In the 1980's and prior to accepting data from the General Instruments Bathymetric Swath Survey System (BS³) into its hydrographic survey database, NOAA conducted extensive characterization tests to determine the accuracy and precision of the BS³, as installed on the NOAA Ship DAVIDSON (Pryor 1982). During those BS³ characterization tests, basic techniques, later termed "Patch Test" (Wheaton 1988), were developed which employed internal consistency of processed bathymetric data, acquired over a common area, along several tracklines with different ship headings and speeds. A patch test is used to estimate systematic adjustments for a particular system installation to improve the accuracy of heading, roll, and pitch sensor values, as well as, any adjustment to navigation time tags that will reduce timing errors between navigation and sonar data. The objective is to determine the set of parameter adjustments which produces minimum variability of the processed data for the common area. The rationale for the patch test is that a major source of variability between several renderings of bathymetry in a common area is bias in one or more of the data sources employed in processing the sonar's slant range acoustic measurements into an X, Y, Z representation of the bathymetry. With slight variations in the prescribed pattern of tracklines and the method of intercomparisons between bathymetry from the several tracklines (Crews 1990), this philosophy is the mainstay behind the patch tests which NOAA regularly performs on its multibeam depth sounders.

The Hydrographic Ground Truth Experiment (HYGRO) series (Mayer, 1993), like the BS³ tests, employed an independent source of depth data, external to the sonars being tested, to provide insight into the accuracy and precision of sonar systems (Huff 1994, Du 1994). HYGRO also led to the development of patch test procedures utilizing point features with a pronounced, localized bathymetric expression (Godin 1996a) to estimate offsets in heading and pitch sensors, as well as, the offset between navigation and bathymetry time tags. However in using point features in the bathymetry, a new complication was introduced into the interpretation of patch tests. The footprint size or spatial resolution capability of the sonar impacts the ability to accurately position a bathymetric point feature using the beam by beam bathymetric data. Consequently, it may not be possible to ascribe apparent shifts in the positions of any particular bathymetric point feature as being the direct result of one or more of the system offsets.

Multibeam bathymetric sonars that provide the signal amplitudes observed in conjunction with each beam's detection of the bottom in slant range (beam by beam imagery) can provide insight concerning spatial variations in backscatter properties of the bottom. It is conceivable that a localized area of the bottom, whose backscatter properties are distinctly different from the backscatter properties of the surrounding bottom, could serve a similar function as a point bathymetric feature in assessing system offsets. However, the spatial resolution of beam by beam imagery is constrained to be the same as the spatial resolution of the bathymetry. Therefore, in a like manner, it may not be possible to ascribe apparent shifts in the position of a particular feature in the beam by beam imagery to one or more of the system offsets.

Some multibeam bathymetric systems provide signal amplitude information, which is independent of vertical (off-nadir) angle, in addition to the beam by beam imagery. Bottom returns from the acoustic pulses transmitted by the sonar for measuring slant ranges to the bottom and beam by beam imagery are also employed in measuring the angle-independent imagery. The difference between the beam by beam imagery and the angle-independent imagery is simple. In the former, the time series of acoustic energy received following any particular transmit pulse is segmented into multiple short time intervals, one for each of the sonar's multiple, narrow off-nadir beams. The result is one amplitude per beam, per pulse with a definite off-nadir angle associated with each amplitude. The angle-independent information is received via one beam to port and one beam to starboard, both formed by the sonar to be broad in the vertical, and are not employed in the measurement of bathymetric slant ranges. In a vertically broad beam there are signal amplitudes available at multiple times after transmit and for essentially the entire interval between successive transmit pulses. The port (or starboard)-side broad beam is intended to receive signals throughout an angular sector which is as large, or larger, than the span of the port (or starboard)-side narrow off-nadir beams. Consequently, the angle-independent signal amplitudes are associated with a range of off-nadir angles which encompasses the range of off-nadir angles for the narrow beams. Furthermore, the range of times after transmit when the angle-independent signals are received (and sampled) encompasses the round-trip travel times associated with the slant ranges to the bottom as observed in the each of the multiple, narrow off-nadir beams. The set of multiple off-nadir angles and slant range travel times to the bottom are regularly processed using navigation, heading, heave, roll, pitch, and the sound velocity profile to provide a set of bottom positions and depths. That set of bottom positions and depths can be associated, through matching slant range travel times, with a set of particular receipt (sampling) times after transmit in the angle-independent signal amplitudes. Those combinations can then be used to guide the designation of positions and depths for the remaining and intervening receipt times of the angle-independent signal amplitudes. This processing scheme incorporates effects of refraction and ship motion in determining the horizontal coordinates of the backscatter area (footprint) associated with each receipt time after transmit.

