基于双模型融合的自适应目标跟踪算法

发布时间：2019-02-13 07:24

【摘要】：针对目标跟踪过程中的光照变化、背景混乱和目标形变等问题,提出一种背景抑制的HS直方图和核相关滤波双模型融合的自适应跟踪算法。首先引入非线性核相关滤波跟踪模型;其次提出背景抑制的HS颜色直方图跟踪模型,通过分离亮度分量以减小光照干扰,并采用背景加权突出目标信息;然后提出一种自适应融合策略,根据目标与背景的HS特征相似度来动态调整两个模型融合权重,以降低背景混乱和目标姿态变化的影响;最后针对目标尺度变化问题,采用尺度金字塔估计策略进行解决。在多个公开数据集下的对比实验表明,与现有算法相比,提出的算法能更好地降低光照、背景混乱等复杂因素干扰,且达到了工程应用的实时性要求。
[Abstract]:An adaptive tracking algorithm based on the fusion of HS histogram and kernel correlation filter is proposed to solve the problems of illumination variation, background confusion and target deformation in the process of target tracking. Firstly, the nonlinear kernel correlation filter tracking model is introduced; secondly, the background suppression HS color histogram tracking model is proposed. The brightness component is separated to reduce the illumination interference, and the background weighting is used to highlight the target information. Then an adaptive fusion strategy is proposed to dynamically adjust the fusion weights of the two models according to the similarity of HS features between the target and the background in order to reduce the influence of background confusion and target attitude change. Finally, the scale pyramid estimation strategy is used to solve the problem of target scale change. Compared with the existing algorithms, the proposed algorithm can better reduce the interference of complex factors such as illumination and background confusion, and meet the real-time requirements of engineering applications.
【作者单位】：国家数字交换系统工程技术研究中心;
【基金】：国家自然科学基金资助项目(61379151) 创新群体资助项目(61521003) 国家科技支撑计划资助项目(2014BAH30B01) 青年科学基金资助项目(61601513)
【分类号】：TP391.41

【相似文献】