Cobalt Crusts In The Nonlinear Ultrasonic Identification Technology

Along with the rise of ocean strategy, the world's major developed countries gradually focused their attention on marine mineral resources. Among them, the cobalt crusts in deep sea as a kind of marine strategic resource which is of the value of commercial exploitation has been drawn focuses on by all the coutries. Currently, the world's major developed countries have begun the research on exploration and mining of cobalt crusts. To speed up our country's pace on the exploration and mining of cobalt crusts, somen relative technical research on the underwater recongnitino on cobalt crusts is tried to do in this paper under the support from the National Natural Science Foundation project "Study on Extraction Mechanism of Collecting Cobalt Crusts Robot and Optimal Design of Extractive Institutions".Firstly, the two representative sediment echo's feature extraction methods in wavelet domain:envelop feature of echoed tail wave and modulus maxima feature of echo are applied to echo feature extraction of cobalt crusts. By comparative experiment, it is indicated that because of the influence of surface topography, the distribution of these two feature samples in wavelet domain degrade and the effects of linear dimensionality reduction and linear classification results are poor by constrast with the sediments of smooth surface. To improve the linear classification results, two nonlinear methods based on kernel space: KFDA (Kernel Fisher Discriminant Analysis) nonlinear dimensionality reduction method and LSSVM (Least Square Support Vector Machine) nonlinear classification method are introduced on the application of the recognition of cobalt crusts. The experimental results show that using KFDA method nonlinear discriminant features can be extracted effectively and the samples'distribution in the reduced dimensional feature space can be improved, application LSSVM seafloor classification method can improve the results, using LSSVM method the seafloor classification method results can be improved. In order to further improve the classification results, a kind of nonlinear kernel features fusion method named as KECCA(Kernel ECCA) is proposed based on CCA(Canonical Correlation Analysis) and ECCA(Enhanced CCA) linear features fusion methods and a kind of nonlinear features fusion and recognition model "KECCA+PLS" is proposed in this paper. Experimental results demonstrate that the classification results can be further improved by using the proposed model. Finally, a comprehensive experiment is done on 19 kinds of major matter in deposits of cobalt crusts using "KECCA+PLS" model. In our experiment, the average correct recognition rate of cobalt crusts is 90.2%, the average error recognition of cobalt crusts is only 3.1%, using "KECCA+PLS" model obtain a good recognition effect.The research of this dissertation provides not only the theoretical base on the recognition of cobalt crusts, but also the good theoretical reference on the recognition of other marine mineral resources and provides the effective technical support to China's deep-sea mining.