应用于行人检测的HLBP与CHLBP纹理特征

根据CSLBP (center-symmetric local binary pattern)和Uniform LBP (local binary pattern)特征描述行人局部纹理互补性的特点, 提出将二者级联的组合特征用于行人检测: 基于灰度图像的纹理特征(hybrid local binary pattern, HLBP)和基于颜色空间的纹理特征(color based hybrid local binary pattern, CHLBP)。实验结果表明, 当FPPW=10^–4时, HLBP特征的检测率为93.96%, 与Uniform LBP和CSLBP特征相比分别提高3.46%和9.68%, 基于颜色空间L′C′C′与HIKSVM分类器结合时的检测率高达98.58%。与传统的纹理特征检测方法相比, 该特征提高了行人检测精度, 降低了误检率, 检测性能得到较大幅度的提升。

The HLBP and CHLBP Features for Pedestrian Detection

Two improved texture features (hybrid local binary pattern, HLBP) and (color based hybrid local binary pattern, CHLBP) which are based on gray image texture and color space are proposed for pedestrian detection. The experimental results show that, when FPPW is 10^–4, the detection rate of HLBP is 93.96% which is about 3.46% and 9.68% higher than Uniform LBP and CSLBP respectively. At the same time, when combined with the HIKSVM classifier, CHLBP feature based on L′C′C′ space makes the detection rate up to 98.58%, and its detection performance has been greatly improved, by this method an good result could be obtained in pedestrian detection.