多特征融合和交叉核SVM的车辆检测方法  

Vehicle detection method based on multi-features fusion and intersection kernel SVM

作　　者：胡庆新[1] 焦伟[1] 顾爱华[1] 

机构地区：[1]合肥工业大学计算机与信息学院,安徽合肥230009

出　　处：《合肥工业大学学报（自然科学版）》2016年第1期84-89,共6页Journal of Hefei University of Technology：Natural Science

基　　金：高等学校博士学科点专项科研基金资助项目(20120111110001)

摘　　要：为了能够提高车辆检测的准确率和效率,利用不同分辨率下图像中的信息,文章提出一种多分辨率下的方向梯度直方图特征(multi-resolution histograms of oriented gradients)和局部二值模式算子(local binary pattern)相融合的车辆检测算法。由于融合后的特征维数明显高于单一特征,使得检测的实时性较差,通过主元分析(principal component analysis)来实现对融合后特征降维,并采用直方图交叉核支持向量机(HIK-SVM)对特征进行分类,完成车辆的检测。实验结果表明,该检测方法有较高的车辆检测率,且实时性较好。In this paper, a vehicle detection method which combines multi-resolution histograms of oriented gradients（HOG） features and local binary pattern（LBP） features as mixed features is proposed by using the image information under different resolutions to improve the accuracy and efficiency of detection. Because the dimension of the mixed features is greater than any single feature and causes the poor real-time performance, the principal component analysis（PCA） is used to decrease the dimension of the mixed features, and the histogram intersection kernel support vector machine（HIK-SVM） is used for the features classification. The experimental results show that the proposed method can detect vehicles with higher detection rate and less calculating time.

关 键 词：车辆检测 多特征融合 方向梯度直方图 局部二值模式算子 主元分析 直方图交叉核SVM 

分 类 号：TP391.413[自动化与计算机技术—计算机应用技术]