摘要
为提高目标亮度突变时的跟踪性能,在每一帧进行目标跟踪时,首先提取可见光图像的颜色特征,红外图像的垂直投影图像和水平投影图像特征,然后利用可见光/红外各自通道的Bhattacharyya系数计算该通道的权值,并对加权mean-shift双通道跟踪方法进行了推导,提出了基于加权mean-shift可见光/红外双通道目标跟踪算法.该方法使前后两帧目标相似度大的通道取大的权值,从而达到有效利用各通道有利信息、提高跟踪性能的目的.实验表明,用本文提出的算法进行可见光/红外双通道目标跟踪时,与基于mean-shift单通道(可见光或红外)目标跟踪算法相比,可提高目标跟踪的准确度,特别是当目标进入树荫区域,引起目标亮度发生显著变化时,跟踪性能基本不受影响.
According to the visible and infrared image sequences, when tracking in every frame, color features of visible image , vertical projective image and horizontal projective image features of infrared image are extracted. The weight in each channel was calculated with Bhattacharyya coefficient, and the weighted mean-shift tracking algorithm in dual-channel videos was derived. So visible/infrared dual-channel target tracking algorithm based on weighted mean-shift was proposed to improve tracking performance while luminance of object is changed. The proposed algorithm can effectively use the better information of each channel,for it makes the channel whose Bhattacharyya coefficient is higher have higher weight. The experiment in object tracking demonstrates that the proposed algorithm overall performance outperforms one channel(visible or infrared) target tracking algorithm based on mean-shift,and it almost remains the tracking performance when luminance of object is changed.
出处
《光子学报》
EI
CAS
CSCD
北大核心
2010年第2期357-363,共7页
Acta Photonica Sinica
基金
国家自然科学基金重点项目(60634030)
高等学校博士学科点专项科研基金(20060699032)
航空科学基金(2006ZC53037
2007ZC53037)资助