郭原东;雷帮军;聂豪;李讷;

• 1:三峡大学

摘要(Abstract)：

针对高精度训练样本缺失场景下图像识别算法泛化能力差的问题,文中提出一种改进的深度置信网络结构(DBNs)。该结构通过在DBN网络中引入随机隐退机制,使得隐含层中的部分单元失效,仅保留其连接权重,防止小样本量训练过程中产生的过拟合现象。为降低引入随机隐退机制后算法的计算复杂度,在该结构中引入基于相邻近算法的降采样机制。采用ORL开放人脸数据集进行仿真实验,结果表明,该机制可以将识别错误率由普通DBN网络的43%降低到5.0%,但计算时间有所增加。对比引入降采样算法后的网络测试结果显示,网络训练时间下降约69.9%;与AlexNet等公开网络的对比测试结果表明,该算法的识别精度可达95.2%,在计算精度与识别效率上均有一定的优越性。

关键词(KeyWords)： 图像识别;深度学习;随机隐退;图像降采样;仿真实验;网络测试

Abstract:

Keywords:

基金项目(Foundation): 湖北省科技厅项目(2019ZYYD007)

作者(Author): 郭原东;雷帮军;聂豪;李讷;

Email:

DOI: 10.16652/j.issn.1004-373x.2021.04.038

参考文献(References)：

• [1]DUAN M,LI K,YANG C,et al.A hybrid deep learning CNN-ELM for age and gender classification[J].Neurocomputing,2018,75(2):448-461.
• [2]LU N,WU Y D,FENG L,et al.Deep learning for fall detection:3D-CNN combined with LSTM on video kinematic data[J].IEEE journal of biomedical&health informatics,2018(7):13-21.
• [3]WANG Ping,JIANG Lei.Deep learning-based object classification through multimode fiber via a CNN-architecture SpeckleNet[J].Applied optics,2018(3):233-242.
• [4]ZHANG M D,LIAO G S,HE X P,et al.Unambiguous forward-looking SAR imaging on HSV-R using frequency diverse array[J].Sensors,2020,20(4):1169-1183.
• [5]LI D,LI R,ZHANG S.A deep learning image recognition framework accelerator based parallel computing[C]//The 2nd International Conference.Beijing:ACM,2018:101-110.
• [6]HUANG H W,LI Q T,ZHANG D M.Deep learning based image recognition for crack and leakage defects of metro shield tunnel[J].Tunnelling and underground space technology,2018,77(7):166-176.
• [7]HUANG Hongwei,LI Peng.Deep learning based image recognition for crack and leakage defects of metro shield tunnel[J].Tunnelling&underground space technology,2018(1):31-40.
• [8]YANG X,ZENG Z,TEO S G,et al.Deep learning for practical image recognition:case study on kaggle competitions[C]//The 24th ACM SIGKDD International Conference.Hangzhou:ACM,2018:42-55.
• [9]AIZEZI Y,JIAMALI A,ABUDUREXITI R,et al.Research on image recognition method based on deep learning algorithm[C]//IEEE International Conference on Advanced Video&Signal Based Surveillance.Boston:IEEE,2018:312-320.
• [10]MEZGEC S,EFTIMOV T,BUCHER T,et al.Mixed deep learning and natural language processing method for fake-food image recognition and standardization to help automated dietary assessment[J].Public health nutrition,2018(6):1-10.
• [11]POSTALCIOGLU S.Performance analysis of different optimizers for deep learning based image recognition[J].International journal of pattern recognition&artificial intelligence,2019(8):1120-1133.
• [12]HUANG L Q,LI J,HAO H,et al.Micro-seismic event detection and location in underground mines by using Convolutional Neural Networks(CNN)and deep learning[J].Tunnelling and underground space technology,2018,81(11):265-276.
• [13]蔡昌许.基于布隆过滤器的WSN链路层地址隐藏方法[J].计算机工程与设计，2016,37(5):1208-1211.