image detection field and summarizes the advantages of CNN in visual image detection field, finally, prospects the future research direction.
作者
蓝金辉
王迪
申小盼
Lan Jinhui;Wang Di;Shen Xiaopan(School of Automation and Electrical Engineering,University of Science and Technology Beijing,Beijing 100083,China)
出处
《仪器仪表学报》
EI
CAS
CSCD
北大核心
2020年第4期167-182,共16页
Chinese Journal of Scientific Instrument
基金
装备预先研究项目(61403110404)资助。
关键词
卷积神经网络
深度学习
机器视觉
图像检测
convolutional neural network(CNN)
deep learning
computer vision
image detection
