纺织学报 ›› 2024, Vol. 45 ›› Issue (05): 94-101.doi: 10.13475/j.fzxb.20230201501

• 染整工程 • 上一篇    下一篇

基于背景拼接与纹理模板的全自动迷彩图案设计

詹宇婷1, 梅琛楠1, 王焰2, 肖红2, 钟跃崎1,3()   

  1. 1.东华大学 纺织学院, 上海 201620
    2.军事科学院系统工程研究院, 北京 100010
    3.东华大学 纺织面料技术教育部重点实验室, 上海 201620
  • 收稿日期:2023-02-08 修回日期:2023-09-19 出版日期:2024-05-15 发布日期:2024-05-31
  • 通讯作者: 钟跃崎(1972—),男,教授,博士。主要研究方向为数字化纺织服装。E-mail:zhyq@dhu.edu.cn。
  • 作者简介:詹宇婷(1998—),女,硕士生。主要研究方向为迷彩图案的自动化设计技术。
  • 基金资助:
    上海市自然科学基金项目(21ZR1403000)

Fully automated camouflage pattern design based on background stitching and texture templates

ZHAN Yuting1, MEI Chennan1, WANG Yan2, XIAO Hong2, ZHONG Yueqi1,3()   

  1. 1. College of Textiles, Donghua University, Shanghai 201620, China
    2. Institute of Systems Engineering, Academy of Military Sciences, Beijing 100010, China
    3. Key Laboratory of Textile Science & Technology, Ministry of Education, Donghua University, Shanghai 201620, China
  • Received:2023-02-08 Revised:2023-09-19 Published:2024-05-15 Online:2024-05-31

摘要:

为提高迷彩伪装图案的设计效率和环境适应性,利用背景拼接与纹理模板融合进行迷彩伪装图案的全自动化设计。通过将输入的若干背景图像自动组合为单张拼接图像,采用均值聚类法提取出背景拼接图像的主色;提出4种迷彩纹理模板自动设计方法(多圆形随机分布法、WGN傅里叶频谱法、肌理图像生成法及分层云彩法),再对所得迷彩纹理模板进行色彩聚类与色彩替换,得到最终的迷彩伪装图案。为验证设计的有效性,针对丛林虚拟场景自动生成了4种伪装图案,并进行了伪装效果的主观和客观评价。结果表明,上述方法仅需17.0~71.0 s即可完成针对目标背景的迷彩图案设计,且最优方案为基于多圆形随机分布法纹理模板的迷彩图案;本文方法相对于普通迷彩伪装,其主观评价搜索时间增加了8.1%,基于Positioning and Focus Network(PF-Net)模型的客观评价发现概率降低了45%。

关键词: 迷彩伪装, 图案设计, 均值聚类, 纹理模板, 伪装检测

Abstract:

Objective In the military, the ability to reponse quickly to the battlefield environment can determine the success or failure of a battle. Therefore, It is need ed to improve the design speed of camouflage patterns and the ability to adapt to the target environment to enhance the camouflage effect of patterns and thus improve the military level of the army.

Method First, we automatically combined the background environment dataset into a stitching map and extracted the primary color of the background stitching map by mean clustering method; then, we proposed four automatic design methods for camouflage texture templates, including multi-circular random distribution method, WGN Fourier spectrum method, texture image generation method, and layered cloud method. Finally, after the mean clustering of camouflage texture templates, the camouflage pattern was obtained by replacing the template colors with the primary colors.

Results Experiments were conducted to construct virtual scenes and a dataset of target background images was created. We stitched the images in the dataset into a single large stitched image. The stitched image was detected based on a color histogram in RGB color space, and a cluster K value of 6 was determined. Six colors were extracted from the eight virtual jungle terrain scenes as primary colors, and four different camouflage patterns were generated based on a texture template method. The experimental time was short, and the design rate was high. In order to reduce the impact of environmental and human factors, camouflage assessment has been relatively carried out using templates based on multi-circular random distribution method, templates based on WGN Fourier spectrum method, templates based on texture image generation method, and templates based on layered cloud method as camouflage targets, and non-environment specific jungle camouflage (FLECKTARN-style jungle camouflage) as a reference. In the subjective evaluation experiments, the experimental results. Each test set had a 100% probability of discovery and different average search times, with the test set of camouflage patterns designed based on the multi-circular random distribution method having the longest average search time of 1.561 4 s. The average search times of all four camouflage patterns designed using this paper's method were greater than those of the reference pattern. Among them, the camouflage search time increased by 8.1% for the template based on the multi-circular random distribution method, 1.1% for the template based on the WGN Fourier spectroscopy method, 2.4% for the template based on the texture image generation method, and 3.7% for the template based on the layered cloud coloring method. In the objective evaluation experiments, the experimental results were shown. The average search time of the PF-Net network for the test set images was the same for both, 0.04 s, but the probability of discovery was different. The detection probability of the test sets designed according to the fully automated camouflage pattern design method based on background stitching and texture templates were both lower than that of the reference target, with the detection probability of the template camouflage based on the multi-circular random distribution method being 45% lower than the detection probability of the reference target.

