空间邻近红外小目标解混资源精选

作者: 许孟泽, 翟曦盟, 韩圣东, 戴一冕*

English Version: Link to English Version

一个关于空间邻近红外小目标解混 (Closely-Spaced Infrared Small Target Unmixing, CSIST Unmixing) 技术的精选资源列表(包含论文、代码、数据集等)。

空间邻近红外小目标解混 是红外搜索与跟踪系统中的一项关键且具有挑战性的任务。它专注于解混和检测焦平面上彼此非常靠近的多个弱小目标,这些目标通常会光学扩散现象导致红外成像混叠严重,多目标可能集中在几个像素中,导致难以准确识别目标的数量和位置。本资源库旨在收集和整理这一特定领域的最新进展。

目录

论文

按年份排序

2025

  • DISTA-Net: Dynamic Closely-Spaced Infrared Small Target Unmixing - Shengdong Han, Shangdong Yang, Xin Zhang, Yuxuan Li, Xiang Li, Jian Yang, Ming-Ming Cheng, Yimian Dai, ICCV 2025
    [论文] [代码 ⭐ ]
  • SeqCSIST: Sequential Closely-Spaced Infrared Small Target Unmixing - Ximeng Zhai, Bohan Xu, Yaohong Chen, Hao Wang, Kehua Guo and Yimian Dai, TGRS 2025
    [论文] [代码 ⭐ ]

2024

  • A Resolution and Localization Algorithm for Closely-Spaced Objects Based on Improved YOLOv5 Joint Fuzzy C-Means Clustering - Li et al., IEEE Photonics Journal, 2024
    [论文]

2023

  • Closely-Spaced Object Classification Using MuyGPyS - Zhang et al., Advanced Maui Optical and Space Surveillance Technologies Conference (AMOS), 2023
    [论文]

2022

  • Closely spaced object detection utilizing spatial information in spectroastrometric observations - J. Zachary Gazak, Ryan Swindle, Zachary Funke, Matthew Phelps, Justin Fletcher, Sensors and Systems for Space Applications XV. SPIE, 2022
    [论文]

2020

  • 采用分裂 Bregman 的空间邻近目标红外超分辨算法 - 左芝勇, 电讯技术, 2020
    [论文]

Pre-2020

  • The infrared image closely spaced objects super resolution method based on sparse reconstruction under the noise environment - J Zeng, J Yang, H Wu, International Conference on Optical and Photonics Engineering (icOPEN 2016). SPIE, 2017
    [论文]

  • Electromagnetic Imaging of Closely Spaced Objects using Matching Pursuit Based Approaches - Şenyuva, R. V., Özdemir, Ö., Kurt, G. K., & Anarım, IEEE Antennas and Wireless Propagation Letters, 2015
    [论文]

  • Bayesian approach to joint super-resolution and trajectory estimation for midcourse closely spaced objects via space-based infrared sensor - Liangkui Lin, Weidong Sheng, Dan Xu, Optical Engineering, 2012
    [论文]

  • QPSO-based algorithm of CSO joint infrared super-resolution and trajectory estimation - Lin, Liangkui and Xu, Hui and Xu, Dan and An, Wei and Xie, Kai, Journal of Systems Engineering and Electronics, 2011
    [论文]

  • 基于 Gibbs 抽样的红外成像小间距目标分辨方法 - 刘涛, 信号处理, 2010
    [论文]

  • Hierarchical Closely-Spaced Object (CSO) Resolution for IR Sensor Surveillance - Macumber, Daniel and Gadaleta, Sabino and Floyd, Allison and Poore, Aubrey, Signal and Data Processing of Small Targets, 2005
    [论文]

  • Model-based superresolution CSO processing - John T. Reagan, Theagenis J. Abatzoglou, Signal and Data Processing of Small Targets, 1993
    [论文]

按方法分类

基于模型 / 优化方法

  • Closely spaced object detection utilizing spatial information in spectroastrometric observations - J. Zachary Gazak, Ryan Swindle, Zachary Funke, Matthew Phelps, Justin Fletcher, Sensors and Systems for Space Applications XV. SPIE, 2022
    [论文]

  • 采用分裂 Bregman 的空间邻近目标红外超分辨算法 - 左芝勇, 电讯技术, 2020
    [论文]

  • The infrared image closely spaced objects super resolution method based on sparse reconstruction under the noise environment - J Zeng, J Yang, H Wu, International Conference on Optical and Photonics Engineering (icOPEN 2016). SPIE, 2017
    [论文]

  • Electromagnetic Imaging of Closely Spaced Objects using Matching Pursuit Based Approaches - Şenyuva, R. V., Özdemir, Ö., Kurt, G. K., & Anarım, IEEE Antennas and Wireless Propagation Letters, 2015
    [论文]

