雨夜条件下的红外可见光视频融合目标跟踪

Infra-visible video fusion object tracking under rainy night condition

针对雨夜条件下目标外观特性衰退,造成目标搜索与跟踪过程中难以适应混杂背景干扰及雨滴突变干扰的问题,提出了一种基于背景抑制的红外可见光融合目标跟踪方法。该方法考虑了目标跟踪时间和空间两方面的外观分布特性,采用了多通道成像数据对前景目标进行特征表示,有效抑制了夜间条件下的混杂背景干扰。在目标模型更新过程中,本文通过分析雨滴突变造成的目标表示变化规律,采用突变检测模型,降低了雨滴突变对模型更新的干扰。实验采用了真实环境采集的雨夜条件的红外可见光视频,实验结果说明,本文方法优于现有目标跟踪方法,能够有效提高雨夜条件下的目标跟踪精度。

Aiming at the problem that object appearance characteristic degenerates under rainy night condition,so that object searching and tracking are susceptible to clutter background and raindrop mutation interferences.In this paper,an infra-visible fusion object tracking method based on background suppression is proposed.The method considers the appearance distribution characteristics in both time and space aspects of object tracking,and adopts multi-channel imaging data to represent the characteristics of the foreground object,which effectively suppresses the clutter background interference under nighttime condition.Furthermore,in the object model update process through analyzing the variation rule of object representation caused by raindrop mutation,the mutation detection model is adopted to reduce the interference of raindrop mutation on model updating.In the experiment the infra-visible videos collected under rainy night condition in real environment were adopted,and the experiment results show that the proposed method is superior to existing object tracking methods and can effectively improve the object tracking accuracy under rainy night condition.