A Study On Classification And Recognition Of Surficial Shallow-Water Sediments Using Acoustical Methods

Most parts of ocean bottoms cover a layer of soft sediment, whose properties are full of important signification to military and civil applications. Therefore, the study on the sediment types of continental shelf and deep oceans is always a classical and interesting issue, and it is also an important content of the new survey to our country's territorial resources.The conventional direct sampling method can obtain the precise information of types of ocean sediments, but it has some shortcoming as follows: l.It's inefficient and uneconomical. 2. It uses the discrete sampling that cannot acquire continuous data. 3. The samples should be easily disturbed so that it's hard to get the accuracy result. Therefore it is being replaced by the indirect methods such as acoustic technique. Acoustic remote sensing is an efficient and economic technique, which can get the data continuously and sufficiently. Combining direct samplings, it offers a rapid and stable method to the classification and recognition of ocean sediments.In this thesis, a study on the classification and recognition of surficial shallow-water sediments off central & northern Yellow Sea by statistical feature extraction of echo signals is made, dealing with the data of experiment in this area. Two new techniques are proposed, one of them is based on the phase plane of acoustic energy flux density, and another is based on the STFT (short time Fourier transform) and SVD (singular value decomposition). The design of pattern classifier and recognizer is also performed, and an adapted classifier based on closest distance is presented, with the training using echo signal samples from 5 positions. The analysis of the recognition result is also discussed. On the same time, a time-domain model of surface scattering at normal incidence is offered and the influence of surface roughness backscattering on signals from different sediment types is analyzed through model/data comparison. The results are summarized as fallows:1) It's found that the characteristic quantities extracted are clear in physics and consistent with the analysis of the sediment samples. Through cluster analysis, it turns out that they are better separable than the quantities extracted using former techniques.2) The types of sediments are recognized using the pattern classifier with testsamples. The recognition is compared with result obtained from traditional methods. It proves that this pattern classifier is effective for the classification and recognition of three types of marine sediments.3) The comparison of surface scattering model with experiment data is discussed, which shows the significant influence on echo signals from sandy sediments and the subordinate influence on those from silty sediments. So this study can also been used to separate the sandy sediments and the silty sediments.