摘要
提出了一种利用BP神经网络用于空中战机目标识别的方法。首先,分别用300张不同姿态的F-16和F-22战斗机图片建立样本图库。其次,利用不变矩理论,提取图片的不变矩作为神经网络的输入量,分别采用基本梯度下降算法、有动量和自适应学习速率梯度下降算法和Levenberg-Marguardt优化算法训练BP网络。然后从样本图库中随机抽取两种型号飞机图片各30张作为空中打击目标进行识别,结果表明采用LM优化算法的BP网络具有一定的抗噪声能力。
A method of aerial targets recognition using BP neural network is presented in this paper.First of all,the sample storage is set up for training the neural network,which is made up of the 300 pictures of the F-16 and F-22 fighter.Secondly,moment invariant of the pictures is taken as the input of the neural network.At the same time,basic gradient descent algorithm,gradient descent with momentum and adaptive learning rate algorithm and Levenberg-Marguardt optimization algorithm are employed to train BP neural network.30 pictures of each fighter are chosen from the sample storage and recognized by BP network.Finally,noise testing is conducted.has certain anti-noise capability.
出处
《火力与指挥控制》
CSCD
北大核心
2012年第12期132-135,139,共5页
Fire Control & Command Control
基金
江苏省科技支撑基金资助项目(BE2010190)
关键词
神经网络
目标识别
BP算法
矩不变量
neural network
target recognition
BP algorithm
moment invariant
