共生矩阵保局投影红外人体目标实时跟踪方法  

Co-occurrence matrix locality preserving projections for pedestrian tracking in infrared image sequences

作　　者：李建福[1,2] 龚卫国[1] 杨金妃[1] 

机构地区：[1]重庆大学光电技术及系统教育部重点实验室,重庆400044 [2]重庆教育学院计算机科学系,重庆400067

出　　处：《红外与激光工程》2010年第6期1012-1017,共6页Infrared and Laser Engineering

基　　金：国家"十一五"基础研究项目(C10020060355);国家"863"计划项目(2007AA1E243);重庆市科技攻关项目(CSTC2009AB0175);重庆市自然科学基金项目(CSTC2008BB2199)

摘　　要：针对红外图像中人体的鲁棒实时跟踪问题,提出了一种新颖的共生矩阵保局投影(COMLPP)方法。为了克服保局投影的红外图像中人体目标信息量不足的弱点,该方法首先构建了训练样本的共生矩阵,然后利用保局投影得到样本目标的共生矩阵保局投影子空间特征的表征向量,最后将上述表征模型与粒子滤波相融合,实现了粒子滤波框架下的人体跟踪算法。利用不同的红外图像数据库进行人体跟踪实验,结果表明:文中提出的方法能够实时有效地跟踪人体目标,并且对有部分遮挡的复杂场景有较强的适应性。A co-occurrence matrix locality preserving projections（COMLPP） method was proposed to improve the robust performance of real-time pedestrian tracking in infrared image sequences.To overcome the weak object feature description ability of a single locality preserving projections（LPP） for pedestrian target in infrared images,the co-occurrence matrices of the training set were first constructed,then the feature representation vector of the LPP subspaces was obtain by applying the LPP method.At last,the above mentioned pedestrian tracking representation model was embedded in a particle filter framework and sample distributions were propagated over time.Experimental results show that the proposed scheme achieves successfully in real-time pedestrian tracking of different infrared image sequences,and is more robust and stable than the classical tracking algorithm.Meanwhile,the proposed method can be adapted to complex and occluding scenes.

关 键 词：人体跟踪 红外 共生矩阵保局投影 粒子滤波 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]