基于时空显著性建模的空中飞行器跟踪方法

Airborne target tracking based on spatio-temporal saliency modeling

以复杂背景下空中飞行器的鲁棒视觉跟踪问题为研究背景,为解决现有跟踪方法目标表征模型不够精确,算法鲁棒性严重受到目标形变、宽高比变化、复杂背景等因素干扰的问题,提出了建模跟踪场景中独立物体的显著性特性,用于构建精确的目标模型。提出的显著性估计方法有别于传统的单帧检测方法,利用跟踪算法提供的背景先验知识以及多帧图像观测数据,使用时空联合的方式进行建模估计,其结果用来指导目标跟踪算法选取有效视觉特征,建立精确目标表征模型,减小背景区域对算法模型的干扰。实验表明,提出的方法为上述难点问题提供了有效的解决方案,对空中飞行器的跟踪精度与鲁棒性优于大多数最先进的主流方法,在其他类型的目标跟踪任务中也有十分优越的性能表现。

This paper studies the robust visual tracking problem of airborne aircraft in complex environments.To improve the robustness in tracking airborne targets against challenging factors like target deformation,aspect ratio variation,and complex background,this paper models the saliency characteristics of the independent objects in the tracking scene and builds a more accurate target representation model.Different from traditional single-frame saliency detection methods,this paper proposes a spatio-temporal saliency estimation model,which makes use of prior knowledge of the target provided by the tracking algorithm and multi-frame observation data.The estimated saliency map is used to select of effective visual features and build accurate target representation in the tracking algorithm,which reduces the interference of the background clutter.Experimental results show that the proposed method provides an effective solution to the above problems in airborne tracking tasks and outperforms most state-of-the-art methods in accuracy and robustness.Moreover,the method shows excellent performance in tracking other objects.