Study On Methods Of Seafloor Geomorphology Identification

Seafloor geomorphology can reveal sedimentary dynamics within the seafloor,and can also provide detailed seafloor geomorphology information for ocean engineering operations.Seafloor geomorphology types are complex.At present,the seafloor geomorphology type recognition mainly depends on manual classification.This method is not only subjective,but also time-consuming.This dissertation is based on previous researches.Based on the terrain macro features and multifractal theory,the features of geomorphology units are extracted.Factor analysis method is used to estimate common factors.Based on support vector machine and fuzzy pattern recognition techniques,an automatic identification method of seafloor geomorphology type is studied.The main research contents are as follows:(1)For platform,gully,landslide,uplift and channel,six terrain macro features are extracted based on multi-beam data.The extracted features are kurtosis,skewness,depth difference entropy,terrain roughness,depth variation coefficient and depth standard deviation.Geomorphological fractal and multifractal analysis results show that gully,landslide and channel have multifractal structure.The multifractal features of these three geomorphology units are extracted,including multifractal spectrum width,multifractal spectrum peak,fractal dimension difference of the maximum and minimum probability subsets,and fractal dimension.The results show that the above features of the experimental seafloor geomorphology types in this studies have differences.So these features can be used to identify seafloor geomorphology type.(2)Based on the factor analysis method,common factors that characterize the geomorphology types are obtained.The research indicates that the terrain macro features and multifractal features can be merged into 3 and 2 common factors,respectively.The cumulative variance contribution rate of these common factors reaches 70%.(3)Based on the principle of support vector machine,seafloor geomorphology classification is implemented by using terrain macro features,terrain multifractal features and their common factors.The RBF kernel function is selected as kernel function,and Grid Search method is used to optimize the related parameters.The experimental results show that based on terrain macro features and multifractal features,seafloor geomorphology recognition accuracy rate can reach 70%;based on the common factors,the recognition accuracy rate can be increased to 95%.(4)Based on maximum membership principle of fuzzy pattern recognition,seafloor geomorphology type is classified by using terrain macro features and their common factors.In this dissertation,Gaussian function is selected as the membership function.The recognition accuracy rate is 81.48%.