复杂场景下动态目标视觉跟踪方法  

Vision based dynamic target tracking under complicated scenes

作　　者：毛建国[1] 柏方超[1] 沈峘[1] 周华鹏[1] 李芳培[1] 卢文玉[1] 

机构地区：[1]南京航空航天大学能源与动力学院,江苏南京210016

出　　处：《传感器与微系统》2010年第3期23-25,28,共4页Transducer and Microsystem Technologies

基　　金：江苏省普通高校研究生科研创新计划资助项目(CX08B_044Z)

摘　　要：当跟踪目标与背景颜色相近或跟踪目标被遮挡时,传统的均值平移算法失效。提出了一种改进的均值平移算法,将RGB颜色空间转换成HSI颜色空间,克服背景颜色与跟踪目标颜色相近时带来的干扰。当发生遮挡时,在较短时间内将目标的运动看作是不变系统,利用Kalman滤波器进行参数辨识使得发生遮挡后的跟踪系统具有后续状态预测的能力。实验结果证明:提出的算法能在复杂场景下准确地跟踪目标,在Intel 1.73 GHz处理器上的处理速率达39帧/s。Traditional mean shift will fail when the goal＇ s color is similar to the background or the goal is blocked. An improved mean shift tracking algorithm is presented. RGB color space is coverted into HSI color space. The interference is overcomed when the goal＇ s color is similar to the background. When the goal is blocked,a time-invariant system is used to describe the movement of the target during a short time sequences. Through Kalman filter, this system is identified so as to make it have the ability of estimating the coming states. Experimental results demonstrate that the proposed approach can accurately track the goal in complicated scenes, and can deal with 39 pictures per second with a Intel 1.73 GHz CPU.

关 键 词：均值平移 HSI颜色空间 卡尔曼滤波 状态预测 

分 类 号：U273.99[机械工程—车辆工程]