HSV与LBP特征融合的行人检测方法研究  

Research of pedestrian detection method based on HSV and LBP feature fusion

作　　者：岳求生 周书仁[1] 李峰[1] 谭飞刚 

机构地区：[1]长沙理工大学计算机与通信工程学院,湖南长沙410004

出　　处：《计算机工程与科学》2014年第10期1997-2001,共5页Computer Engineering & Science

基　　金：长沙市科技计划资助项目(K1203015-11);湖南省标准化战略项目(2011031)

摘　　要：提出一种融合HSV颜色空间特征与局部二元模式特征LBP的特征的HSV-LBP行人检测方法。HSV特征是一种全局特征,它能简单地描述一幅图像中颜色的全局分布,LBP特征能很好地描述图像局部空间结构,所以该算法既考虑了全局特征也考虑了局部特征,且该算法具有维数少、计算速度快的优点。在Matlab环境下实验,利用Adaboost分类器对算法的性能进行实验仿真,与经典的梯度方向直方图HOG特征、LBP特征、分层梯度方向直方图PHOG特征及HOG-LBP特征进行对比,结果表明HSV-LBP方法的识别性能较好。A pedestrian detection method is proposed based on the fusion of HSV color space feature and LBP feature. HSV feature is global which can simply describe the global distribution of colors in an image while LBP feature is a good description of the spatial structure of a local image,therefore the pro- posed method takes both the global features and the local features into account, which has the advanta- ges of low dimension and quick computing speed. In Matlab,Adaboost classifier is used to carry out the experiments. Results show that, compared with the classic Gradient Direction Histogram （HOG） fea- ture,LBP feature,Pyramid Histogram of Oriented Gradients （PHOG） feature and HOG-LBP feature, the proposed method is of high recognition rate.

关 键 词：颜色空间 LBP HSV 特征融合 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]