Geocoder: An Efficient Backscatter Map Constructor

The acoustic backscatter acquired by multibeam and sidescan sonars carries important information about the seafloor morphology and physical properties, providing valuable data to aid the difficult task of seafloor characterization, and important auxiliary information for a bathymetric survey. One necessary step towards this characterization is the assemblage of more consistent and more accurate mosaics of acoustic backscatter. For that, it is necessary to radiometrically correct the backscatter intensities registered by these sonars, to geometrically correct and position each acoustic sample in a projection coordinate system and to interpolate properly the intensity values into a final backscatter map. Geocoder is a software tool that implements the ideas discussed above. Initially, the original backscatter time series registered by the sonar is corrected for angle varying gains, for beam pattern and filtered for speckle removal. All samples of the time series are preserved during all the operations, ensuring that the full data resolution is used for the final mosaicking. The time series is then slant-range corrected based on a bathymetric model, in the case of sidescan, or based on beam bathymetry, in the case of the multibeam. Subsequently, each backscatter sample of the series is geocoded in a projected coordinate system in accordance to an interpolation scheme that resembles the acquisition geometry. An anti-aliasing algorithm is applied in parallel to the mosaicking procedure, which allows the assemblage of mosaics at any required resolution. Overlap among parallel lines is resolved by a priority table based on the distance of each sample from the ship track; a blending algorithm is applied to minimize the seams between overlapping lines. The final mosaic exhibits low noise, few artifacts, reduced seams between parallel acquisition lines and reduced clutter in the near-nadir region, while still preserving regional data continuity and local seafloor features. Fonseca and Calder, Geocoder: An Efficient Backscatter Map Constructor 2 ACOUSTIC BACKSCATTER AND GEOCODER STRUCTURE The acoustic backscatter acquired by multibeam and sidescan sonars carries important information about the seafloor geomorphology and physical properties. With the proper radiometric and geometric correction, acoustic backscatter mosaics can aid in the mapping of surficial seafloor features and facies, an important task toward remote seafloor characterization. Backscatter mosaics can also provide important auxiliary information not only for marine geological and environmental studies but also for hydrographic surveys. The acoustic backscatter registered by sidescan sonars is normally logged as two long time series of intensity values, one for the port side the other for the starboard side, recorded at the reception transducer (Tyce, 1986). On other hand, multibeam sonars register the acoustic backscatter in three different forms: 1) one measurement of average backscatter strength for each beam; 2) one time series of backscatter strength around the detection point of each received beam; 3) two long time series of backscatter strength (port and starboard) for each received ping, which will generate data very similar to a sidescan backscatter (Beaudoin, 2002). Geocoder is software tool conceived to accept all these different sources of acoustic backscatter and to construct more consistent and more accurate mosaics with the processed data. The data processing starts with the raw acquisition data, that is, the original data registered during the survey, without any additional processing. So far, the system can process Simrad, GSF and XTF formats. The implemented algorithm radiometrically corrects the backscatter intensities registered by sidescan and multibeam sonar, and then geometrically corrects and positions each acoustic sample in a final backscatter mosaic in a well-defined projection coordinate system. The Geocoder system is implemented with an interactive graphical user interface that allows the visualization of the navigation tracks and the backscatter mosaics, and is structured using object-oriented methods. The main objects of its data structure are the sonar lines, the sonar mosaics and the backscatter cells. The sonar objects have Fonseca and Calder, Geocoder: An Efficient Backscatter Map Constructor 3 information about the sonar equipment, the navigation, the transducer attitude, the gains and the sidescan backscatter samples. The sonar mosaic objects have information about projection, resolution and histograms of the final mosaic. A mosaic object is defined as an array of backscatter cells objects and not just pixels. One cell object can store up to two sidescan samples, each sample consisting of the backscatter value, the sample source (the acquisition line) and the sample quality. In order to define quality, samples closer to the nadir and far off nadir are attributed low quality values, while