摘要
近年来神经网络在图像分类上取得了成绩,然而军事图像具有数据量少、图像清晰度不高、军事目标与环境相似度较高等特点,导致传统的人工神经网络在军事图像数据集处理方面表现不佳,因此急需提高神经网络在军事图像分类方面的性能。结合自动搜索神经网络技术,提出了一种基于自动搜索神经网络技术的军事图像分类方法,并采用强化学习算法、参数共享和推进式搜索策略等思想,设计了神经网络结构搜索算法。试验结果表明,该方法在提高军事图像分类性能方面具有有效性和准确性。
In recent years,the neural networks have achieved excellent results in image classification.However,the military image has the characteristics of less data,low image resolution,and high similarity between military target and environment,which leads to the poor performance of traditional artificial neural network in military image data set processing.Therefore,it is urgent to improve the performance of neural network in military image classification.Combines with the automatic search neural network technology,a military image classification method based on automatic search neural network technology is proposed,and a neural network structure search algorithm is designed by using reinforcement learning algorithm,parameter sharing and propelling search strategy.Experimental results show that this method is effective and accurate in improving the performance of military image classification.
作者
周川
陈雷霆
陈雪地
ZHOU Chuan;CHEN Leiting;CHEN Xuedi(College of Computer Science&Engineering,University of Electronic Science&Technology of China,Chengdu 611731,China;Sichuan Provincial Key Laboratory of Digital Media,Chengdu 611731,China)
出处
《指挥信息系统与技术》
2021年第1期16-21,共6页
Command Information System and Technology
基金
装备发展部“十三五”装备预研课题资助项目。
关键词
军事图像分类
神经网络结构搜索
强化学习
military image classification
neural architecture search
reinforcement learning
