摘要
提出一种基于金字塔梯度方向直方图的分层投票方法,将其用于人脸识别、对象识别和对象检测等领域。利用分层投票机制选出分类效果最好的层,并为它们赋予相应的投票权值,最终对图像类别进行投票。在ORL和UIUC Car数据库上进行实验,结果表明,改进方法优于传统的特征向量分类方法。
On the basis of Pyramid Histogram of Oriented Gradient（PHOG）, a new layered vote method is presented. It is used in the area of face recognition, object recognition, and object detection. By using the layered vote, it can concentrate the attention on the levels which have high recognition rate, and the weights of vote are set to these chosen levels. The final result is got by vote for the object category. This method is tested on ORL and UIUC Car database, and the results show that it is better than traditional classification method.
出处
《计算机工程》
CAS
CSCD
2012年第8期156-158,共3页
Computer Engineering
基金
国家自然科学基金资助重点项目(61033012)
关键词
梯度直方图
空间金字塔
分层投票
图像块
支持向量机
Histogram of Gradient（HOG）
spaeial pyramid
layered vote
image block
Support Vector Machine（SVM）
