无特定身份背景下的目标跟踪算法仿真  

Target Tracking Algorithm Simulation under the Background without Specific Identity

作　　者：龙文光[1] 

机构地区：[1]内江师范学院现代教育技术中心,四川内江641112

出　　处：《计算机仿真》2014年第8期254-258,共5页Computer Simulation

摘　　要：研究追踪对象特征不明显,无特定身份特征的目标准确跟踪问题。无特定身份背景会导致跟踪目标出现时无明显各种特征,需要为其选定特征的跟踪特征。传统的目标跟踪算法在这种环境下采集的跟踪目标灰度特征信息单一,并且缺乏描述目标的信息,直接导致目标跟踪的失败或者带来较大的误差。提出一种改进的无特定身份背景下的目标跟踪方法。对图像进行边缘梯度融合,计算目标图像区域中基于局部二值模式(LBP)的纹理特征模型,通过Mean shift算法来建立目标直方图并实现目标的跟踪。实验结果表明,改进的算法能有效地克服无特定身份背景下的光照变化、多目标干扰的影响,改进的算法取得了较好的目标跟踪精度。The background without specific identity leads to track the target appeared no significant various char- acteristics, it is needed for tracking characteristics of the selected features. The traditional tracking algorithm in this environment takes single track target gray characteristic information, and can directly lead to the failure of the target tracking or bring bigger error. An improved target tracking method is presented under the background without specific identity, image edge gradient is calculated based on local binary pattern （LBP） texture feature model, the target his- togram is established through the Mean shift algorithm, and the goal of tracking is achieved. The experimental results show that the algorithm can effectively overcome the interferences of the multi - targets and the illumination transfor- mations under the background without a specific identity, and achieve good accuracy of target tracking.

关 键 词：目标跟踪 纹理表示方法 局部二值模式 目标直方图 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]