Method for Generating Infrared Image from Visible Image of Water Surface Targets
-
摘要: 红外警戒系统、红外成像制导导弹等军事装备在进行性能评估和模拟训练过程中都需要大量红外仿真图像,但目前红外仿真软件普遍存在生成红外仿真图像逼真度差、软件普适性不好等问题,且国外技术封锁造成我国红外仿真软件发展缓慢。因此,针对国内可见光图像仿真技术日趋成熟的现状,为提高红外仿真图像质量,本文提出了一种采用循环生成对抗网络、由可见光图像生成红外仿真图像的方法,并通过实验验证该算法是有效可行的。该算法首先通过区域生长算法从采集的可见光图像中提取水上目标,建立了水上目标可见光图像生成红外图像的训练数据集;然后利用训练好的网络生成红外仿真图像。测试实验表明,采用这种方法所生成的水上目标红外仿真图像视觉效果接近真实红外图像,可实际应用于海军红外军事装备模拟试验和训练系统。Abstract: Military equipment, such as infrared warning systems and infrared image-guided missiles, requires a number of infrared simulation images during performance evaluation and simulation training. However, the infrared simulation image generated by infrared simulation software has some problems, such as poor fidelity and poor universality. The domestic infrared simulation technology is hampered by foreign technology blockade. Currently, the development situation is difficult in meeting the requirement of the application. To solve this problem, this experiment proposed a visible-infrared image transform method. First, the region seeds growing (RSG) algorithm was used to extract the water target from the collected image. Subsequently, the visible and infrared image water target datasets were established. Finally, the trained cycle generative adversarial network (CycleGAN) was used to generate the infrared simulation image from the visible image. The test results show that the visual effect of the infrared simulation image generated by this method is close to that of the real infrared image and can be applied to the naval infrared military equipment simulation test and training system.
-
Key words:
- infrared image simulation /
- image style migration /
- CycleGAN /
- RSG
-
表 1 模型参数值
Table 1. Model parameters
Parameters Learning rate Epoch N_epochs N_epochs decay Beta1 Beta2 Decay iters Value 0.0002 200 100 100 0.5 0.999 50 -
[1] 罗海波, 史泽林. 红外成像制导技术发展现状与展望[J]. 红外与激光工程, 2009, 38(4): 565-573. https://www.cnki.com.cn/Article/CJFDTOTAL-HWYJ200904004.htmLUO H B, SHI Z L. Status and prospect of infrared imaging guidance technology[J]. Infrared and Laser Engineering, 2009, 38(4): 565-573. https://www.cnki.com.cn/Article/CJFDTOTAL-HWYJ200904004.htm [2] 董敏周. 用于红外成像系统抗干扰性能评估的红外图像仿真[D]. 西安: 西北工业大学, 2004.DONG M J. Infrared Image Simulation for Anti-Jamming Performance Evaluation of Infrared Imaging System[D]. Xi'an: Northwest University of Technology, 2004. [3] 张静元, 赵非玉, 吴刚刚, 等. 国内外红外对抗仿真及模型研究[J]. 光电技术应用, 2021, 36(2): 5-9. https://www.cnki.com.cn/Article/CJFDTOTAL-GDYG202102003.htmZHANG J Y, ZHAO F Y, WU G G. Research on Simulation and model of infrared countermeasure at home and abroad[J]. Electro-optic Technology Application, 2021, 36(2): 5-9. https://www.cnki.com.cn/Article/CJFDTOTAL-GDYG202102003.htm [4] ZHAO C, YANG H. A survey on development trends and key technologies of infrared guidance systems [J]. Electronics Optics & Control, 2008(5): 48-53. [5] 于伟杰, 刘华, 彭群生. 计算机红外成像仿真[J]. 计算机应用与软件, 2004(7): 1-2. https://www.cnki.com.cn/Article/CJFDTOTAL-JYRJ200407000.htmYU W J, LIU H, PENG Q S. Computer simulation of infrared image[J]. Computer Applications and Software, 2004(7): 1-2. https://www.cnki.com.cn/Article/CJFDTOTAL-JYRJ200407000.htm [6] 周方方, 司斌, 王晖. 红外图像仿真的发展综述[J]. 航空兵器, 2013(2): 29-37. https://www.cnki.com.cn/Article/CJFDTOTAL-HKBQ201302005.htmZHOU F F, SI B, WANG H. Review for simulation system of infrared image[J]. Aero Weaponry, 2013(2): 29-37. https://www.cnki.com.cn/Article/CJFDTOTAL-HKBQ201302005.htm [7] Goodfellow I, Pouget Abadie J, Mirza M, et al. Generative adversarial nets[J]. Advances in Neural Information Processing Systems, 2020(11): 139-144. [8] ZHU J Y, Park T, Isola P, et al. Unpaired image-to-image translation using cycle-consistent adversarial networks[C]//Proceedings of the IEEE International Conference on Computer Vision, 2017: 2223-2232. [9] Isola P, Zhu J Y, Zhou T, et al. Image-to-image translation with conditional adversarial networks[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2017: 1125-1134. [10] Mnih V, Kavukcuoglu K, Silver D, et al. Human-level control through deep reinforcement learning[J]. Nature, 2015, 518(7540): 529-533. [11] FAN J, ZENG G, Body M, et al. Seeded region growing: an extensive and comparative study[J]. Pattern Recognition Letters, 2005, 26(8): 1139-1156.