In many instances, a reasonable estimate of the positions of points on the bottom associated with the angle-independent signal amplitudes can be made without relying on the bathymetry. To do so requires assumptions that the port and starboard angle-independent imagery beams are orthorgonal to the ship's heading, that the bottom is flat in the port and starboard cross-track directions, and that the depth is defined by the time after transmit to the onset of "significant, sustained" levels of signal amplitudes in both the port and starboard angle-independent beams.

Figure 1 illustrates the along-track and cross-track footprint dimensions for the beam by beam and angle-independent imagery for a RESON SeaBat 9001, operating with 15 meters of water below the sonar head. The along-track footprint dimensions of the two imagery types are identical. In the cross-track direction there are significant differences between the two types of imagery. It is obvious that angle-independent imagery should be capable of identifying cross-track demarcations between distinct backscatter regions with a higher spatial precision than beam by beam imagery.

A patch test can be devised to determine offsets in a shallow water multibeam bathymetry system via angle-independent imagery. In general one would expect the sonar to sense acoustic energy backscattered from any point on the bottom when that point is within the intersection of the transmit and receive footprints of the sonar (Miller 1997, Hughes Clark 1998). It is necessary that the acoustic backscatter signal from the bottom be sufficiently greater than system noise for it to be reliably detected. However for the backscatter from any particular point to stand out compared to backscatter from the surrounding area, it is necessary that there be a distinct spatial pattern in the backscatter properties associated with that point. The pattern may be composed of either reduced or enhanced backscatter. Such features do exist and can be repeatedly imaged, port and starboard, on parallel tracklines with headings that are identical and reciprocal, and over a range of ship speeds.

Assume that one distinct spatial pattern in the backscatter (feature) can be uniquely identified with a time "T" and cross-track distance "C-TD" in each of the several tracklines. In simple form, the horizontal positions assigned to each of the detected manifestations of the feature can be estimated as follows: 1) using the ship geometry parameters and the ship heading(H) at time "T", compute the sonar position at time "T" using the GPS antenna position at time "TNAV"="T" ; followed by 2) using the cross-track distance "C-TD" and sonar position, ship heading(H), and ship pitch(P) at time "T", compute the horizontal position of the feature on the bottom.

The set of several positions assigned to the feature can be used to estimate the heading and pitch offsets, as well as, any systematic offset between the sonar and navigation time tags. The conceptual steps to arrive at the offsets are: I) compute the average easting and average northing of the multiple positions, determined according to steps 1 and 2 above; II) compute the RMS distance "RD" from that central point to each of the multiple positions; III) perform multiple repeated positionings of the feature but with small independent zero-mean random perturbations in steps 1 and 2, added to "TNAV", "H", and "P", each followed by recomputation of "RD" according to steps I and II (this causes the computed positions to migrate horizontally by amounts which depend on ship speed, course made good, and relative position of the feature in the port/starboard scene when the imagery was acquired); and IV) determine the minimum value of "RD" from step III and the associated particular perturbations to "TNAV", "H", and "P". The net effect of some migrations will be more widely dispersed positions and the net effect of others will be more closely clustered positions. The perturbations which produced the minimum "RD" (tightest cluster of positions) are the expected values of the system offset between the time tags assigned to bathymetry and navigation and the system offsets in heading and pitch.

