《计算机应用》唯一官方网站 ›› 2024, Vol. 44 ›› Issue (2): 563-571.DOI: 10.11772/j.issn.1001-9081.2023020167

• 多媒体计算与计算机仿真 • 上一篇    

基于孪生网络和Transformer的红外弱小目标跟踪方法

崔晨辉1, 蔺素珍1(), 李大威2, 禄晓飞3, 武杰1   

  1. 1.中北大学 计算机科学与技术学院,太原 030051
    2.中北大学 电气与控制工程学院,太原 030051
    3.酒泉卫星发射中心,甘肃 酒泉 735000
  • 收稿日期:2023-02-23 修回日期:2023-04-24 接受日期:2023-05-05 发布日期:2024-02-22 出版日期:2024-02-10
  • 通讯作者: 蔺素珍
  • 作者简介:崔晨辉(1998—),男,山西晋城人,硕士研究生,CCF会员,主要研究方向:目标跟踪、图像处理
    李大威(1980—),男,河北衡水人,副教授,博士,主要研究方向:模式识别、机器学习、图像处理
    禄晓飞(1981—),男,河南许昌人,博士,主要研究方向:测量数据处理
    武杰(1998—),男,山西吕梁人,硕士研究生,CCF会员,主要研究方向:红外弱小目标检测。
  • 基金资助:
    山西省研究生教育创新项目(2022Y631);山西省专利转化专项计划项目(202302001);山西省级教改项目(J20220593);中北大学第十八届研究生科技立项项目(20221845)

Infrared dim small target tracking method based on Siamese network and Transformer

Chenhui CUI1, Suzhen LIN1(), Dawei LI2, Xiaofei LU3, Jie WU1   

  1. 1.College of Data Science and Technology,North University of China,Taiyuan Shanxi 030051,China
    2.School of Electrical and Control Engineering,North University of China,Taiyuan Shanxi 030051,China
    3.Jiuquan Satellite Launch Center,Jiuquan Gansu 735000,China
  • Received:2023-02-23 Revised:2023-04-24 Accepted:2023-05-05 Online:2024-02-22 Published:2024-02-10
  • Contact: Suzhen LIN
  • About author:CUI Chenhui, born in 1998, M. S. candidate. His research interests include target tracking, image processing.
    LI Dawei, born in 1980, Ph. D., associate professor. His research interests include pattern recognition, machine learning, image processing.
    LU Xiaofei, born in 1981, Ph. D. His research interests include measurement data processing.
    WU Jie, born in 1998, M. S. candidate. His research interests include infrared dim small target detection.
  • Supported by:
    Graduate Education Innovation Project of Shanxi Province(2022Y631);Shanxi Province Patent Transformation Special Plan Project(202302001);Education Reform Project of Shanxi Province(J20220593);18th Graduate Technology Project of North University of China(20221845)

摘要:

针对红外弱小目标跟踪准确性较低这一问题,提出一种基于孪生网络和Transformer的红外弱小目标跟踪方法。首先,构建多特征提取级联模块分别提取红外弱小目标模板帧和搜索帧的深度特征,并将二者分别与其对应的HOG特征进行维度层面的串联;其次,引入多头注意力机制Transformer进行模板特征图和搜索特征图的互相关操作,生成响应图;最后,通过响应图上采样网络和边界框预测网络,获得目标在图像的中心位置和回归边界框,完成对红外弱小目标的跟踪。在包含13 655张红外图像数据集上的测试结果表明:与KeepTrack跟踪方法相比,成功率提高5.9个百分点,精确率提高1.8个百分点;与TransT(Transformer Tracking)方法相比,成功率提高14.2个百分点,精确率提高14.6个百分点,证明所提方法对复杂背景下的红外弱小目标跟踪准确性更高。

关键词: 目标跟踪, 红外弱小目标, 孪生网络, Transformer, 多特征提取

Abstract:

A method based on Siamese network and Transformer was proposed to address the low accuracy problem of infrared dim small target tracking. First, a multi-feature extraction cascading moduling was constructed to separately extract the deep features of the infrared dim small target template frame and the search frame, and concatenate them with their corresponding HOG features at the dimension level. Second, a multi-head attention mechanism Transformer was introduced to perform cross-correlation operations between the template feature map and the search feature map, generating a response map. Finally, the target’s center position in the image and the regression bounding box were obtained through the response map upsampling network and bounding box prediction network to complete the tracking of the infrared dim small targets. Test results on a dataset of 13 655 infrared images show that compared with KeepTrack tracking method, the success rate is improved by 5.9 percentage points and the precision is improved by 1.8 percentage points; compared with TransT (Transformer Tracking) method, the success rate is improved by 14.2 percentage points and the precision is improved by 14.6 percentage points. The proposed method is proved to be more accurate in tracking infrared dim small targets in complex backgrounds.

Key words: target tracking, infrared dim small target, Siamese network, Transformer, multi-feature extraction

中图分类号: