Machine learning methods for acoustic-based automatic Posidonia meadows detection by means of unmanned marine vehicles

This work describes the exploitation of a Remotely Operated Vehicle (ROV), equipped with a multi-parametric sensors package (acoustic and video), for the exploration and characterisation of sea-bottoms covered with Posidonia oceanica seagrass, which represents a valuable indicator of the environmental health. The data collection is achieved by the employment of a single beam echosounder and a down-looking underwater camera. An acoustic data procedural analysis based on machine learning methods was developed to automatically detect the Posidonia presence, so that in future works it will be possible to operate also in low-visibility conditions, using only the acoustic sensors. Data acquisition was carried out over different seafloor types in coastal area near Biograd Na Moru (Croatia) and the preliminary results are reported in the paper.

[1]  Massimo Caccia,et al.  Seabed classification using a single beam echosounder , 2015, OCEANS 2015 - Genova.

[2]  G. Fader,et al.  An overview of seabed-mapping technologies in the context of marine habitat classification , 2000 .

[3]  Vladimir N. Vapnik,et al.  The Nature of Statistical Learning Theory , 2000, Statistics for Engineering and Information Science.

[4]  Gabriele Bruzzone,et al.  Development of an automatic sampler for extreme polar environments: first in situ application in Svalbard Islands , 2016, Rendiconti Lincei.

[5]  R. O'Neill,et al.  The value of the world's ecosystem services and natural capital , 1997, Nature.

[6]  E. Pouliquen,et al.  Time-evolution modeling of seafloor scatter. I. Concept , 1999 .

[7]  F. Bonin-Font,et al.  Towards Visual Detection, Mapping and Quantification of Posidonia Oceanica using a Lightweight AUV , 2016 .

[8]  B. Bett,et al.  Autonomous Underwater Vehicles (AUVs): Their past, present and future contributions to the advancement of marine geoscience , 2014 .

[9]  A. Atiya,et al.  Learning with Kernels: Support Vector Machines, Regularization, Optimization, and Beyond , 2005, IEEE Transactions on Neural Networks.

[10]  G. Bernard,et al.  Monitoring methods for Posidonia oceanica seagrass meadows in Prove n c e and the French Riviera , 2007 .

[11]  Marco Bibuli,et al.  Integrated Tele-Operation & Mission Control: preliminary experiments with a small USV , 2009 .

[12]  Colin Brown,et al.  Neural network classification of multibeam backscatter and bathymetry data from Stanton Bank (Area IV) , 2009 .

[13]  E. Pouliquen,et al.  Time-evolution modeling of seafloor scatter. II. Numerical and experimental evaluation , 1999 .

[14]  Markus Diesing,et al.  Mapping seabed sediments: Comparison of manual, geostatistical, object-based image analysis and machine learning approaches , 2014 .

[15]  Guillaume Bernard,et al.  Préservation et conservation des herbiers à Posidonia oceanica , 2006 .

[16]  L. Hamilton,et al.  Acoustic seabed segmentation for echosounders through direct statistical clustering of seabed echoes , 2011 .

[17]  Henrik Berg,et al.  A comparison of different machine learning algorithms for automatic classification of sonar targets , 2016, OCEANS 2016 MTS/IEEE Monterey.

[18]  Daniel Ierodiaconou,et al.  Evaluation of Four Supervised Learning Methods for Benthic Habitat Mapping Using Backscatter from Multi-Beam Sonar , 2012, Remote. Sens..