留言板

尊敬的读者、作者、审稿人, 关于本刊的投稿、审稿、编辑和出版的任何问题, 您可以本页添加留言。我们将尽快给您答复。谢谢您的支持!

姓名
邮箱
手机号码
标题
留言内容
验证码

红外偏振成像系统性能评估模型

王霞 赵家碧 孙琪扬 金伟其

王霞, 赵家碧, 孙琪扬, 金伟其. 红外偏振成像系统性能评估模型[J]. 红外技术, 2023, 45(5): 437-445.
引用本文: 王霞, 赵家碧, 孙琪扬, 金伟其. 红外偏振成像系统性能评估模型[J]. 红外技术, 2023, 45(5): 437-445.
WANG Xia, ZHAO Jiabi, SUN Qiyang, JIN Weiqi. Performance Evaluation Model for Infrared Polarization Imaging System[J]. Infrared Technology , 2023, 45(5): 437-445.
Citation: WANG Xia, ZHAO Jiabi, SUN Qiyang, JIN Weiqi. Performance Evaluation Model for Infrared Polarization Imaging System[J]. Infrared Technology , 2023, 45(5): 437-445.

红外偏振成像系统性能评估模型

基金项目: 

国家自然科学基金资助项目 62171024

详细信息
    作者简介:

    王霞(1972-),女,副教授,博士,主要从事图像处理、红外偏振成像、光电探测等方向的教学和研究工作。E-mail: angelniuniu@bit.edu.cn

  • 中图分类号: TN219

Performance Evaluation Model for Infrared Polarization Imaging System

  • 摘要: 红外偏振成像系统快速发展且应用广泛,但评估其性能的成像系统性能模型发展不足。迫切需要能够与先进的偏振成像系统相匹配的性能模型。利用深度学习网络的训练过程与人脑提取认知信息过程的相似性,本文首次将深度学习方法引入系统性能模型领域,提出了一种基于二维图像的可自动评估系统性能的红外偏振成像系统性能模型。该模型主要包含两个主要模块:退化模块、性能感知模块。在评估一个新的系统时,需要输入高质量的原始图像,并根据系统的硬件参数量身定制成像系统退化模块,退化完成后输入性能感知模块,从而得到最终的目标获取性能。为验证模型有效性,本文基于红外辐射理论自建了面向海面场景的红外偏振数据集,训练网络并进行测试。应用该模型对红外偏振成像系统的性能进行评估,评估结果与主观感知具有较好的一致性。
  • 图  1  基本系统性能模型结构(绿色)。灰色:两个主要模块; 蓝色:性能感知模块

    Figure  1.  Basic system performance model structure(green). Gray: two main components; Blue: performance awareness module

    图  2  PRI-YOLOv5的结构及输出定义示意图

    Figure  2.  Illustration of the PRI-YOLOv5 structure and output definition

    图  3  预测网络示意图

    Figure  3.  Illustration of our prediction network

    图  4  场景仿真示意图

    Figure  4.  Illustration of scene simulation

    图  5  仿真结果示例

    Figure  5.  An example of the simulation results

    图  6  PRI-YOLOv5训练结果

    Figure  6.  Training results of PRI-YOLOv5

    图  7  预测网络测试结果

    Figure  7.  Test results of the prediction network

    图  8  两组数据预测差值与其相似性的关系

    Figure  8.  The difference of the predicted results v.s. NMI

    图  9  几款待评估偏振系统的退化效果示意图

    Figure  9.  Schematic diagram of degradation effects of several polarization systems

    图  10  不同距离处的目标获取概率(三款系统)

    Figure  10.  Target acquisition probability as a function of range(three systems)

    表  1  红外偏振成像系统性能模型研究现状

    Table  1.   Research status of performance models for infrared polarization imaging systems

    Name Year Principle Illustration
    Edson Guimaraes[9] 1999 MRTD, Johnson Criterion Based on the MRTD, this model further consider the transfer function and transmittance of the polarizer.
    Mehmet Yildirim[10] 2000 Designed for the second generation forward looking infrared sensor.
    Zhou Chenghao[11] 2013 Emphatically analyzed the models of different targets such as point source targets and extended source targets.
    Xia Runqiu[12] 2016 Ignore registration errors and the impact of the polarizer on MTF, the MRTD model of the polarization system is calculated.
    Liang Jianan[13] 2019 Modify the MRTD model and Johnson criterion based on interference factors such as background clutter.
    下载: 导出CSV

    表  2  船模型及类别

    Table  2.   Ship models and classes

    Target class
    Frigate Destroyer Patrol
    Target type ship1 Size
    X: 0.280m
    Y: 1.948m
    ship3 Size
    X: 0.436m
    Y: 3.919m
    ship5 Size
    X: 0.367m
    Y: 2.001m
    ship2 Size
    X: 0.344m
    Y: 2.830m
    ship4 Size
    X: 0.395m
    Y: 3.108m
    ship6 Size
    X: 0.482m
    Y: 2.000m
    Note: All 3D models are downloaded from https://www.3d66.com/
    下载: 导出CSV

