基于自适应耦合模型的运动目标跟踪方法

在跟踪目标快速运动和外形显著变化的任务中,现有跟踪方法经常存在目标跟踪丢失、误差较大等问题.为此本文提出一种新的跟踪方法,该方法采用的是基于改进码本模型的背景相减法检测运动目标,对检测出的目标构建耦合模型进行实时跟踪.另外,本文还讨论了模型中参数改变对跟踪失败次数的影响.实验结果表明：该方法在处理目标上述情景下的跟踪效果比较理想,并且一定范围内的参数改变对该方法的稳定性影响不大.

A moving object tracking method based on adaptive coupled model

Current tracking methods on tracking objects which undergo rapid motion and significant appearance changes always lead to target tracking loss and large error. This paper proposes a new tracking method which detects moving targets by improved codebookmodelbased background subtractionmethod, and provides realtime tracking of detected targets by building coupling model. Furthermore,the influence on tracking effect caused by parameter changes is also discussed in this paper. The experimental results show the tracking effect of tracking targets in the mentioned situation is ideal and theproposed algorithm is nearly stable over a certain range of parameter changes.