基于Gist和PHOG特征的场景分类  

Scene Classification Based on Gist and PHOG Feature

作　　者：刘静[1] 郭建[1] 贺遵亮 

机构地区：[1]湘潭大学材料与光电物理学院,湖南湘潭411105

出　　处：《计算机工程》2015年第4期232-235,240,共5页Computer Engineering

摘　　要：局部Gist方法提取的特征维数过高、计算复杂,单一的Gist特征不能很好地描述全局场景。为此,提出一种将改进的局部Gist特征与梯度方向直方图特征进行组合的场景描述方法。采用支持向量机作为分类器,在WS场景库中考察单一特征和组合特征的分类精度,在OT场景库下研究不同数量训练样本对于分类精度的影响。实验结果表明,与全局Gist、局部Gist等方法相比,该方法能降低计算的复杂度,且提高分类正确率。In view of complex computation caused by extracting high dimension characteristics with local Gist method,as w ell as the problem that the sole Gist characteristic can not describe global scenes w ell,a kind of improved method to describe the scenes is proposed,w hich combines local Gist characteristics w ith Histograms of Oriented Gradient（ HOG）characteristics. Classification accuracy of the sole characteristics and the combination of characteristics are inspected in the WS scene database using Support Vector M achine（ SVM） as the classifier. On this basis,classified precision influenced by different quantity training samples is also studied in the OT scenes database. Experimental results show that this method reduces the computational complexity,and improves the classified accuracy compared w ith the global Gist,local Gist methods,etc.

关 键 词：局部Gist特征 梯度方向直方图 特征组合 场景描述 支持向量机 场景分类 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]