    表  3  仿真中变量和常量参数设置

    Table  3.   Variables and constant settings during simulation

    Variables
    Group 1: Model ship Group 2: Real ship
    Name Number of values Values(unit) Number of values Values(unit)
    Focal length(f) 2 60, 85(mm) 1 105(mm)
    Pixel size(p) 3 14, 17, 20(μm) 3 14, 17, 20(μm)
    View radius(r) 2 100, 150(m) 4 2, 3, 4, 5(km)
    View zenith angle 4 30, 45, 60, 75(°) 4 30, 45, 60, 75(°)
    Wind speed 1 2(m/s) 1 13(m/s)
    Sea surface size 1 8 m×8 m 1 338 m×338 m
    Target class 6/3 see Table 1 6/3 see Table 1
    Constants
    Name Values Name Values
    Solar zenith angle/° 45 Solar azimuth angle/° 60
    Max reflected number 3 Detected wavelength/μm 3.8
    Image size 300×300 Sea temperature/ K 300
    Deck temperature/K 318.15 Hull temperature/K 303.15
    Hot part temperature/K 328.15 Others temperature/K 313.15
    下载: 导出CSV

    表  4  待评估红外偏振成像系统主要参数

    Table  4.   The main parameters of infrared polarization imaging systems to be evaluated

    System Focal length/mm Wave-length/μm F# Resolution Pixel size/μm
    System A 105 3~5.2 3.5 160×120 20
    System B 135 3~5.2 1.5 640×480 17
    System C 105 3~5.2 2 320×240 20
    下载: 导出CSV
  • [1] Scott L B, Condiff L R. C2NVEO advanced FLIR systems performance model[C]//Proc. SPIE, 1990, 1309: 168-180.
    [2] Scott L B, D'Agostino J A. NVEOD FLIR92 thermal imaging systems performance model[C]//Proc. SPIE, 1992, 1689: 194-203.
    [3] Maurer T, Driggers R G, Vollmerhausen R H. 2002 NVTherm improve-ments[C]//Proc. SPIE, 2002, 4719: 15-23.
    [4] Edwards T C, Vollmerhausen R H, Driggers R G, et al. NVESD time-limited search model[C]//Proc. SPIE, 2003, 5076: 53-59.
    [5] Bijl P, Valeton J M. Triangle orientation discrimination: the alternative to minimum resolvable temperature difference and minimum resolvable contrast[J]. Optical Engineering, 1998, 37: 1976-1983. doi:  10.1117/1.601904
    [6] Vollmerhausen R H, Jacobs E and Driggers R G. New Metric for Predicting target acquisition Performance [J]. Optical Engineering, 2004, 43(11): 2806-2818. doi:  10.1117/1.1799111
    [7] Greif H J, Weiss A R, Wittenstein W. pcSitoS: A new tool for image-based IR system simulation [C]//Proc. SPIE, 2009, 7481: 748107.
    [8] Hogervorst M A, Bijl P, Valeton J M. Capturing the sampling effects: a TOD sensor performance model [C]//Proc. SPIE, 2001, 4372: 62-73.
    [9] Guimaraes E F. Investigation of minimum resolvable temperature difference formulation for polarized thermal imaging range prediction[D]. America: Naval Postgraduate School, 1999.
    [10] Yildirim M. Modeling second generation FLIR sensor detection recognition and identification range with polarization filtering[D]. America: Naval Postgraduate School, 2000.
    [11] 周程灏. 红外偏振成像系统作用距离建模与分析[D]. 哈尔滨: 哈尔滨工业大学, 2013.

    ZHOU Chenghao. Modeling and Analysis of the Operating Range of Infrared Polarization Imaging System[D]. Harbin: Harbin Institute of Technology, 2013.
    [12] 夏润秋, 王霞, 金伟其, 等. 海面环境中红外偏振成像系统作用距离模型[J]. 红外与激光工程, 2016, 45(3): 74-78. https://www.cnki.com.cn/Article/CJFDTOTAL-HWYJ201603010.htm

    XIA Runqiu, WANG Xia, JIN Weiqi, et al. Distance model of infrared polarization imaging system used in sea-surface environment[J]. Infrared and Laser Engineering, 2016, 45(3): 74-78. https://www.cnki.com.cn/Article/CJFDTOTAL-HWYJ201603010.htm
    [13] 梁建安. 基于红外偏振成像技术的水面杂波抑制方法研究[D]. 北京: 北京理工大学, 2019.

    LIANG Jianan. Research on Water Surface Clutter Suppression Method Based on Infrared Polarization Imaging Technology[D]. Beijing: Beijing Institute of Technology, 2019.
    [14] ZHANG J Q, WANG X R. Modeling and Performance Evaluation Theory of Photoelectric Imaging System[M]. Xi'an: Xidian University Publishing House, 2010.
    [15] Redmon J, Divvala S, Girshick R. You only look once: Unified, real-time object detection[C]//IEEE Conference on Computer Vision and Pattern Recognition, 2016: 779-788.
    [16] Jocher G. YOLOv5 release v4.0[EB/OL]. [2023-01-20]. https://github.com/ultralytics/yoloV5.
    [17] SU Shaolin, YAN Qingsen, ZHU Yu, et al. Blindly assess image quality in the wild guided by a self-adaptive hyper network[C]//IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2020: 3664-3673.
    [18] HE Si, WANG Xia, XIA Runqiu, et al. Polarimetric infrared imaging simulation of a synthetic sea surface with Mie scattering[J]. Applied Optics, 2018, 57(7): B150-B159. doi:  10.1364/AO.57.00B150
  • 加载中
图(10) / 表(4)
计量
  • 文章访问数:  196
  • HTML全文浏览量:  63
  • PDF下载量:  79
  • 被引次数: 0
出版历程
  • 收稿日期:  2023-04-14
  • 修回日期:  2023-04-28
  • 刊出日期:  2023-05-20

目录

    /

    返回文章
    返回