【作者】 肖敬若； 胡伏原； 郑江滨； 张艳宁；

【Author】 XIAO Jing-ruo HU Fu-yuan ZHANG Yan-ning ZHENG Jiangbin (Computer School,Northwest Poly-technology University,Xi’an 710072)

【机构】 西北工业大学计算机学院；

【摘要】 本文针对自然环境下的多目标跟踪,提出了一种有效的实时跟踪算法。通过改进MeanShift算法,并结合Kalman滤波器,实现了对多类目标的高精度的稳健跟踪,解决了目标的遮掩问题。并提出了一种用于自动初始化目标的三层区域链交互结构,能有效的去除伪目标和提取真实目标。大量自然环境下的实验结果表明,本文算法对于复杂背景的目标跟踪具有有效性,实时性和实用性。更多还原

【Abstract】 An efficient multi-target tracking method has been presented,especially dealing with the tracking under complex nature environment.This method integrates the improved Mean Shift tracking method and Kalman Filtering,to stably track the targets,especially dealing with the ones occluded by the trees or buildings.It also proposes a Three-MotionRegion-List Interaction structure to initialize targets,which can erase out the false alarms and set up the real targets.Lots of experiments have shown that this method is efficient, practical and can perform in real-time.更多还原