摘要： 提出一种基于组件词表的物体识别方法，通过AdaBoost从物体样本图像的组件中选取一些最具区分性的组件，构成组件词表。每幅图像都用词表中的组件来表征，在此基础上用稀疏神经网络来训练分类器。实验结果表明，该方法识别精度较高，对于遮挡和复杂背景有较强的鲁棒性。

关键词: 物体识别, 基于组件的表示, 组件词表, AdaBoost算法, 稀疏神经网络

Abstract: This paper presents a novel method for object recognition based on part vocabulary. A vocabulary of object parts is automatically constructed from sample images of the object class by AdaBoost. Images are represented by using parts from the vocabulary. Based on it, Sparse Network of Winnows (SNoW) learning architecture is employed to learn to recognize the instances of the object class. Experimental results show that the method achieves high recognition accuracy on different data sets, and it is robust to partial occlusion and background clutter.

Key words: object recognition, part-based representation, part vocabulary, AdaBoost, Sparse Network of Winnows(SNoW)

中图分类号: 

• TP391.41