Conclusion For automatic camouflage to achieve excellent results, it is necessary to react quickly to generate camouflage colors and textures according to changes in the target environment. This paper proposes a fully automatic camouflage pattern design method based on background splicing and texture templates with a high design rate, rapid response to the target environment, and good camouflage effect according to the modern army's demand for camouflage combined with computer technology. Through the subjective and objective evaluation of camouflage detection, the camouflage effect of the camouflage pattern is evaluated with the search time and detection probability as the primary indexes. The feasibility and effectiveness of the method are verified, and the generated pattern has a good camouflage effect. The camouflage design method in this paper provides various ideas for constructing camouflage design and achieves a more accurate and convenient camouflage pattern design for target area environments.

Key words: camouflage, pattern design, mean clustering, texture template, camouflage detection

中图分类号: 

  • TS941.2

图1

迷彩设计流程"

图2

多圆形随机分布法流程"

图3

WGN傅里叶频谱法流程"

图4

肌理图像生成法流程"

图5

分层云彩法流程"

图6

目标背景图像示例"

图7

不同k值下颜色直方图与原图的相似度"

图8

4种方法自动生成的迷彩图案"

图9

非针对目标环境的丛林迷彩"

图10

士兵模型"

图11

测试集图像示例"

图12

干扰图像示例"

表1

测试集图像主观评价"

实验测试集图像 发现概率/% 平均搜索时间/s
MC-RD 100 1.561 4
WGN-FFT 100 1.460 4
TIG 100 1.478 2
LC 100 1.497 3
非针对目标环境的丛林迷彩 100 1.443 5

图13

结果显著性图示例"

表2

测试集图像PF-Net客观评价"

