Author Login Editor-in-Chief Peer Review Editor Work Office Work

Computer Engineering

Previous Articles     Next Articles

Image Recognition Method Based on Self-labeling Online Sequential Extreme Learning Machine

WEI Tao 1,2,JI Xinsheng 1   

  1. (1.National Digital Switching System Engineering and Technological R & D Center,Zhengzhou 450002,China; 2.School of Computer Science,Henan University of Engineering,Zhengzhou 451191,China)
  • Received:2015-04-08 Online:2016-06-15 Published:2016-06-15

基于自标注在线顺序极速学习机的图像识别方法

魏涛1,2,季新生1   

  1. (1.国家数字交换系统工程技术研究中心,郑州 450002; 2.河南工程学院 计算机学院,郑州 451191)
  • 作者简介:魏涛(1981-),男,讲师、博士研究生,主研方向为模式识别、图像融合;季新生,教授、博士、博士生导师。
  • 基金资助:
    河南省基础与前沿技术研究计划基金资助项目“基于多源光学图像的信息融合关键技术研究”(142300410374)。

Abstract: In image recognition field,the fact is that the labeled data is far less than the unlabeled data.In order to make full use of unlabeled data to improve the ability of image recognition,a Self-labeling Online Sequential Extreme Learning Machine(SLOSELM) algorithm is proposed.Based on the labeled data in the source domain,an Extreme Learning Machine(ELM) model is trained to recognize the unlabeled data in the target domain.The high confidence samples in the recognition results are selected to adaptivly adjust the trained ELM model using the SLOSELM algorithm,thus can enhance the recognition accuracy.Experimental results on real data sets show that the average recognition ability of the ELM model is improved by about 18% after using SLOSELM algorithm,and the recognition time is shorter than that of the Co-training algorithm.

Key words: machine learning, Extreme Learning Machine(ELM), transfer learning, self-labeling algorithm, image recognition

摘要: 针对图像识别领域目标域标注数据较少而未标注数据较多的情形,为能充分利用未标注数据以提高模型识别能力,提出一种自标注在线顺序极速学习机(SLOSELM)算法。基于源域中已标注数据构建极速学习机(ELM)模型以识别目标域中未标注数据,选取识别结果中置信度高的样本,并采用SLOSELM算法对ELM模型进行自适应调整,提高图像识别能力。在真实数据集上的实验结果表明,应用SLOSELM算法后ELM模型的图像平均识别能力提高约18%,相比Co-training算法识别时间更短。

关键词: 机器学习, 极速学习机, 迁移学习, 自标注算法, 图像识别

CLC Number: