作者投稿和查稿 主编审稿 专家审稿 编委审稿 远程编辑

计算机工程 ›› 2010, Vol. 36 ›› Issue (20): 170-172. doi: 10.3969/j.issn.1000-3428.2010.20.060

• 人工智能及识别技术 • 上一篇    下一篇

基于遗传算法的高分辨率遥感分类器融合

马永聪,方 涛   

  1. (上海交通大学图像处理与模式识别研究所遥感科学研究室,上海 200240)
  • 出版日期:2010-10-20 发布日期:2010-10-18
  • 作者简介:马永聪(1985-),男,硕士研究生,主研方向:高分辨率遥感图像特征选择,多分类器融合;方 涛,教授、博士生导师
  • 基金资助:
    国家“863”计划基金资助项目(2006AA12Z105);国家“973”计划基金资助项目(2006CB701303)

Multiple Classifier Fusion in High-resolution Remote Sensing Based on Genetic Algorithms

MA Yong-cong, FANG Tao   

  1. (Remote Sensing Group, Institute of Image Processing and Pattern Recognition, Shanghai Jiaotong University, Shanghai 200240, China)
  • Online:2010-10-20 Published:2010-10-18

摘要: 提出利用遗传算法设计的2种多分类器融合模型:在模型1中,各单分类器选择不相交的特征子空间;模型2则取消了上述限制。通过实验,对2种模型与2种单分类器进行了对比。结果表明,2种多分类器模型能有效地提高分类精度,且模型1利用了不相交特征空间,更有利于提高特征相关性较大的高分辨率遥感图像的分类精度。

关键词: 遗传算法, 多分类器融合, 不相交特征子空间, 高分辨率遥感

Abstract: Two models of classifier-fusion systems designed by genetic algorithms are proposed. 1st model utilizes disjoint feature subspaces, whereas 2nd one withdraws the restriction. Through the experiments, the two models are compared against single classifiers. Results show that, the two classifier-fusion systems yields better classification accuracy against the two single classifiers, and 1st model, which utilizes disjoint feature subspaces, can help improve the classification of high-resolution remote sensing with significant feature correlation.

Key words: genetic algorithm, multiple classifier fusion, disjoint feature subspaces, high-resolution remote sensing

中图分类号: