摘要
针对SAR图像的自动目标识别问题,研究了基于小波分析和神经网络的识别算法。由非线性小波基作为网络中神经元的激励函数,隐层结点数由小波分解次数和处理目标类别数决定,输出层由目标的类别数决定,同时利用目标的方位角来限定被识别目标的范围。实验结果表明,该方法有效降低了训练和识别的难度,取得了优于BP网络的识别结果,具有广阔的应用前景。
A method based on wavelet analysis and neural network is presented,to the question of SAR image automatic target recognition.The wavelet function is used as the active function in the network.The number of hidden units is determined by the wavelet decomposition and the dimension of input signal.The output dimension is determined by the classes of the targets.The pose of the target is used for restrict the scope of the identified targets.Experiment results show that the property of the wavelet network is better than that of BP network in computational cost and identify rate.
出处
《测控技术》
CSCD
2005年第7期14-16,23,共4页
Measurement & Control Technology
基金
国防预研课题
关键词
合成孔径雷达
自动目标识别
小波神经网
方位角估计
synthetic aperture radar(SAR)
automatic target recognition(ATR)
wavelet neural network(WNN)
pose estimation
