系统仿真学报 ›› 2016, Vol. 28 ›› Issue (6): 1386-1393.

• 仿真系统与技术 • 上一篇    下一篇

基于多特征组合与优化BoW模型的影像分类技术研究

李科1, 游雄1, 杜琳1, 2   

  1. 1.信息工程大学,郑州 450052;
    2.72515部队,济南 250014
  • 收稿日期:2015-01-05 修回日期:2015-03-02 出版日期:2016-06-08 发布日期:2020-06-08
  • 作者简介:李科(1977-),男,河北沧州,博士,副教授,研究方向为地理空间数据工程;游雄(1962-),男,福建罗源,博士,教授,研究方向为虚拟地理环境;杜琳(1982-),女,河南开封,硕士,工程师,研究方向为地理信息工程。
  • 基金资助:
    国家自然科学基金(41201390),国家“863”计划(2013AA12A202)

Research of Remote Sensing Image Classification Technology Based on Multi-feature Combining and BoW Model

Li Ke1, You Xiong1, Du Lin1, 2   

  1. 1. The Information Engineering University, Zhengzhou 450052, China;
    2. Troop 72515, PLA, Jinan 250014, China
  • Received:2015-01-05 Revised:2015-03-02 Online:2016-06-08 Published:2020-06-08

摘要: 提出一种基于多特征组合与优化BoW模型的影像地物分类新方法。提取影像的SIFT、GIST、颜色、Census和Gabor等多种类型特征,通过实验分析确定最佳特征组合。针对一般K-Means算法没有考虑各个特征值的权重,提出利用自动加权k-Means算法计算不同特征分量的权值,分别对SIFT、GIST、Gabor特征构建了基于权重的影像特征词汇表,采用基于Soft的词汇编码算法进行影像编码,使用SVM算法完成影像分类。通过实验表明方法能有效提高遥感影像分类准确性,并且具有较好的稳定性和鲁棒性。

关键词: 特征组合, K-means, 视觉词袋, 支持向量机

Abstract: A new algorithm of image classification of multi feature combination and BoW model was proposed. SIFT, GIST, Census and Gabor color, and many other types of features were extracted from the images, and then through the experimental analysis to determine the best feature combination. According to the general K-means algorithm which did not consider the weight of each features, different feature component was put forward by using automatic weighted k-means algorithm, respectively SIFT, GIST, Gabor feature construct weights based on image features of vocabulary, using the soft coding algorithm for image coding, and using the SVM algorithm to complete the image classification. Experiments show that this method can effectively improve the classification accuracy of images.

Key words: multi-feature combining, k-means, bag of words, SVM

中图分类号: