| 物体分类识别中视觉词汇生成方法研究 |
| |
| 引用本文： | 王红霞,杨克俭,高峰,朱荣.物体分类识别中视觉词汇生成方法研究[J].武汉理工大学学报(信息与管理工程版),2011,33(6):945-947,957. |
| |
| 作者姓名： | 王红霞 杨克俭 高峰 朱荣 |
| |
| 作者单位： | 1. 武汉理工大学计算机科学与技术学院,湖北武汉,430063
2. 武汉大学计算机学院,湖北武汉,430072 |
| |
| 基金项目： | 武汉市科学技术局科技攻关计划基金资助项目 |
| |
| 摘 要： | 针对目前Bag of words模型将聚类中心作为视觉单词,而导致语义信息表达不完全的问题,提出了一种新的改进的视觉词汇生成方法。首先,提取图像的SIFT特征点并聚类;然后利用核函数进行核密度估计,选取每个聚类中若干个有代表性的特征点;最后,通过SVM训练生成视觉词汇。实验结果表明,改进后的视觉词汇生成方法,在物体分类识别中,与以聚类中心为视觉单词的生成方法相比,增强了语义信息的表达,提高了查全率,使得物体分类识别率大大增加。
|
| 关 键 词： | Bag of words模型 视觉词汇 核函数 支持向量机 |
Visual Vocabulary Generation Method for Object Classification |
| |
| WANG Hongxia,YANG Kejian,GAO Feng,ZHU Rong.Visual Vocabulary Generation Method for Object Classification[J].Journal of Wuhan University of Technology(Information & Management Engineering),2011,33(6):945-947,957. |
| |
| Authors: | WANG Hongxia YANG Kejian GAO Feng ZHU Rong |
| |
| Affiliation: | WANG Hongxia,YANG Kejian,GAO Feng,ZHU Rong Doctorial Candidate,School of Computer Science and Technology,WUT,Wuhan 430063,China. |
| |
| Abstract: | Aiming at the problem of incomplete expressions of image semantic information caused by the visual word which are taken by cluster center in the bag of words model,a novel visual vocabulary generation method was proposed.Firstly,after they were extracted,SIFT feature points of images were clustered.Then,some representative feature points were selected from each cluster by soft-decision based on kernel function.Finally,SVM was employed to train these selected feature points to generate a visual word of each ... |
| |
| Keywords: | bag of words model visual vocabulary kernel function SVM |
| 本文献已被 CNKI 万方数据 等数据库收录！ |
|
