 |
VIDEO
METRIC
SYSTEMS
from NorthWest Research Associates, Inc. |
US Army Corps of Engineers |
 |
Argus Beach Monitoring Station
|
VIDEO
METRIC
SYSTEMS
from NorthWest Research Associates, Inc. |
US Army Corps of Engineers |
|
INTRODUCTION: The key parts of an Argus
Beach Monitoring Station (ABMS) are one or more video cameras looking down at
oblique angles along a beach, and an image processing and analysis system built
around a SGI Unix workstation.
The ABMS produces three standard
image files: a snapshot (SNAP), a digitally averaged time-exposure (TIMEX), and a
pixel standard deviation (SIGMA) image. Example images are presented below.
In addition, the ABMS can produce pixel time series records at specified locations
to measure hydrodynamic processes such as wave run up, surface currents and nearshore
wave direction.
Analysis of shoreline position is presented from 11 months of
images acquired from Camera 3 at Lake Worth Inlet (http://www.planetargus.com/ABMS_sites.html - click on
Florida in the map).
(Click on image to enlarge)
TIMEX IMAGE: An example of a time-exposure ("TIMEX") image. This image is created by averaging pixel intensities collected once per second over a 10-minute averaging period. TIMEX images filter variations in wave breaking due to wave groups, providing more distinct locations of wave dissipation. Locations of wave breaking (dissipation) serve as a proxy for bar and shoreline positions.
(Click on image to enlarge)
SIGMA IMAGE: A SIGMA image displays the standard deviation of pixel intensity over the same averaging period as the TIMEX image. In a SIGMA image, areas where the illumination is changing during the sampling period appear bright, whereas areas where the illumination is relatively constant appear dark. For example, a sandy beach appears bright in a snap-shot or TIMEX image, but dark in the SIGMA image. Similar to the TIMEX images, SIGMA images reveal locations of wave breaking which serve as a proxy for morphology feature location.
(Click on image to enlarge)
Sometimes the SIGMA image shows changes other than the changing shore break. For instance, boat tracks appear as bright lines in the image above.
With intermittent sun caused by passing clouds during the sampling period, the dry beach appears bright.
(Click on image to enlarge)
Camera 3 is an excellent example of the maximum range obtainable with an ABMS while still providing survey-quality spatial resolution (nominally, pixel resolution is 1m in the cross-shore, 8m in the longshore). The area of interest in the image (the apex of the bend in the shoreline) is 2.5 km from the ABMS. This range and resolution is possible because of both the height of the cameras (150 m) and the power of the lens (focal length=50 mm).
(Click on image to enlarge)
RECTIFIED AND MERGED TIMEX
IMAGES: Through sophisticated camera models and image processing techniques, ABMS images
may be converted from image coordinates to real world coordinates to present a rectified view
of the coastal zone. Via the coordinate transformation process, ABMS data yield nearly
continuous and quantifiable information on beach morphology characteristics and nearshore
hydrodynamic conditions.
Below is an example rectified image produced
from an oblique TIMEX image from Camera 3. The coordinates shown are the
ABMS station coordinates. However, any user-requested coordinate system
can be presented since the camera location and images are ground-controlled and referenced
to local bench marks.
(Click on image to enlarge)
Below is an example rectified and merged TIMEX image from Cameras 1,2,3, and 4. Approximately 3 km of shoreline are observed with this image. This broad view provides the coastal engineers and managers a "system-wide view" of the area of interest.
(Click on image to enlarge)
DATA ANALYSIS - SHORELINE MAPPING: Shoreline identification from video involves mapping between image coordinates (pixel space) and local coordinates (e.g., NAD27). Transformation between the two-dimensional pixel space (UV) and three-dimensional world coordinates (XYZ) is typically an underdetermined photogrammetric problem with a monoscopic camera view. To solve the transformation for this application, the elevation (Z) is set to measured tide level relative to Mean Low Water (MLW). The tide level is assumed to correspond with the land-sea interface. With the Z-coordinate known, each pixel UV coordinate of interest can be mapped to a XY world coordinate. This transformation includes corrections for optical distortions in the lens based on laboratory camera-lens calibrations. Shoreline position is determined for each horizontal scan line across a regiona of interest in the image. ABMS images have 640 horizontal pixels (U-coordinate) and 480 vertical pixels (V-coordinate). The figure below shows a region of interest outlined by polygons and the location of an example pixel transect or horizontal scan line. The companion figure shows each transect that passes through the area of interest is searched seaward from the beach for the pixel intensity gradient caused by swash. A 5-point zero-phase running average filter is applied to each transect to reduce the effect of individual bright pixels on the beach that are not of interest (e.g., umbrellas, beach towels, birds, or a bad pixel in the camera's CCD). Pixel intensity maximum is identified on the scan line, then sequences of pixels that are within 90% of the maximum intensity are least squares curve-fit with a combined quadratic and Gaussian function(after Plant, N. G. and Holman, R. A., 1997. Interannual shoreline variations at Duck, NC, USA. In: Proceedings of the 25th International Conference Coastal Engineering, pp3521-3533). The first landward point on the curve that exceeds 95% of the maximum is selected as the cross-shore (V) coordinate. Selecting a threshold on the landward side improves the shoreline determination if the swash zone is wide and peak intensity is biased seawards (energetic conditions).
