The use of predictors in seabed mapping: A simulation approach Ghedhban Swadi a , Dave Holifield a,* , Ivan Jordanov b a Department of Product and Engineering Design, University of Wales Institute, Cardiff, Western Avenue, Llandaff, Cardiff CF5 2YB, UK b Computing School, University of Portsmouth, UK article info Article history: Received 11 May 2010 Received in revised form 24 November 2010 Accepted 24 February 2011 Available online 11 March 2011 Keywords: Seabed mapping Sonar systems Echo sounder Neural network K-Nearest-Neighbour Prediction abstract This paper compares two methods of prediction applied to seabed mapping; the K-Nearest- Neighbour (KNN) and the Adaptive Linear Neural Network (ADALINE). In order to study the performance of these predictors, a simulated sonar system platform was developed. The platform includes a seabed simulator based on fractal geometry, and an echo sounder whose outcome is the measured depth of the seabed. Matlab was used to build the simu- lator and to assess the performance of the predictors. The results show the dynamic ADA- LINE gives a better performance than KNN. Ó 2011 Elsevier Ltd. All rights reserved. 1. Introduction Seabed mapping is one of the most active domains of modern underwater acoustics, attracting the attention and exploration in many branches of science and technol- ogy. The seabed is a platform for engineering structures, a source of raw materials, a repository for unwanted materi- als, route of communications, a potential background, a laboratory where physical, chemical and biological processes of great importance can be studied and many more things besides. The science related to the measuring of sea depth is bathymetry. The production of good quality, accurate bathymetric maps are essential for ensuring safe navigation at sea, both commercial and leisure. Such charts were made before the advent of the acoustic techniques, but they have greatly gained from them, as much as in accuracy as in completeness. Nowadays, the seabed general topography can be im- aged or detected by several different techniques, each with their own merits. Sonar systems which are based on acoustical methods include Single Beam Echo Sounder (SBES), Multi-Beam Echo Sounder (MBES), Side Scan Sonar (SSS), Synthetic Aperture Sonar (SAS) and Interferometric Synthetic Aperture Sonar (InSAS). There are some other technologies capable of mapping small areas of the seabed dedicated for the study of microstructure and composi- tion of the seabed such as benthic corers, grabs and probes. These devices are used to ground-truth acoustic data [10]. Introducing prediction techniques to enhance the qual- ity of the measured bathymetric data will help in filtering the measured data and excluding noise from actual data to ensure higher seabed mapping accuracy and more secure navigation. 2. Echo sounders The simulated sonar platform adopted in this work is based on echo sounder. Echo sounders are the simplest and most used of all sonar equipments. They are used to measure the depth of water below the sonar platform as shown in Fig. 1. A typical operating frequency of an echo sounder work- ing in shallow water is 40 kHz. In deep water, as in oceans, 0263-2241/$ - see front matter Ó 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.measurement.2011.02.020 ⇑ Corresponding author. Tel.: +44 02920416709. E-mail address: dmholifield@uwic.ac.uk (D. Holifield). Measurement 44 (2011) 1038–1043 Contents lists available at ScienceDirect Measurement journal homepage: www.elsevier.com/locate/measurement