基于多通道融合HOG特征的全天候运动车辆检测方法  

A Method of Moving Vehicle Detection in All-weather Based on Melted Multi-channel HOG Feature

作　　者：刘操[1,2] 郑宏[1,2] 黎曦[2,3] 余典[1,2] 

机构地区：[1]武汉大学电子信息学院,湖北武汉430072 [2]湖北省视觉感知与智能交通技术研究中心,湖北武汉430072 [3]武汉工程大学电气信息学院,湖北武汉430025

出　　处：《武汉大学学报（信息科学版）》2015年第8期1048-1053,共6页Geomatics and Information Science of Wuhan University

基　　金：国家973计划资助项目(2012CB719905)~~

摘　　要：对传统的梯度方向直方图(HOG)特征提取方法进行改进,提出了一种基于多通道特征提取的加权HOG特征融合方法。首先采用基于亮度均值的方式对彩色车辆图像增强处理,其次采用自适应加权法将H、S、V三通道的梯度方向直方图(HOG)特征融合成多通道融合HOG特征,最后采用支持向量机(SVM)对融合后的特征进行车辆分类器训练和车辆检测。该方法与直接运用HOG特征进行车辆检测以及其他车辆检测方法相比,具有检测率高、鲁棒性强等特点,并且在各种气候环境下都能实现较好的检测效果,效果优于其他方法,达到了全天候车辆检测的要求。In this article, an algorithm for vehicle detection is proposed using the improved histogram of oriented gradients （HOG） features, in which a fusion of weighted multi-channel HOG features is employed. First, the color image is enhanced by the method based on the mean of brightness. After that the image is remapped into HSV color space. In consequence, the multi-channel HOG feature is obtained by combining the adaptive-weight HOG feature for each channel. Finally, the support vector machine （SVM） is trained by such feature to detect the vehicles. Experiments prove the robustness and precision of this algorithm and it reaches the requirement for all-weather vehicles detecting in terms of the one using the traditional HOG features.

关 键 词：多通道融合HOG 支持向量机 车辆检测 全天候 

分 类 号：P237.3[天文地球—摄影测量与遥感] TP751[天文地球—测绘科学与技术]