samples in the midrange are attributed higher values. The cell structure has an important function during the mosaicking procedure when multiple sidescan samples are mapped into the same mosaic cell. In Geocoder, instead of storing only the last mapped sample, or averaging all the samples inside one cell, the data structure stores the two most significant samples, with the two highest sample quality values. Finally, the mosaicking procedure is defined as the method that maps multiple sonar objects into one mosaic object. SLANT RANGE CORRECTION Slant-range distortions are inherent to backscatter acquisition geometry, and are a result of the echo return being registered in time and not in horizontal range to the transducer. Another related source of distortion is the water column data. Some sidescan systems start recording the backscatter time series just after the transmitting pulse, so that there is a period of time in which the echoes are coming from the water column and not from the seafloor. In Geocoder, the water column data is removed based on the depth of the first return and the knowledge of the sampling rate. In the case of backscatter time series from sidescan sonars, a typical slant-range correction is applied to the data (Miller et al. 1991). For that, a flat bottom is assumed and the transformation from slant range time samples to horizontal range distances is computed by simple geometry, depending only on the value of the sound speed. If a bathymetric model is available, the average slope in the across-track direction of each ping replaces the flat bottom assumption for the slant range correction. In the case of acoustic backscatter from multibeam sonars, the backscatter time series for each beam is added based on the time and range of each detection point, in Fonseca and Calder, Geocoder: An Efficient Backscatter Map Constructor 4 order to assemble a time series equivalent to a sidescan trace (Hughes-Clark et al. 1996). During the assemblage of this time series, if two samples arrive at the same time, the preference is given to the sample closest to a detection point. The final beam solutions of multibeam sonars include the detection time and the horizontal range to the transducer, so this information can be used to compute a more accurate slant-range correction. In Geocoder, multibeam backscatter time series can be slant-range corrected by parts, as the ranges for a certain number of samples (at the detection time of the beams) are known. The horizontal ranges for all samples are then computed by linear interpolation between consecutive beam solutions or by splines. This same procedure is used to slant-range correct sidescan datagrams registered by multibeam sonars, as the acquisition time of the sidescan samples can be synchronized to the detection times for the bathymetric beams. When processing average beam backscatter, the same rationale is used, with the limitation that the number of samples is restricted to the number of beams (one average backscatter per beam), and no interpolation is necessary. RADIOMETRIC CORRECTIONS AND SPECKLE REMOVAL In Geocoder, the processing sequence starts with the original acquisition data, so that all the logged parameters will be considered for the radiometric correction. Each raw backscatter sample is then corrected for the removal of variable acquisition gains, power levels and pulse widths, according to manufacturer’s specifications. The backscatter strength is calculated per unit of area and per unit of solid angle, so that the actual footprint area of the incident beam should be taken into account for proper radiometric reduction. During acquisition, logging systems normally simplify the geometry by assuming a flat bottom for the incident beams, which causes a radiometric distortion in the data. With this flat bottom assumption a simple Lambert’s law correction in normally applied during acquisition to reduce the angular dependency of the backscatter. In Geocoder, the backscatter values are corrected to the true footprint projection area, as the detailed bathymetry is known from the multibeam time-of-flight beam measurements. For each beam footprint, the along and across track slope are calculated with respect to a bathymetric model. The effective area of insonification is calculated Fonseca and Calder, Geocoder: An Efficient Backscatter Map Constructor 5 based on these slopes, the transmit and receive beamwidths, pulse length and range to the transducer. In a similar manner, the effective incident angle is calculated from the scalar product of the beam vector (form the transducer to the footprint) and the normal to the bathymetric surface at the footprint. The effective incident angle is used for correcting the Lambert’s law correction applied during acquisition. In the case of sidescan backscatter, due to the absence of beam bathymetry information, an external bathymetric model is used for the footprint area and slope corrections. Finally, a residual beam pattern correction is