基于多特征和最小二乘支持向量机相融合的行人检测模型  

Pedestrian detection model based on multi-features and least squares support vector machine

作　　者：周浩[1] 王天志[2] 

机构地区：[1]暨南大学电子工程系,广州510632 [2]云南师范大学,昆明650500

出　　处：《激光杂志》2014年第10期46-50,56,共6页Laser Journal

基　　金：广东省科技专项(NO.2009B90600036)

摘　　要：为了提高行人检测正确率,提出一种基于多特征融合和最小二乘支持向量机的行人检测模型。首先提取行人的相位一致性特征和梯度直方图特征,然后采用粒子群算法选择最优特征子集,最后将最优行人检测特征子集输入到最小二乘支持向量机对学习和分类,并采用对模型性能采用仿真实验进行测试。结果表明,相对于其它行人检测模型,本文模型不仅提高了行人检测率、降低了虚警率,而且加快行人检测效率,具有较强的鲁棒性。In order to improve the pedestrian detection rate, a new pedestrian detection model based on multi-features and least squares support vector machine is proposed in this paper. Firstly, the Phase Congruency feature and histograms of oriented gradients features are extracted, secondly, the optimal features are selected by particle swarm optimization algorithm, finally, the features are input least squares support vector machine to train and classify, and the simulation experiments are carried out to test the performance of model. The results show that compared with other pedestrian detection model, the proposed model not only improve the pedestrian detection rate, but also can fasten the speed,and it has good robust.

关 键 词：行人检测 最小二乘支持向量机 相位一致性特征 梯度直方图特征 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]