There are multiple possible variations on the conceptual imagery-based patch test described above. Those variations might involve other object functions than minimum RMS distances, employing rotations and horizontal shifts in the actual imagery rather than the detected positions, or using a specifically deployed manmade target rather than relying on natural features. Several variations have been investigated. The results from two variations will be presented in the following discussions.

Figure 2 illustrates a manmade target which was deployed to serve as a marker in the angle-independent imagery. It is nominally 30 centimeters in diameter and composed of twenty triplane corner reflectors assembled in the form of an icosahedron. The target is so bright (highly reflective) as to be detected by the angle-independent imagery over a wide range of azimuth. Consequently acoustic reflections from the target are present in the angle-independent imagery at a wide range of times after transmit as the sonar advances on any trackline in the region of the target.

Consider a straight trackline with a point target on the bottom to starboard of the ship. The slant ranges from the multibeam sonar to that target, as the ship travels past the target, will be a specific hyperbola defined by the depth of the target and the cross-track distance to the target at the closest point of approach (CPA). If the ship were to attempt moving along that exact straight line past the target, while being subjected to heave, roll, pitch, yaw, crabbing, deviations from the prescribed trackline, etc, the slant ranges would deviate somewhat from the specific hyperbola of the ideal uniform straight trackline.

Figure 3 is an example of the RESON SeaBat 9001 starboard channel angle-independent imagery from a "straight" trackline past the marker target. It was acquired on the NOAA S/V Bay Hydrographer, under the command of Lt. Shepherd Smith, NOAA. The image has been corrected for speed but has not been converted from slant range to cross-track distance. The dimensions of the image are 25 meters along-track and 54 meters cross-track slant range. The image includes a small school of fish in the water column, but is visually dominated by a set of highlights arranged along a line with smoothly varying curvature. The highlights in the region of CPA represent the mainlobe response of the sonar's transmit and receive beam patterns which are nominally along a line perpendicular to the trackline. The highlights in the "wings" before and after CPA represent different combinations of mainlobe and sidelobe responses of the sonar's transmit and receive beam patterns. More importantly, each of the individual highlights is potentially associated with a different vertical angle between the sonar and the marker target. Therefore, and in a manner similar to the principle of Cross-Fans (Gutbertlet 1989), the exact positions of the highlights contain information related to local refraction. Additional research is underway to determine the best method to utilize this refraction information.

The NOAA S/V Bay Hydrographer ran a series of tracklines in easterly and westerly directions, to the north and south of the marker target, and at different speeds. The navigation control was provided via the STARLINK DNAV 212 and post-cruise positioning of the ship was performed using On-The-Fly post-processing of GPS observables (OTF-GPS) from a pair of Ashtech Z-12 receivers. The shore-based Ashtech Z-12 reference station used in the OTF-GPS post-processing was located less than 8 kilometers from the survey area. The angle-independent images acquired on those tracklines were first processed to determine the systematic difference in time tags which the system had assigned to the bathymetry (and imagery) and to the navigation. They were subsequently processed to determine the heading and pitch offsets.

The analysis started with isolating and extracting the set of highlights associated with the marker target in each of the several tracklines. The data were first subjected to a threshold criteria that reduced the number of candidate imagery pixels to approximately two percent of the available pixels. Figure 4 illustrates the locations in sample space of the pixels which passed the thresholding applied to port and starboard channels of imagery from line 1847. Those pixels were further down-sampled by manual selection to arrive at the target pixels presented in Figure 5. An initial estimate of the location of the marker target was made from visual inspection of plots (like Figure 5) for each trackline's imagery.

