地面运动目标分类的模式特征与评价
|
摘要
为了进一步对车辆目标分类 ,对实验获得的典型地面运动目标—轮式车、履带式车的地震动信号从频域、时—频域等多方面进行特征提取。在频域上 ,应用傅立叶变换、经典功率谱分析等常用的信号处理方法对信号进行处理 ,提取了信号的 FFT特征和功率谱特征。在时 -频域应用短时傅立叶变换、小波及小波包分析方法对信号进行处理 ,得到时频分布矩阵奇异值分布特征和小波包分解能量分布特征。之后基于距离可分性设计了一个模式特征可分性测度 ,对时域和时—频域所提取的各种特征进行对比评价 ,结果表明 FFT特征、功率谱特征和小波分解后的能量特征具有更好的可分性。该结果与将各特征应用神经网络进行目标识别的结果是一致的。这表明所设计的模式特征可分性测度是有效的。
To classify the wheeled vehicle from tracked vehicle, the seismic signals of two kinds of vehicle are analyzed in different ways in the paper. In frequency domain, the methods of FFT and classical PSD are used to process the signals, and the features of FFT and PSD are obtained. In time-frequency domain, the signals are processed by the methods of the STFT, the wavelet and wavelet package, and the feature of SVD and the feature of the energy spectrum are obtained, After then, a separable measure is proposed based on the space, and it is applied into the evaluation of the above features, the result of the evaluation shows that the features the of FFT, PSD, and the energy spectrum all have the better separation ability, and this result is validated by the result of the target identification by neural network. Accordingly, we make sure that the proposed separable measure is effective.
To classify the wheeled vehicle from tracked vehicle, the seismic signals of two kinds of vehicle are analyzed in different ways in the paper. In frequency domain, the methods of FFT and classical PSD are used to process the signals, and the features of FFT and PSD are obtained. In time-frequency domain, the signals are processed by the methods of the STFT, the wavelet and wavelet package, and the feature of SVD and the feature of the energy spectrum are obtained, After then, a separable measure is proposed based on the space, and it is applied into the evaluation of the above features, the result of the evaluation shows that the features the of FFT, PSD, and the energy spectrum all have the better separation ability, and this result is validated by the result of the target identification by neural network. Accordingly, we make sure that the proposed separable measure is effective.
引文
[1] 凯依S M.现代谱估计—原理与应用[M].北京:科学出版社,1994.
[2] 李建平.小波分析及信号处理理论、应用及软件实现[M].重庆:重庆出版社,1997.
[3] RichardA Gramann,MaryBethBennett,ThomasD.O'Brien.Vehicle andPersonnelDetectionUsingSeismicSensors.Partof theSPIE Conference onSensors,C3I,Information,andTrainingTechnologies forLawEnforce-ment.Boston,Massachusetts[J].1998,SPIE 3577:73-85.
[4] HlawatschF,UrbankeR L.Bilinear time-frequency representation of signal:the shift-scale invariantclass[J].IEEE Trans onSignalProcessing,1994,42(2):356-366.
[2] 李建平.小波分析及信号处理理论、应用及软件实现[M].重庆:重庆出版社,1997.
[3] RichardA Gramann,MaryBethBennett,ThomasD.O'Brien.Vehicle andPersonnelDetectionUsingSeismicSensors.Partof theSPIE Conference onSensors,C3I,Information,andTrainingTechnologies forLawEnforce-ment.Boston,Massachusetts[J].1998,SPIE 3577:73-85.
[4] HlawatschF,UrbankeR L.Bilinear time-frequency representation of signal:the shift-scale invariantclass[J].IEEE Trans onSignalProcessing,1994,42(2):356-366.
版权所有：© 2023 中国地质图书馆 中国地质调查局地学文献中心