融合HOG特征和注意力模型的孪生目标跟踪算法
作者:
作者单位:

沈阳理工大学

作者简介:

通讯作者:

中图分类号:

TP273

基金项目:

国家重点研发计划基金资助项目(2017YFC0821001);辽宁省教育厅科学研究经费项目(LG201909);辽宁省教育厅高等学校基本科研项目(LJKZ0275)


Twin target tracking network combining HOG features and attention model
Author:
Affiliation:

SHENYANG LIGONG UNIVERSITY

Fund Project:

Projects funded by the National Key R&D Program Fund(2017YFC0821001);Scientific Research Funding Project of the Education Department of Liaoning Province(LG201909);Basic Scientific Research Project of Higher Education Institutions of the Education Department of Liaoning Province(LJKZ0275)

  • 摘要
  • |
  • 图/表
  • |
  • 访问统计
  • |
  • 参考文献
  • |
  • 相似文献
  • |
  • 引证文献
  • |
  • 资源附件
  • |
  • 文章评论
    摘要:

    为提高跟踪算法在目标发生形变和被遮挡时的准确性,本文提出一种融合HOG(Histogram of Oriented Gradient,HOG)特征和注意力模型的孪生目标跟踪算法。首先,采用对ResNet残差模型改进后的CIR模型塑造孪生目标跟踪网络的骨干网络,充分利用不同层次的特征图,同时加深网络。其次,融入HOG特征,增强网络对图形几何变化的鲁棒性。再次,加入CBAM注意力模型,让网络能够在结合上下文信息的同时调节HOG特征在特征图中所占比例,增强特征图中的有效特征,弱化无效特征,使网络中各特征图发挥出最好的效果。最后,定义算法的损失函数。实验结果表明,所提算法在GOT-10k数据集上进行训练后,能够在OTB100上获得较好的跟踪效果,在该数据集中精确率和成功率分别达到81.9%和60.6%。在目标物体发生形变和被遮挡的情况下,所提算法仍能取得较好的跟踪效果。

    Abstract:

    In order to improve the accuracy of the tracking algorithm when the target is deformed and occluded, a twin target tracking algorithm integrating HOG (Histogram of Oriented Gradient, HOG) feature and attention model is proposed in this paper. First, the CIR model improved by ResNet residual model is used to shape the backbone network of twin target tracking network, and make full use of different levels of feature maps to deepen the network. Secondly, the HOG feature is integrated to enhance the robustness of the network to the geometric changes of graphics. Thirdly, the CBAM attention model is added to enable the network to adjust the proportion of HOG features in the feature map while combining the context information, enhance the effective features in the feature map, weaken the invalid features, and make each feature map in the network play the best effect. Finally, the loss function of the algorithm is defined. Experimental results show that the proposed algorithm can achieve good tracking effect on OTB100 after training on GOT-10k dataset, and the accuracy and success rates in this dataset are 81.9% and 60.6%, respectively. When the target object is deformed and occluded, the proposed algorithm can still achieve better tracking results.□□□□□□Key words:Object tracking;HOG feature;Attention model;Siamese network;Feature fusion;Residual network

    参考文献
    相似文献
    引证文献
引用本文
分享
文章指标
  • 点击次数:
  • 下载次数:
  • HTML阅读次数:
  • 引用次数:
历史
  • 收稿日期:2021-07-15
  • 最后修改日期:2022-09-25
  • 录用日期:2021-10-27
  • 在线发布日期: 2021-12-01
  • 出版日期: