一种基于CNN的样本不足战场包围态势认知方法
发布时间：2018-12-15 02:14
【摘要】：为研究面对战场视图如何捕捉到指挥员认知经验的问题,深度学习中CNN可提供有力支持。但CNN的训练需要足够的样本数据,目前难以获得。针对战争中常见的战场包围态势认知及样本不足问题进行了剖析,提出一种基于CNN的样本不足包围态势认知新方法,该方法利用CNN的非线性拟合功能及包围态势图像的对称特性,可在一定程度上获得指挥员对包围态势的认知经验。仿真实验结果证明了方法的有效性和鲁棒性。
[Abstract]:In order to study how to capture the commander's cognitive experience in the battlefield view, CNN can provide support in depth learning. But the CNN training needs enough sample data, is difficult to obtain at present. In this paper, the common problems of battlefield encirclement situation cognition and lack of samples are analyzed, and a new method based on CNN is proposed. In this method, the nonlinear fitting function of CNN and the symmetry characteristics of the encircling situation image can be used to obtain the commander's cognitive experience on the encircling situation to a certain extent. Simulation results show that the method is effective and robust.
【作者单位】： 中国人民解放军国防大学信息作战与指挥训练教研部;中国人民解放军93682部队;中国人民解放军91053部队;空军工程大学防空反导学院;
【基金】：国家自然科学基金(61374179);国家自然科学基金青年科学基金(61703412) 军民共用重大研究计划联合基金(U1435218) 中国博士后科学基金(2016M602996)
【分类号】：E11;TP18
，
本文编号：2379766
[Abstract]:In order to study how to capture the commander's cognitive experience in the battlefield view, CNN can provide support in depth learning. But the CNN training needs enough sample data, is difficult to obtain at present. In this paper, the common problems of battlefield encirclement situation cognition and lack of samples are analyzed, and a new method based on CNN is proposed. In this method, the nonlinear fitting function of CNN and the symmetry characteristics of the encircling situation image can be used to obtain the commander's cognitive experience on the encircling situation to a certain extent. Simulation results show that the method is effective and robust.
【作者单位】： 中国人民解放军国防大学信息作战与指挥训练教研部;中国人民解放军93682部队;中国人民解放军91053部队;空军工程大学防空反导学院;
【基金】：国家自然科学基金(61374179);国家自然科学基金青年科学基金(61703412) 军民共用重大研究计划联合基金(U1435218) 中国博士后科学基金(2016M602996)
【分类号】：E11;TP18
，
本文编号：2379766
本文链接：https://www.wllwen.com/kejilunwen/zidonghuakongzhilunwen/2379766.html
