The Ultra-High Resolution Future of Hydrography

The Ultra-High Resolution Future of Hydrography
Thomas Meurling, The RESON Group
Marlene Baldwin, The RESON Group
Doug Lockhart, Fugro Pelagos, Inc.
Chris Malzone, The RESON Group
Abstract: As sensor technology improves, so does the face of the hydrographic
industry. The next generation of Multibeam Echo Sounders, with their ultra-high
resolution , multiple frequency capabilities and new signal processing functionality,
offer surveyors greater flexibility and the ability to identify features that would not
have been visible with older sonar systems. In addition, the availability of raw water
column and sediment data (i.e. snippets) facilitates improved post processing and
supports more scientific research applications. New processing methods, such as
adaptive bottom detection and CUBE algorithm, make data more accurate, thus
providing more reliable survey results.
This paper explores recent improvements in Multibeam Echo Sounder systems in
comparison with their older counterparts and discusses the potential effects these
advancements will have on the future of hydrographic survey activities both in
shallow and deep waters.
I. Introduction
Resolution in modern Multibeam Echo Sounder (MBES) systems is increasing at
a steady pace as newer generation systems are developed. Increasing beam
counts, sample rates, and bottom coverage result in a vastly higher volume and
accuracy level of data output, which in turn provides significant value in hydrographic
applications, offering more reliable results and showing features on the sea floor that
were not previously visible.
A significantly higher processing burden, however, can counteract these benefits
and raise survey costs to unmanageable levels. It is important, therefore, to develop
processes and algorithms that will assist scientists with the evaluation and
compilation of survey results.
II. Ultra-High Resolution
Improvements in sonar resolution result from a number of technology
advancements, including increasing the number of soundings against the sea floor,
improving the quality of the bottom detection process, and creating techniques that
yield better bottom coverage.
A. Increasing the Number of Soundings
In traditional equiangular beam forming, the number of beams per degree of
swath is kept constant as the steering angle moves away from nadir. This results in
an elongated distance between soundings – thus a degradation of data quality – at
the outer edges of the swath. To compensate for this, modern MBES often utilize
equidistant beam forming techniques. This method keeps the distance between
soundings consistent by gradually increasing the number of soundings per degree of
swath. As an example, in the SeaBat 7125 400 kHz system, there are 256 beams
formed in equiangle mode and 512 beams formed in equidistant mode over the
same 150° swath.[1]
Figure 1 presents the number of beams as a function of steering angle for a 1°
nadir beam spacing.
Figure 1: Beam Density
Figure 2 demonstrates the difference in beam spacing as the steering angle
moves away from nadir.
Figure 2: Beam Spacing
Figure 3 demonstrates the difference in sounding density between the two
methods.
Figure 3: Across-Track Sounding Location on a Flat Bottom
B. Improved Data Quality
A few simple models demonstrate the effect of the improved vertical and
horizontal resolution of the SeaBat 7125. The following samples model the behavior
of the SeaBat 8101 and SeaBat 7125 sounders in a water depth of seventy meters.
At this depth, the SeaBat 8101 horizontal resolution is about 1 sounding every 3
meters across track. The SeaBat 7125 horizontal resolution is about 1 meter. Figure
4 shows a noise free surface roughly 250 meters on a side with a ten meter vertical
feature. A small tide error of 20 cm can be seen running through the center of the
feature. At this scale, and with no systematic noise, the modeled surface for both
systems are indistinguishable.
Figure 4: Surface at 70 meters with a ten meter high feature and a 20 cm vertical tide error.
Adding noise to the model gives us a remarkably different view. For these tests,
only digitizer and bottom detect accuracy was considered. The resolution of the
SeaBat 7125 was assumed to be 0.02m while the resolution of the SeaBat 8101 was
assumed at 0.08m, both modeled as normal distributions. Figure 5 shows the
theoretical data density and noise distribution for the SeaBat 7125 and SeaBat 8101
in 70 meters of water. At this depth, the improved horizontal and vertical resolution
of the SeaBat 7125 is clearly evident. For consistency, both figures are drawn using
the same gray scale look up table.
Figure 5: Simulated flat sea floor at 70m showing bottom detect noise for the 7125 (left) and 8101
(right).