实验测试
集图像
发现概率/% 平均搜索
时间/s
MC-RD 40 0.04
WGN-FFT 45 0.04
TIG 50 0.04
LC 55 0.04
非针对目标环境的丛林迷彩 85 0.04
[1] 蔡云骧, 薛士强, 周志勇, 等. 基于分形布朗模型的数码迷彩图案生成方法研究[J]. 兵工学报, 2016, 37(1):186-192.
doi: 10.3969/j.issn.1000-1093.2016.01.026
CAI Yunchang, XUE Shiqiang, ZHOU Zhiyong, et al. Research on digital camouflage pattern generation method based on fractal Brownian model[J]. Journal of Military Engineering, 2016, 37(1):186-192.
[2] 徐英. 基于背景代表色提取的迷彩伪装颜色选取算法[J]. 光电工程, 2007(1):100-103,144.
XU Ying. Camouflage color selection algorithm based on background representative color extraction[J]. Optoelectronic Engineering, 2007(1):100-103,144.
[3] CHU M. Research on color feature extraction method of digital camouflage based on visual perception[C]// 2021 6th International Conference on Communication, Image and Signal Processing (CCISP) Chengdu: IEEE, 2021: 188-192.
[4] QIN L, DUAN Z, HU J, et al. Background self-adapted digital camouflage pattern generation algorithm for fixed target[M]. International Conference on Image and Graphics. Cham: Springer, 2017: 545-551.
[5] XUE F, XU S, LUO Y T, et al. Design of digital camouflage by recursive overlapping of pattern temp-lates[J]. Neurocomputing, 2016, 172: 262-270.
[6] 喻钧, 杨武侠, 胡志毅, 等. 数码迷彩的生成法研究[J]. 光电工程, 2010(11):110-114.
YU Jun, YANG Wuxia, HU ZhiYi, et al Research on the generation algorithm of digital camouflage[J]. Optoelectronic Engineering, 2010(11):110-114.
[7] 周颖, 谢振平, 蒋晓军. 基于演化计算的迷彩目标隐蔽策略仿真设计[J]. 兵工学报, 2021, 42(9):1987-1997.
doi: 10.3969/j.issn.1000-1093.2021.09.019
ZHOU Ying, XIE Zhenping, JIANG Xiaojun. Simulation design of camouflage target concealment strategy based on evolutionary computation[J]. Journal of Military Engineering, 2021, 42(9):1987-1997.
[8] GAN Y, LIU C, HE Z, et al. Digital camouflage pattern design based on the biased random walk[J]. Modelling and Simulation in Engineering, 2022. DOI: 10.1155/2022/2986346.
[9] LOESDAU M, CHABRIER S, GABILLON A. Hue and saturation in the RGB color space[C]// International Conference on Image and Signal Processing. [S.L.]: Springer, 2014: 203-212.
[10] YU J, CHU M, TIANG S, et al. Comparison and analysis for image clustering methods of camouflage design[J]. Journal of Applied Sciences-Electronics and Information Engineering, 2009, 27(5): 475-479.
[11] ZENG M, LI J, PENG Z. The design of top-hat morphological filter and application to infrared target detection[J]. Infrared Physics & Technology, 2006, 48(1): 67-76.
[12] BOATO G, DANG-NGUYEN D T, DE NATALE F G B. Morphological filter detector for image forensics applications[J]. Ieee Access, 2020, 8: 13549-13560.
[13] HASKELL J P, RITCHIE M E, OLFF H. Fractal geometry predicts varying body size scaling relationships for mammal and bird home ranges[J]. Nature, 2002, 418(6897): 527-530.
[14] BRUNYÉ T T, EDDY M D, CAIN M S, et al. Masked primingfor the comparative evaluation of camouflage conspicuity[J]. Applied Ergonomics, 2017, 62: 259-267.
[15] DICKINSON S J, CHRISTENSEN H I, TSOTSOS J K, et al. Active object recognition integrating attention and viewpoint control[J]. Computer Vision and Image Understanding, 1997, 67(3): 239-260.
[1] 池盼盼, 梅琛楠, 王焰, 肖红, 钟跃崎. 基于边缘填充的单兵迷彩伪装小目标检测[J]. 纺织学报, 2024, 45(01): 112-119.
[2] 许晗, 沈雷, 陈涵. 居里叶分形在艾德莱斯裙装图案设计中的应用[J]. 纺织学报, 2023, 44(06): 191-199.
[3] 邓椿山 李琴 周莉 张龙琳. 体型分析在观测服装号型适应性上的应用[J]. 纺织学报, 2017, 38(01): 111-115.
[4] 张瑜 李群华 李智勇. 感温变色材料在纺织品图案设计中的应用[J]. 纺织学报, 2016, 37(4): 49-53.
[5] 杨韧 付少海 谭莹 梁惠娥. 中老年女装图案设计及其发展趋势[J]. 纺织学报, 2016, 37(10): 107-0112.
[6] 曹叶青 钱晓农 张宁 张苕珂. 淮阳泥泥狗艺术形式的男衬衫图案设计中的应用[J]. 纺织学报, 2016, 37(09): 111-116.
[7] 罗戎蕾 洪潘. 基于数学方法的流行几何图案设计[J]. 纺织学报, 2014, 35(3): 141-0.
[8] 鲍小龙 刘月蕊. 基于数码技术的蓝印花布图案创新设计研究[J]. 纺织学报, 2013, 34(5): 100-106.
[9] 吴巧英 冯婉婉. 长江下游地区学龄前女童体型的分类[J]. 纺织学报, 2013, 34(4): 111-116.
[10] 姜会钰;郑春玲;刘素一. 基于分形理论中Mandelbrot集变换的印花图案设计[J]. 纺织学报, 2010, 31(12): 139-142.
[11] 黄鹤玲;李霆;易积政;倪兵兵. 基于颜色空间与聚类分析相结合的袜带辨色分级方法[J]. 纺织学报, 2010, 31(11): 99-103.
[12] 张春明;房宽峻;贡利华;赵珍玉;吴菡. 数字喷墨印花对服装图案设计的革新[J]. 纺织学报, 2008, 29(5): 75-79.
[13] 张聿;李玲;付岳莹. 基于IFS随机迭代算法的印花图案设计[J]. 纺织学报, 2008, 29(2): 106-109.
[14] 潘如如;高卫东;张星烨. 基于动态聚类的织物疵点识别算法[J]. 纺织学报, 2008, 29(10): 122-126.
[15] 万永菁;万光逵. 图像分割算法在花型自动识别中的应用[J]. 纺织学报, 2007, 28(5): 63-65.
Viewed
Full text


Abstract

Cited

  Shared   
  Discussed   
No Suggested Reading articles found!