基于运动特征的颜色粒子滤波器  

Color Particle Filter Based on Motion Feature

作　　者：蒋斐[1] 程玉宝[1] 李宏斌[1] 

机构地区：[1]合肥电子工程学院脉冲功率激光技术国家重点实验室,合肥230037

出　　处：《计算机工程》2011年第9期221-222,225,共3页Computer Engineering

摘　　要：传统颜色粒子滤波器不能对跟踪状态进行自我判断,粒子容易失效,导致目标跟偏甚至跟丢。针对该问题,设计一种基于运动特征的颜色粒子滤波器。在分析目标运动特征的基础上,改进系统运动模型,建立自适应背景颜色尺度;在跟踪过程中通过最优估计速度和加速度来判断粒子滤波器的跟踪状态,建立相应的捕获机制以跟踪目标。实验结果证明,改进的颜色粒子滤波器能对与背景相似的运动目标进行有效准确的跟踪。An improved color particle filter is designed in order to solve the problem that traditional color particle filter can not proceed self-judgment for tracking state,produce incorrect particle and last cause inaccuracy or even failure.System motion model is improved and an adaptive color-scale for local background is established based on motion target feature.Tracking state of particle filter is judged according to optimal estimation of the speed and acceleration during the tracking period.Corresponding capture system is founded for tracking object.Experimental result shows that the improved color particle filter is more effective when both the motion target and the background are similar.

关 键 词：颜色粒子滤波器 跟踪状态 运动特征 系统运动模型 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]