基于特征融合与背景加权的红外目标跟踪

针对传统均值漂移算法无法对与背景相近红外目标进行有效跟踪的问题,提出了一种改进均值漂移(Mean Shift)算法。首先,融合了灰度和纹理两方面的信息以增加目标描述的信息量,接着为了减少背景像素对跟踪定位的影响,通过目标区域周围像素的颜色直方图定义背景加权系数,并将该系数引入到目标模型的灰度直方图和纹理直方图的计算中,进而实现目标的准确定位,最后,给出了目标模型更新方法。实验结果表明,文中算法能够抑制背景干扰,对与背景相似的目标进行有效的跟踪。

Infrared target tracking based on feature fusion and background weighting

For the target tracking based on Mean Shift may be lost when the infrared target has a low signal-to-noise ratio(SNR),an improved Mean Shift tracking method is approached.First the features of gray and texture are fused to enhance the target information.Then to reduce the localization error of object tracking,coefficients based on the color histograms of the background pixels around the target are computed and incorporated into the computation of gray and texture spatial histograms.Meanwhile,the criterion o...