基于信息融合的运动目标跟踪算法  

Fusion information-based moving tracking algorithm

作　　者：曹洁[1] 李延林[1] 

机构地区：[1]兰州理工大学计算机与通信学院,兰州730050

出　　处：《电子测量与仪器学报》2011年第3期211-217,共7页Journal of Electronic Measurement and Instrumentation

摘　　要：针对复杂背景下目标遮挡后跟踪不稳定的问题,提出了一种在粒子滤波框架下融合异类信息进行目标跟踪的算法。该方法建立了声源特征模型和颜色特征模型,在此基础上将声音信息和图像的颜色信息融合到粒子滤波跟踪算法中,通过粒子更新策略实现跟踪。实验结果表明与基于单一特征的跟踪算法相比均方误差降低了18.3%。在低信噪比情况下,该算法更加稳健、有效。Aimed at the tracking-unstable problem when object is shaded in complex background,an object tracking algorithm based on alien information fusion is presented under particle filter frame.Sound source feature model and color feature model are built by this method.Besides,audio and image color information are fused into the particle filter track-ing algorithm based on this method and tracking is realized using particle update strategy.Experiments show that the mean square error decreases by 18.3% compared with the single feature tracking algorithm and this algorithm is more stable and effective in low SNR situation.

关 键 词：粒子滤波 目标跟踪 信息融合 声源定位 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]