欢迎访问《兵工学报》官方网站,今天是 分享到:

兵工学报 ›› 2021, Vol. 42 ›› Issue (9): 1987-1997.doi: 10.3969/j.issn.1000-1093.2021.09.019

• 论文 • 上一篇    下一篇

基于演化计算的迷彩目标隐蔽策略仿真设计

周颖1,2, 谢振平1,2, 蒋晓军3   

  1. (1.江南大学 人工智能与计算机学院, 江苏 无锡 214122; 2.江苏省媒体设计与软件技术重点实验室, 江苏 无锡 214122;3.近地面探测与感知技术国防科技重点实验室, 江苏 无锡 214035)
  • 上线日期:2021-10-20
  • 作者简介:周颖(1996—),女,硕士研究生。E-mail:1144606707@qq.com
  • 基金资助:
    国家自然科学基金项目(61872166);江苏省“六大人才高峰”项目(XYDXX-161)

Simulation Design of Camouflage Target Concealing Strategy Based on Evolutionary Computation

ZHOU Ying1,2, XIE Zhenping1,2, JIANG Xiaojun3   

  1. (1.School of Artificial Intelligence and Computer Science,Jiangnan University,Wuxi 214122,Jiangsu,China;2.Jiangsu Key Laboratory of Media Design and Software Technology,Wuxi 214122,Jiangsu,China;3.Science and Technology on Near-Surface Detection Laboratory,Wuxi 214035,Jiangsu,China)
  • Online:2021-10-20

摘要: 针对典型自然环境下军事装备的隐蔽问题,提出一种新颖的演化计算策略,将其与图像仿真设计手段相结合,根据环境状况,构建一种快速仿真计算进而给出迷彩目标隐蔽策略的新方法。该方法主要包括迷彩目标与场景图像的融合仿真计算,引入视错觉条带覆盖的迷彩目标隐蔽计算,基于粒子群优化算法和概率分布采样的搜索计算,以及基于深度神经网络图像特征的融合度计算。运用深度神经网络图像分割模型,结合迷彩目标的分割识出率,评估新方法的性能。仿真实验结果表明:在林地和荒漠环境中获得的隐蔽仿真图像平均融合度可达0.99以上,平均分割识出率低于0.90;新方法能够为设计给定目标在场景图像中的隐蔽策略提供有效的依据,具有较高的实用价值和可扩展性。

关键词: 迷彩目标, 隐蔽策略, 演化计算, 概率分布采样, 深度神经网络

Abstract: For the concealment problem of military equipment in typical natural environment, a novel simulation design method is proposed by combining a modified evolutionary computation strategy and image simulation design technique. The proposed method contains following parts: the fusion simulation computation of camouflage objects and scene images,the camouflage target concealment strategy simulation using optical illusion strips coverage,the search computation based on particle swarm optimization algorithm and probability distribution sampling algorithm,and the calculation of fusion degree between camouflage objects and background scenes based on deep neural network image features. Based on the deep neural network image segmentation model,the recognition rate of camouflage target segmentation is introduced to evaluate the performance of the proposed method. Simulation experimental result shows that the average fusion degree of concealment simulation images obtained in woodland and desert environment can reach above 0.99,and the average segmentation recognition rate of camouflage targets is lower than 0.90. The proposed method can provide very effective results for designing the concealment strategy of a given target in a scene image,and has good practical value and scalability.

Key words: camouflagetarget, concealingstrategy, evolutionarycomputation, probabilitydistributionsampling, deepneuralnetwork

中图分类号: