基于码本模型和压缩跟踪算法的遗留物检测  

Abandoned Object Detection Based on Codebook Model and Compressive Tracking Algorithm

作　　者：张辉[1] 徐伟[1] 谢正光[1] 蒋小燕[1] 马文萱 

机构地区：[1]南通大学电子信息学院,江苏南通226019

出　　处：《电视技术》2014年第23期146-151,共6页Video Engineering

基　　金：国家自然科学基金面上项目(61171077);交通部应用基础研究项目(2011-319-813-510);南通大学创新人才基金项目(2009);南通大学研究生科技创新计划项目(YKC13003)

摘　　要：针对一般遗留物检测算法复杂度高和跟踪效果不理想的问题,提出了一种基于码本模型和压缩跟踪算法相结合的遗留物检测方法。首先通过码本模型建模和适时匹配背景更新算法来获取静止目标区域信息;然后利用稀疏测量矩阵对静止目标区域的多尺度特征进行降维,得到分类器的正负样本;最后用朴素贝叶斯分类器对提取的特征进行分类,当分类器响应最优时得到当前帧中跟踪到的目标位置,即使目标被部分遮挡,也能实现对遗留目标的准确跟踪。实验结果表明,该方法不仅简单高效、实时性好,而且可以消除由物体短暂停留而带来的干扰。The general abandoned object detection algorithms proposed are either highly complex or weak in tracking abandoned object. Therefore, an abandoned object detection method based on codebook model and compressive tracking algorithm is proposed. Firstly, the stationary target area information is obtained through codebook background modeling and timely matching background update algorithm. Then, a very sparse measurement matrix is used to reduce the dimensions of multi - scale image features, which are acted as the positive and negative samples in the classifier, for the sta- tionary target area. At last, those features extracted above are classified via a naive Bayes classifier. As a result, the tracking location is found in the current frame with the maximal classifier response. Even when partially covered, the abandoned object can still be accurately tracked. The results show that the proposed method is not only simple, efficient and real - time, but also can efficiently eliminate the interference caused by the short - term stay of the objects.

关 键 词：遗留物检测 码本模型 背景更新 压缩跟踪 实时 

分 类 号：TN911.73[电子电信—通信与信息系统]