Research On Classification Method Of Submarine Substrate Based On Echo Data

Seabed sediment is an important aspect of oceanographic research.Multi-beam detection is the main observation method in the field of seabed classification,providing massive data support for the study of seabed classification in national seas.Seabed classification is of great significance and practical research in areas such as seabed mapping and submarine concealment value.Nowadays,the inversion of the seabed environment has received extensive attention,and it is also a hot spot in the field of marine bottoms science research.Therefore,this paper selects the echo data from the Meizhou Bay in Fujian Province,which is rich in sediment types,and conducts in-depth research on the classification method of the ocean floor.Firstly,the background and significance of the classification of seabed sediments in the echo data of this subject are introduced,and the status quo of related studies on seabed sediments at home and abroad is analyzed.Combined with the relevant theories of multi-beam subsea detection,the multi-beam data recording method and the characteristics of the echo data are analyzed,and the echo data and sources used in this paper are briefly introduced.Secondly,for the extraction of feature values of echo data,the feature quantities of echo data are mainly extracted from three angles: time series,frequency domain sequence,and time-frequency domain to solve the problem of classification of single type feature quantities.In terms of time series feature quantity extraction angles,nine feature quantities based on time series waveform extraction are mainly studied.In frequency sequence feature quantity extraction angles,two types of subband energy feature extraction methods based on fast Fourier transform are mainly studied.In the time-frequency domain feature quantity lifting angle,the feature extraction method based on EMD decomposition and Hilbert transform is mainly studied.From these three perspectives,the feature quantities of 14 echo data were obtained and experimentally analyzed and evaluated,in order to select suitable echo features and prepare effective feature quantities for echo data feature fusion research.Thirdly,the feature quantities extracted from seabed bottom echo data are researched and analyzed from two aspects,the classic fusion algorithm and the optimized m RMR algorithm.The redundancy judgment criterion of feature quantity is introduced.The feature quantity is analyzed by redundant judgment theory.Experiments are performed to determine that the feature quantity is redundant.It is determined that a fusion algorithm is required to reduce the feature quantity to reduce the redundancy.Further,the theoretical analysis of feature quantity fusion algorithms including serial,PCA algorithm,Fisher algorithm and m RMR algorithm is proposed,and an optimized m RMR algorithm is proposed.The experimental analysis and evaluation of five feature fusion algorithms are performed.The fusion algorithm can better fuse the features.It has a higher application value for improving the fusion effect.Finally,three methods for classification of seabed sediments are introduced,including KNN algorithm,BP neural network algorithm and random forest algorithm,and the feasibility of the algorithm is analyzed.Experiments are performed on the fusion results of five types of feature data of the seabed bottom echo data,and the accuracy rates under the three classification methods are obtained.After a comparative analysis,the optimized m RMR fusion algorithm proposed in this paper is used to conduct random forests.The classification has obtained good results,which confirms that the theoretical analysis of seabed bottom echo data improved classification effect through feature fusion has practical application value.