湖底沉积物分类的新方法研究

A New Practical Way for Classifying LakeBottom Sediment

从小波分析——多分辨率分解的观点出发，对不同的实测湖底沉积物的回波，用Ｄａｕｂｅｃｈｉｅｓ小波进行了Ｍａｌａｔ塔式分解，从中提取某些特征，用人工神经网络进行了分类，取得了对三类沉积物识别，平均正确识别率９０％以上，对五类沉积物识别，平均正确识别率８７％以上的较好结果。另外，对采取不同的分解特征进行分类的效果。

Classification of lakebottom and continentalshelf sediment is essential to certain applications, both civil and military. Such classification is the information indispensable for understanding shallowwater sound propagation and for shallowwater target identification. We utilized wavelet analysis to do a better job of lakebottom sediment classification. We obtained wideband sediment echo features, both in the time and frequency domain, with Mallat pyramid algorithm based on Daubechies type wavelet. Classification and recognition of experimentally measured lakebottom sediment echoes were done with: (1) an improved algorithm in the M.S. thesis of Huang Haining (H.N.Huang) , (2) several multilayer perceptrons (MLP) designed by us. A high recognition accuracy was obtained by the abovementioned practical way. For three kinds of sediment, it was above 90%; for five kinds, above 87%.