  • Bayesian approach to joint super-resolution and trajectory estimation for midcourse closely spaced objects via space-based infrared sensor - Liangkui Lin, Weidong Sheng, Dan Xu, Optical Engineering, 2012
    [论文]

  • QPSO-based algorithm of CSO joint infrared super-resolution and trajectory estimation - Lin, Liangkui and Xu, Hui and Xu, Dan and An, Wei and Xie, Kai, Journal of Systems Engineering and Electronics, 2011
    [论文]

  • 基于 Gibbs 抽样的红外成像小间距目标分辨方法 - 刘涛, 信号处理, 2010
    [论文]

  • Hierarchical Closely-Spaced Object (CSO) Resolution for IR Sensor Surveillance - Macumber, Daniel and Gadaleta, Sabino and Floyd, Allison and Poore, Aubrey, Signal and Data Processing of Small Targets, 2005
    [论文]

  • Model-based superresolution CSO processing - John T. Reagan, Theagenis J. Abatzoglou, Signal and Data Processing of Small Targets, 1993
    [论文]

基于深度学习的方法

  • DISTA-Net: Dynamic Closely-Spaced Infrared Small Target Unmixing - Shengdong Han, Shangdong Yang, Xin Zhang, Yuxuan Li, Xiang Li, Jian Yang, Ming-Ming Cheng, Yimian Dai, ICCV 2025
    [论文] [代码 ⭐ ]

  • SeqCSIST: Sequential Closely-Spaced Infrared Small Target Unmixing - Ximeng Zhai, Bohan Xu, Yaohong Chen, Hao Wang, Kehua Guo and Yimian Dai, TGRS 2025
    [论文] [代码 ⭐ ]

  • A Resolution and Localization Algorithm for Closely-Spaced Objects Based on Improved YOLOv5 Joint Fuzzy C-Means Clustering - Li et al., IEEE Photonics Journal, 2024
    [论文]

  • Closely-Spaced Object Classification Using MuyGPyS - Zhang et al., Advanced Maui Optical and Space Surveillance Technologies Conference (AMOS), 2023
    [论文]

数据集与基准

  • CSIST-100K - 一个用于空间邻近红外点目标解混的大规模合成数据集(10万样本)。模拟每张图像包含 1-5 个目标,扩散函数标准差为 σ=0.5 像素,最小间距 ≥0.52 瑞利单位,且强度随机。目标在 3×3 区域内显著重叠,对计数和定位提出了严峻挑战。(按 80k/10k/10k 划分)
    [百度网盘 OneDrive]

  • SeqCSIST - 序列空间邻近红外小目标解混数据集
    一个用于序列空间邻近红外小目标解混的序列基准数据集。这是一个合成数据集,生成的图像尺寸为11×11像素。每张图像包含2到4个目标,强度从特定范围内随机采样。目标遵循随机轨迹。目标渲染基于84%能量集中分辨率标准和0.5像素的扩散标准差。XML文件中提供每个目标的精确坐标和强度等真实值。
    [百度网盘 OneDrive]

  • 自定义仿真数据集 - 自定义红外目标仿真数据生成方法 \

    • 中段弹道目标群的红外成像仿真研究 - 林两魁, 谢恺, 徐晖, 红外与毫米波学报, 2009. [论文]
    • 弹道目标识别的红外辐射数据仿真研究 - 刘俊良, 陈尚锋, 卢焕章, 红外与激光工程, 2016. [论文]
    • 深空动态场景目标红外图像仿真研究 - 李志军, 王卫华, 陈曾平, 红外技术, 2007. [论文]
    • 空间目标在轨红外成像仿真 - 王盈, 黄建明, 魏祥泉, 红外与激光工程, 2015. [论文]
    • 红外运动目标轨迹重构动态仿真平台 - 姚成喆, 郭伟兰, 陈钱, 红外与激光工程, 2022. [论文]
    • 天基空间小目标复杂场景数字成像仿真 - 李鹏飞, 徐伟, 朴永杰, 系统仿真学报, 2025. [论文]
    • 红外成像系统噪声测量仿真研究 - 邹前进, 戴睿, 刘鑫, 红外技术, 2008. [论文]
    • Exploring Video Denoising in Thermal Infrared Imaging: Physics-Inspired Noise Generator, Dataset, and Model - Cai L, Dong X, Zhou K,IEEE Transactions on Image Processing, 2024. [论文]

评估指标

  • CSO-mAP: 空间邻近红外小目标平均检测精度 在多个亚像素距离阈值 (δ=0.05 - 0.25px) 上的平均精度的均值。 受平均检测精度(mAP)启发的指标,旨在评估目标间距小于瑞利判据场景下的精确定位性能。[论文] [代码 ⭐ ]

相关研究团队

  • GrokCV - 南开大学 - 由戴一冕副教授领导。长期致力于红外弱小目标检测和遥感多模态视觉感知。