In Figure 6, the sounder bottom detect noise is added to the seafloor feature
image. In both cases, the noise adds false roughness to the feature. In the case of
the SeaBat 8101 model, the roughness begins to distort the feature and completely
masks the tide error.
Figure 6: Seafloor feature with bottom detect noise for the SeaBat 7125 (left) and SeaBat 8101
(right)
This simple example highlights a problem that SeaBat 7125 and 8125 users have
recently discovered - improving vertical resolution in the sounder makes other error
sources, such and tides or refraction, appear to be larger.
1. Full Array Calibration
Since none of the elements in the receive array can have absolutely identical
characteristics or mounting position, there will be small magnitude and phase
differences from one to the next. Without calibrating the sonar array, it is not
possible to relate the received data to any real-world measurements with a high level
of accuracy.
During the calibration process, a tone generator injects a tone of known
amplitude and phase into the Receiver channels. When this calibration signal is
received, the Sonar Processor generates a lookup table of both gain and phase
offsets by comparing the received signal to that transmitted. All subsequent signals
from that channel have these offsets applied to them before further processing. This
compensates for minor inconsistencies in all receiver channels.[2]
Newer techniques calibrate every channel using complex directional responses to
compensate for variations in the beam forming. The beam patterns shown below
illustrate difference between raw and calibrated returns.[3]
(a)
(b)
Figure 7: (a) Simulated center beam without compensation and (b) with compensation. [3]
(a)
(b)
Figure 8: (a) Simulated beam steered to 60° without compensation and (b) with compensation. [3]
2. Snippets
In 2001, Fugro Pelagos Inc, working with RESON Inc. and TritonElics, Inc.
developed Footprint Time Series (“Snippets”) processing. [4]
A Snippet is the series of amplitude values in the signal reflected from a beam‘s
footprint on the seafloor. One Snippet is produced for each beam for each sonar pin,
with the length of each Snippet varying as a function of the individual beam angle,
seafloor depth, and the Snippets operating mode. If the Snippets data for each swath
is concatenated, with each individual beam centered at its corresponding bottom
detect point (footprint) location, the combined series will provide a result similar to
slant range corrected sidescan imagery. In most cases, this will provide 100%
coverage of the seafloor.
MBES Sidescan Backscatter Sampling
MBES Sidescan Imagery
Snippets Backscatter Sampling
Snippets Imagery
Figure 9: Side-by-Side Comparison of Sidescan (left) and Snippets (right) Imagery [3]
Snippets processing has a significant impact on system resolution, providing
improved SNR, allowing for the detection of small features, and assisting in sea floor
characterization.
C. Improved Bottom Coverage
1. Multi-Ping
Deep water systems suffer from the effects of long two-way travel times. This
tends to result in decreased along-track sounding density, thus causing gaps
(“holidays”) in the collected data. Vessel motion – especially pitch – only serves to
exacerbate the problem. To remedy this problem, RESON has implemented multiping processing.
Multi-ping processing employs up to four separate frequencies transmitted during
one pulse interval, each with a slightly different steering angle. This alleviates the
problem without requiring sacrifices in survey speed.
The data plots below show -3dB beam footprints on the sea floor. Parameters
used for the modeling are real data taken from a SeaBat installation on “RV
Davidson”. Plot 1 shows holidays in the coverage when in single ping mode. Vessel
speed is 10 kts, water depth 4000m, 30 pings with a ping period of 20 seconds, pitch
±1.6°, yaw ±2.8°. Plot 2 shows the same model using 4-ping multi-ping. Vessel
dynamics are the same as plot 1. [5]
Figure 10: Comparison of Single Ping (left) and Multi-Ping (right) Bottom Coverage[5]
III. Multi-Purpose Survey
A. Raw Element Data Recording
The next generation of SeaBat sonars (also known as the 7000 series) provides
for the option of recording raw element (I and Q) data in addition to raw beamformed
(magnitude and phase) data. This data is recorded and stored either on the Sonar
Processor itself or on a separate, dedicated RAID array, then played back later
during post-processing,
This supports multi-disciplinary survey activities by giving surveyors the
opportunity to process the same set of raw data using various resolutions and
processing methods to meet different survey requirements.
