Mapping seabed features from multibeam echosounder data using autocorrelation and multi-scale wavelet analyses

MAPPING SEABED FEATURES FROM MULTIBEAM ECHOSOUNDER DATA USING AUTOCORRELATION AND MULTISCALE WAVELET ANALYSES JAROSàAW TĉGOWSKI1, JAROSàAW NOWAK1, BENEDYKT HAC1, MATEUSZ ZAMARYKA2, KAZIMIERZ SZEFLER1 1 Maritime Institute in GdaĔsk Dáugi Targ 41/42, 80-830 GdaĔsk, Poland 2 Institute of Oceanography, University of Gdansk, al. Marszalka Pilsudskiego 46, 81-378 Gdynia, Poland In the paper we propose the method of seabed morphological features extraction, which we have obtained from bathymetric and backscatter data, recorded by multibeam echosounder. Presented results of acoustical recognition of the southern Baltic Sea bottom are the part of measurements conducted in the band of 220 km length in the central part of Polish coastal water. The detailed analysis of seabed features were performed for area located in the vicinity of Koáobrzeg harbour. The degree of seafloor corrugation was determined by autocorrelation analysis of seafloor bathymetry. To which, we used estimation of autocorrelation length and fractal dimension, based on the shape of autocorrelation function. Moreover, the parameters of wavelet decomposition of bottom backscattering strength were the input to fuzzy logic clustering system allowing for outline of seafloor areas of similar morphological features. Both presented methods have confirmed its effectiveness in identifying morphological characteristics and types of the bottom surface. INTRODUCTION The use of multibeam echosounders (MBES) for bottom recognition allows for investigation of detailed seafloor geomorphologic features of extensive seafloor areas. The idea of MBES seafloor classification based on angular dependency of the backscattered intensity is known in several classification systems [e.g. 1,2,3,4]. Other, but not so popular methods utilised for research of seabed morphology (slope and roughness) are based mostly on texture analysis [e.g. 4,5]. The authors of this paper made attempt to both mentioned analyses, where the first was based on parameterization shape of bathymetric 253 transect (wavelet, statistical and fractal parameters) in sliding windows and used parameters as the input to fuzzy logic clustering system [4]. The second system was based on parameterisation of shape of angular dependencies of backscattered signal intensities. Computed parameters were the input to the segmentation system [4]. Positive results of both mentioned procedures allowed for classification of seabed types in the Rowy polygon (southern Baltic Sea area). Need to identify the morphological characteristics of the bottom is important in mapping of the bottom habitats and description of morphological processes, which determine the shape of the bottom surface (eg, direction and speed of bottom currents). The main idea of this paper is to find the method of fast and complex description of scales of bottom corrugations. To perform this task we examined correlation features of seabed surface, measuring the autocorrelation radii and fractal dimensions based on the autocorrelation function. Moreover, to the spatial distribution of bottom backscattering strength (BBS) we applied a segmentation algorithm based on BBS spatial wavelet decomposition and fuzzylogic clustering algorithm. Results of seafloor segmentation and classification for both algorithms gave a good correspondence with types of the bottom and its sediments. 1. AREA AND METHOD OF ACOUSTICAL MEASUREMENT The acoustical measurements of seafloor of the Polish Exclusive Economic Zone within the Baltic Sea were conducted on board of the rv Imor from Maritime Institute in GdaĔsk. The research vessel was well equipped with instruments for seafloor research as eg. MBES, side scan sonars, subbotom profilers, cameras mounted on the ROV, sediment core and grab samplers and precise GPS navigation systems. Special interest was focused on the narrow euphotic zone of the depth up to 20m elongated parallel to the Polish cost and containing Fig.1. MBES bathymetry of investigated area - polygon located in the vicinity of the Koáobrzeg harbour 254 different morphological forms of the bottom. The total length of the surveyed area was about 220 km and of a width approximately 1 km. But in some particular areas, as e.g. vicinity of Koáobrzeg harbour, the measurements covered areas of larger distance than 1km from the coast (see Fig.1.). The bathymetry of surveyed area measured using MBES at a water depth of not less than 4 m and other techniques in surf zone (single beam echosounder and geodesic measurements) is presented in Fig.1. From the huge set of registered acoustical data we concentrated on MBSE bathymetric and BBS data. The measurements were performed with multibeam echosounder Reson 8125 with working frequency of 455 kHz, range - 0,5m - 120m, no. of beams - 248, scan width - 120º and beam width - 0.5º. Due to spatial spread of investigated area, huge volume of registered bathymetric data and possibilities, as well as limitations of used computers, the spatial resolution of bathymetric 3D map was chosen on 2m by 2m, which is sufficient for investigated scales of seabed corrugations. 2. AUTOCORRELATION AND FRACTAL ANALYSIS OF SEAFLOOR CORRUGATION The corrugation of rough surface is usually described by statistical, spectral or fractal techniques [6]. In our approach we combined statistical and fractal descriptions of bottom roughness. Efficient and reliable indicator of surface shape is the correlation length cr, which & is defined as distance over which the autocorrelation function C r falls by 1/e. The other roughness parameter used in this analysis is fractal dimension D of the surface calculated on the base of slope of autocorrelation function. The seafloor bathymetry map (matrix) was divided for squares of side lengths of 25 or 50 meters. For each isolated square of seafloor surface (see Fig.2.a.) was computed the autocorrelation function (Fig.2.b). The most of calculated autocorrelation functions of square areas are not isotropic, which confirms anisotropy of seabed undulations. For further analysis, & we estimated the autocorrelation functions C r and autocorrelation lengths cr in x horizontal and y - vertical directions for each isolated seafloor square (Fig. 2c). Many shapes and forms in nature satisfy assumptions of the fractal geometry, that is, there are the self-similar at different scales. It has been shown that the landscape is a fractal surface [7, 8, 9]. The measure of object fractality is fractal dimension, so-called Hausdorff dimension [7, 10]. The Hausdorff dimension of a subset X of Euclidean space is defined as a limit: D  log N ( r ) , r o0 log r lim (1) where N(r) denotes the smallest number of open balls of radius r needed to cover subset X; an open ball B(p, r) = {x: dist(x, p) < r}, where dist(x, p) is the distance between points x and p. It is practically impossible to measure fractal dimension using the above definition (1), and it is the reason for use of equivalent methods. (...truncated)