Slant range travel times between the sonar position at the sampling time of each target pixel and the marker target were computed using the heave, roll, pitch, heading, and positions of the ship. However, since the position of the marker target was not known precisely, the slant ranges were computed from the sonar to each point on a series of closely spaced grids, each surrounding the initial estimate of the marker target's horizontal position, but separated in the vertical. An object function was formed consisting of the difference between the actual observed slant range travel time of each target pixel and the computed travel time to the X, Y, Z coordinates of each grid point. The coordinates associated with the object function's minimum RMS for any given trackline were accepted as the best estimate of the marker target's X, Y, Z position based on the target pixels for that particular trackline. This process was repeated twice, once using positions provided by the STARLINK and once using the post-processed OTF-GPS positions. The several resulting estimates of the marker target's horizontal position, based solely on either of the two sources of positions, did not agree. The RMS distance from the average easting and northing of the marker target positions using post-processed OTF-GPS and using STARLINK were both in excess of 3 meters. This indicated there was an offset between the time tags assigned to the bathymetry (and imagery) and to the navigation. Figure 6 presents the along-track component of the several estimated marker target positions, projected onto the average trackline, as a function of the component of ship's velocity parallel to the average trackline. This figure is based on the post-processing OTF-GPS source of positions. The results obtained with the STARLINK positions was similar but exhibited greater scatter. The slope of the straight line, least mean square fit, through the data was 0.54 seconds, which is the estimated offset between the two sets of relevant time tags.

After the navigation time tags were adjusted with a 0.54 second offset, the data were reprocessed with a resultant slope of 0.002 seconds. The RMS distance from the average easting and northing of the revised marker target positions using post-processed OTF-GPS and using STARLINK were 0.13 and 0.52 meters, respectively. Since the two analyses employed the same set of target pixels and observed slant range travel times, the difference in RMS distances must be attributed to differences between the accuracy and/or precision of the STARLINK DNAV 212 and Ashtech Z-12 with OTF-GPS post-processing. Further investigations are underway to determine the growth rate in the RMS distance when known levels of a particular statistical distribution of random noise are added to the ship positions. It is anticipated that this may lead to a method for investigating the performance of a navigation system in any particular survey area.

After adjusting the navigation time tags, the angle-independent imagery from the set of tracklines run by the NOAA S/V Bay Hydrographer near the marker target was further processed to estimate the system offsets in heading and pitch. For that analysis, a set of natural features in the imagery was selected. The analysis returned to and proceeded in a different manner from the point where the number of pixels in sample space had been reduced by the thresholding process. All target pixels were removed. The remaining pixels primarily represented first bottom returns of the mainlobe and first sidelobe of the transmit pattern which are readily recognized in Figure 4. However, this set also contained pixels associated with specific distinct features, like the ones present in Figures 7 and 8. The angle-independent images in those two figures are from two reciprocal tracklines and have been corrected for speed and slant range. The non-target, thresholded pixels were "cookie cut" in sample space to isolate five specific distinct features as they appeared in several different renderings of the imagery. Figure 9 shows three distinct features which were isolated from the thresholded pixels of line 1847, previously presented in Figure 4. For each trackline and each of the five features that may appear along the particular trackline, the average easting and average northing were computed for the pixels representing each feature. The RMS distance (radius) from the central position of each feature to each of that feature's pixels was also computed. An RMS across all tracklines was computed of the RMS radii of each feature and finally an RMS of those radii was computed across all features. This final RMS was the RMS distance (radius) to each of the pixels of each of the features, as if their central positions were co-located. It is subsequently referred to as the base radius.

The position of each pixel was recomputed by means of a systematic perturbation applied to the reported ship's heading and pitch at the time when that pixel was initially sampled. The various perturbations cause the pixels defining each feature to migrate. The direction of the migrations depends on course made good and whether a particular feature was to port or starboard of the trackline when it was initially imaged. The net effect of the migrations was a set of pixel positions for each of the five features which were either more widely dispersed or more closely clustered. The perturbations which produce the tightest collective cluster of pixel positions for the features are the expected values of the system offsets in heading and pitch. The previously described determination of RMS radii was made for each of the perturbation iterations in heading and pitch. An object function was computed as the square root of the difference between the base radius squared and the squared RMS radius to each of the repositioned pixels for each of the features. Figure 10 presents contours of the object function associated with perturbations to ship's heading and pitch. The minimum of the object function occurs when the perturbation in heading is 1.75 deg and the perturbation in pitch is 0.52 deg. Those are the estimated offsets in the ship's heading and pitch.

