基于在线集成学习的红外点目标跟踪研究  

Infrared Small Target Tracking Algorithm Based on Online Ensemble Learning

作　　者：武文成 艾斯卡尔·艾木都拉[1] WU Wencheng;ASKAR Hamdulla(Institute of In formation Science and Engineering,Xinjiang University,Urumqi 830046,China})

机构地区：[1]新疆大学信息科学与工程学院

出　　处：《山西大学学报（自然科学版）》2019年第4期755-761,共7页Journal of Shanxi University(Natural Science Edition)

基　　金：国家自然科学基金(61563049)

摘　　要：通过研究红外点目标特征,提出了一种基于在线集成学习的红外点目标跟踪算法,在跟踪过程中调整分类器,具有较强的鲁棒性。同时,由于红外点目标只有灰度信息、无纹理、无形状。选择最具辨别红外点目标的MS-LSGM算法计算特征,从而产生稳定的跟踪结果。实验结果证明,在复杂多变的背景下,方法具有较好的鲁棒性,实时性。Through studying the feature of infrared point target,this paper proposes an infrared point target tracking algorithm based on online ensemble learning, which adjusts the classifier during the tracking process, so that the algorithm has strong robustness.Moreover, since the infrared point target has only grayscale features, no texture, no shape, this paper selects the MS-LSGM algorithm with the ability to distinguish the infrared point target to calculate the characteristics, thus producing stable tracking results. The experimental results show that the proposed method has good robustness and real-time performance under the complex and variable background.

关 键 词：红外点目标 点目标跟踪 集成学习 目标增强 

分 类 号：TP751[自动化与计算机技术—检测技术与自动化装置]