摘要
针对机场环境,提出一种特定目标跟踪算法。研究核相关滤波算法,在外观特征提取方面融合多种特征,克服背景与目标相似度较高以及光照变化的影响;同时引入一种尺度自适应方法,以更好地适应目标尺度变化;还设计一种模型实时更新策略,以避免分类器学习到错误的目标信息。实验结果表明,所提出的算法在机场环境下跟踪精度较高、鲁棒性也更强。
Aiming at the airport environment, proposes a specific target tracking algorithm. Studies the kernel correlation filtering algorithm, and in the aspect of appearance feature extraction, a variety of features are fused to overcome the high similarity between background and target and illumination changes. At the same time, introduces a scale adaptive method to better adapt to the change of target scale. Designs a realtime updating strategy to avoid the classifier from learning the wrong target information. Experimental results show that the proposed algorithm has higher tracking accuracy and stronger robustness in airport environment.
作者
赵康
王正勇
何小海
熊杰
郑新波
ZHAO Kang;WANG Zheng-yong;HE Xiao-hai;XIONG Jie;ZHENG Xin-bo(College of Electronics and Information Engineering,Sichuan University,Chengdu 610065;Dongguan Institute of Advanced Technology,Dongguan 523000)
出处
《现代计算机》
2018年第21期11-15,共5页
Modern Computer
基金
国家自然科学基金(No.11176018)
成都市产业集群协同创新项目(No.2016-XT00-00015-GX)
东莞市社会科技发展项目(No.2017507102428)
关键词
核相关滤波
尺度自适应
模型更新
鲁棒性
Kernel Correlation Filtering
Scale Adaptation
Model Updating
Robustness