B. Multiple Frequency
Historically, MBES systems were capable only of operating in one frequency.
Future systems will use two or more transmit arrays operating at different
frequencies and using interleaved transmission sequences, while a single
hydrophone receives and processes bottom returns for all frequencies.
Multi-frequency MBES systems will combine long-range, low or medium
frequency operation with shorter range high frequency / high resolution operation.
This gives surveyors the flexibility of carrying out multi-disciplinary surveys at a
variety of depths and with a variety of resolutions.
IV. Less Operator Intervention
A. Automatic Processing (AutoPilot)
More and more often, less-skilled operators perform physical survey operations
while expert surveyors perform just the on-shore post-processing activities. To
maintain survey quality, it is essential to have automatic processing routines (such
as RESON AutoPilot) monitor and adjust depth gate information as necessary during
a survey.
Once an autopilot function is initiated, the sonar processor analyzes bottom
detect data from a specified number of pings. If a significant number of those pings
are determined to have good quality, the average bottom depth and maximum
usable swath angle are determined. Based on this information, an appropriate range
setting is automatically chosen. The process then repeats, constantly updating the
range to maintain high quality survey data. [1]
B. CUBE Algorithm
With the extremely high volume of data produced by modern MBES systems (the
SeaBat 7125 can provide up to 2.2 billion soundings per day), it is necessary to find
automated methods for processing the majority of the data. In 2003, Dr. Brian
Calder of UNH proposed a new method for statistical processing of MBES data.
This algorithm is known as “Concurrent Uncertainty Bathymetric Estimator” or
“CUBE”. [6]
The CUBE method uses statistical redundancy to compute the most likely
measured depths (a hypothesis) from the multibeam data with the use of all the
information that is available. [7]
Figure 11: Sample CUBE Edit Window [7]
By incorporating the CUBE algorithm into post-processing software such as
RESON PDS2000 (shown above), the amount of manual review time can be
decreased drastically, thus making processing of survey data more economical.
It is important to note, however, that utilizing methodology such as CUBE
requires significant attention to detail when creating the model, as it is very easy to
miss good quality data or incorporate poor quality data into the survey information if
the wrong parameters are set.
V. Conclusion
Advancements in multibeam technology present both opportunities and
challenges for hydrographic survey teams.
Increasing the sounding density and data quality provides more detailed
information about features on the sea floor while also making surveys more cost
effective by reducing the minimum required distance between survey lines. In a
similar manner, improving bottom coverage through the use of multi-ping processing
also makes surveys more cost effective, because a survey can be conducted at
higher speed and still have sufficient bottom coverage for the data to be useful.
Raw element data recording and multiple frequency systems make it possible for
a single data set to be utilized for more than one application, paving the way for
multi-disciplinary survey activities. The ability to do this makes a single survey much
more valuable.
Of course, with these advancements, there are certain obstacles to be
addressed. The most significant of these is the fact that, as resolution and sounding
density increase, so does the sheer volume of data collected during a survey. If all
of this data must be processed manually, the cost savings realized during the survey
activities themselves would be outweighed by the significant increase in man hours
required to perform the post-processing. To counteract this, innovations are
constantly being made in automatic processing methods (such as Autopilot and
CUBE) which remove a large portion of the heavy post-processing burden.
The overall result of these improvements, however, is increasing quality and
usefulness of data along with decreasing operating cost.
References
[1] RESON Inc., “SeaBat 7125 Operator Manual, v5”, © 2007.
[2] RESON Inc., “SeaBat 8125 Operator Manual, v4.01”, © 2002.
[3] T. Meurling and B. Volberg, “The Evolution and Future of Multibeam Echo
Sounder Technology”, Proceedings of Underwater Technology, 2007; Tokyo,
Japan.
[4] D. Lockhart, E. Saade, and J. Wilson, “New Developments in Multi-beam
Backscatter Data Collection and Processing”.
[5] R. Lear, “Multiping Description v 1.3”, RESON Internal Memo, May, 2007.
[6] Calder, BR (2003). ‘Automatic Statistical Processing of Multibeam
Echosounder Data’, Int. Hydro. Review, Vol. 4, No. 1, pp. 53-68.
[7] RESON B.V., “PDS2000 User Manual, v. 3.1”, © 2007.