Geostatistical and Neural-Net Seafloor Classification at High Resolution andRelated Scaling Properties

摘要

The objective of my research under this grant is the development of anintelligent system for the automated classification of the seafloor using acoustic data. The work is a contribution to the Acoustical Reverberation Special Research Program (ARSRP). We have developed a method for surface classification, incorporating ideas from the theory of geostatistics (in short, the method has been called Geostatistical classification method). Under this project, we have finalized parameter selection and software development for the automated geostatistical seafloor classification method. Thereafter, we applied this method in a geomorphologic segmentation of the Western Flank of the Mid-Atlantic Ridge at 26 deg. North, the area of geophysical survey under the ARSRP in 1992. High-resolution bathymetric data from the 1993 geophysical experiment were also analyzed and compared to the (low resolution) HYDROSWEEP bathymetric data. Surface structures were found to be scale-dependent.