用小波变换进行水下回波边缘特征提取与分类识别

Transient feature extraction and discrimination of wideband echos based on wavlet transform

摘要: 提出一种运用小波变换提取宽带回波信号中的一种暂态特征-边缘特征以进行目标分类识别的方法.文中探讨了宽带回波的边缘与其所对应的目标特性之间的联系,提出运用离散二进小波变换提取出回波的多尺度边缘特征,并在此基础上构造了一个良好的特征空间.对实际采集的四种对应于不同湖底沉积物的宽带回波信号进行特征提取及分类识别,平均正确率可达95%以上.

Abstract: The performance of wavelet transform on extracting edge of underwater sonar echoes is investigated.The theoretical and experimental results show that WT has strong ability to characterize different edges with high noise immunity.A method to extract edge features of wideband echoes using discrete dyadic wavelet transform is proposed,and experimental results based on real data demonstrate the efficiency of the method.