计算机工程与应用 ›› 2018, Vol. 54 ›› Issue (20): 208-212.DOI: 10.3778/j.issn.1002-8331.1707-0031

• 图形图像处理 • 上一篇    下一篇

提取目标区域词袋特征的图像分类方法

王娜娜1,李晓旭1,2,曹  洁1   

  1. 1.兰州理工大学 计算机与通信学院,兰州 730050
    2.北京邮电大学 信息与通信工程学院,北京 100876
  • 出版日期:2018-10-15 发布日期:2018-10-19

Image classification method based on extracting bag of words features of object area

WANG Nana1, LI Xiaoxu1,2, CAO Jie1   

  1. 1.School of Computer and Communication, Lanzhou University of Technology, Lanzhou 730050, China 
    2.School of Information and Communication Engineering, Beijing University of Posts and Telecommunications, Beijing 100876, China
  • Online:2018-10-15 Published:2018-10-19

摘要: 规则网格是视觉词袋模型中常用的图像检测方法,该方法抽取图像所有区块,获得背景区块和目标区块完整的图像信息。事实上,抽取的背景区块信息对类别的判定往往会有一定的混淆作用。以“摩托车”类和“小汽车”类的图像为例,这两类图像背景特征相似,大多都是道路,一般的分类方法很可能将它们分为相同类别。可见,背景信息会干扰图像分类结果。因此,提出一种提取目标区域词袋特征的图像分类方法。利用图像分割去除背景信息提取目标区域;对目标区域构建视觉词袋模型;使用SVM分类器对图像进行分类。PASCAL VOC2006及PASCAL VOC2010数据集上的实验结果表明,提取目标区域词袋特征的图像分类方法具有较好的分类性能。

关键词: 视觉词袋特征, 图像分割, 图像分类

Abstract: The regular grid is a popular image detection method in visual bag of words model. This method extracts all image blocks, and obtains complete image information of background blocks and object blocks. In fact, the extracted background block information could confuse classification result. Taking “motorbike” and “cars” images as examples, their background features are similar, and mostly are roads. Some classical classification methods possible divide them into the same category. It can be seen that background information will affect image classification results. So, this paper proposes an image classification method based on extracting bag of words features of object region. Firstly, the background information is removed by image segmentation, and the object area is extracted. Then, bag of words features are extracted for object area. Finally, SVM is used to classify images. The method proposed in this paper is evaluated in PASCAL VOC2006 and PASCAL VOC2010 datasets, and experimental results show that the proposed method has better performance.

Key words: visual bag of words feature, image segmentation, image classification