Figure 10. Contours of Object Function used to Determine Heading and Pitch Offsets
4. Comparison Of Offsets Determined By Different Techniques

The angle-independent imagery discussed in this report was acquired shortly after the initial installation of the RESON SeaBat 9001 aboard the NOAA S/V Bay Hydrographer. The system offsets determined at installation were: time, 0.5 sec; heading, 2.0 deg; pitch, 1.5 deg; and roll, 1.7 deg. Those offsets were determined using positions provided by the STARLINK DNAV 212. The calibration procedure was performed by the ship's survey personnel using the calibration features in CARIS HIPS (Godin 1996b). The system offsets determined by the angle-independent imagery described above are: time, 0.54 sec; heading, 1.74 deg; and pitch, 0.52 deg.

A technique has been described which was developed to use angle-independent imagery from a suitably equipped shallow water multibeam survey system to estimate the offset between time tags which the system assigns to the bathymetry and navigation, as well as, offsets in heading and pitch. The work presented has utilized natural bottom features and a passive manmade marker target. Conceptually, when using a bright manmade reflector like an icosahedron or possibly an active acoustic repeater, the technique can also estimate the system roll offset. The angle-independent imagery patch test is easy to implement and has been shown to produce results comparable to the present shallow water patch test method used by NOAA. Additional investigations are ongoing which include extending the angle-independent imagery patch test technique to provide quality assurance on sound velocity refraction correctors and positioning systems.

Crews, N. L.,"Depth Accuracy Analysis on the Intermediate Depth Swath Sonar System (IDSSS) Hydrochart II", The Hydrographic Journal, No. 57 July 1990, pp.11-16.

Du, Z., L. Mayer, D. Well, and L. C. Huff, 1994, "Uncertainty in SIMRAD EM1000 and RESON SeaBat9001 Multibeam Echo Sounders in Shallow Water--A Case Study", Proceedings of the US Hydrographic Conference, April 18-23, 1994, Norfolk VA, pp.103-110

Godin, A., 1996a, " The Calibration of Shallow Water Multibeam Echo-Sounding Systems", Proceedings of the Canadian Hydrographic Conference, June 3-5, 1996, Halifax, Nova Scotia, pp.25-31.

Godin, A., 1996b, "Field and Processing Procedures for the Calibration of Multi-beam Echo-sounding Systems", Universal Systems, Ltd., Fredericton NB, Canada, p.44.

Gutbertlet, M., H. W. Schenk, 1989, "Hydrosweep: New Era in High Precision Bathymetric Surveying in Deep and Shallow Water", Marine Geodesy, Vol.13, pp.1-23.

Huff, L. C., H. Orlinsky, and S. Matula, 1994,"Using Near-Shore Multibeam Systems for Survey Investigations", Proceedings of the US Hydrographic Conference, April 18-23, 1994, Norfolk VA, pp.32-39.

Hughes Clarke, J. E., 1998, "The effect of fine scale seabed morphology and texture on the fidelity of SWATH bathymetric sounding data", Proceedings of the Canadian Hydrographic Conference, March 10-12, 1998, Victoria, BC, pp.168-181

Mayer, L., J. E. Hughes Clarke, D. Wells, and HYGRO'92 Team,

1993, "A Multi-faceted Acoustic Ground-Truthing Experiment in the Bay of Fundy", Acoustic Classification and Mapping of the Seabed, Proceedings of the Institute of Acoustics Volume 15 Pt 2 1993, pp.203-219.

Miller, J., J. E. Hughes Clarke, and J. Patterson, 1997, "How Effectively Have You Covered Your Bottom?", The Hydrographic Journal, No. 83, pp.3-10.

Pryor, D. E., 1982, "Performance Characteristics of the Bathymetric Swath Survey System", NOAA Technical Report OTES-9, May 1982, p.205.

Wheaton, G. E., 1988,"PATCH TEST, A system check for Multibeam Survey Systems", Proceedings of the US Hydrographic Conference, April 12-15, 1998, Baltimore MD, pp.85-90.

Revised Thursday October 25 2001by OCS Webmaster