(Click on images to enlarge)
A series of QA/QC routines are used for editing unrealistic shoreline determinations in image space. First, a polynomial curve (typically 6th -8th order) is fit to the entire shoreline and outliers that deviate from the curve by a selected threshold are excluded. Next, the shoreline is divided into segments that overlap by 50%. Polynomials are fit to these sections and outliers exceeding a deviation limit from the curve are rejected. Three iterations of curve fitting are done on each segment, with tighter tolerance of the deviation limit applied to successive iterations. Data gaps of less than 3 pixels are linear interpolated and multiple solutions from the overlapped segments are averaged. This analysis produces sequence of UV- coordinates of shoreline position as shown in the following figure.
(Click on image to enlarge)
DATA ANALYSIS - MAPPING ACCURACY: Transformation of shoreline UV coordinates to XY coordinates requires a "geometry solution" for the image. A geometry solution accounts for camera angles (tilt, pitch, and roll), horizontal field of view, lens distortion and the image processor. Small variations in camera angles can occur with movement of the camera mount (e.g., thermal expansion of the mount) and are corrected for analysis. At the Lake Worth Inlet ABMS, variations in camera angles are caused by a combination of the powerful lens (50mm) on Camera 3 and the 0.5m-long horizontal arm of the camera mount. Daily heating and cooling of this arm causes the camera's tilt angle to change enough to cause alongshore movement of targets in the field of view on the order of 30m, 2.5km away from the camera. Movement of the 150m high condominium may also contribute to camera angle variations. To correct for variations in camera angles, a unique geometry solution is computed for each image using an "AUTOGEOM" program that compares each image to an image with a geometry solution that is known to be accurate. The automated co-registration routine determines the relative shift and rotation in each image and calculates a new geometry solution. Following which, each shoreline UV-coordinate is converted to a XYZ-coordinate (Z=tide level) based on the AUTOGEOM solution. These coordinates are converted to NAD27 and then to State Plane, NAD83 with the Corpscon program for this project. The two figures below demonstrate the improved geometry solutions using the AUTOGEOM program. An edge detection routine was used to locate the image coordinates (UV) on a section of the south jetty (yellow box). These pixel coordinates (lower plot) and the corresponding XYZ coordinates (upper plot) were with AUTOGEOM (blue) and without (magenta). AUTOGEOM solutions reduced the pixel variation from about ±3 pixels to about 1 pixel. In this example, at that range, that amounts to a reduction of longshore position errors of ±10m to ±2m.
(Click on images to enlarge)
DATA ANALYSIS - SHORELINE EVOLUTION: A simple way to present
shoreline changes as a function of time is to overlay the maps of shoreline
location. This type of presentation can be used to quickly identify areas
of high and low variability in shoreline change. The figure below shows shorelines
derived from images collected from September 2001 through August 2002 by Camera 1 (looking
at shoreline north of the inlet) and Camera 3 (looking at shoreline south of the inlet).
The shoreline positions shown are 5-day averages at tide levels of Z=1 m (+/- 0.15 m).
On the south side of the inlet, most of the shoreline variation occurs within 1 km of the jetty.
The large envelope of shoreline position in this region is caused by the placement and diffusion
of dredged material during channel maintenance.
(Click on image to enlarge)
DATA ANALYSIS - TIME-SERIES OF SHORELINE LOCATION: Another
way to present shoreline change is to plot the shoreline location as a function of time. The
figure below is an example of the kind of
information one can glean from such a presentation. The figure presents a time series
surface plot of the difference in shoreline position from the mean shoreline. The x and z axes
are northing (alongshore) and easting (cross shore) in FL State Plane coordinates, respectively.
The y-axis is time from September 2001 through August 2002. The placement of dredged material
on the intertidal beach immediately adjacent to the south jetty occurred during the month of
February. This time history figure may be used to quantify the rate of alongshore diffusion
of the fill along the Z=1 m MLW contour (~3.5 m/day from Feb thru June).
(Click on image to enlarge)
DATA ANALYSIS - TEMPORAL EVOLUTION OF NEARSHORE FEATURES: MOVIES OF TIMEX
IMAGES: ABMS images not only provide quantitative information about
shoreline location, offshore wave direction, alongshore current magnitude and
direction, and shoreline swash magnitude and frequency, but it also maps
sub-aerial and sub-acqueous nearshore features such as dry-beach area, submerged
sand bar location and movement. And, to aid the engineer and manager in
looking at system-wide trends and behavior, the ABMS can provide movies of
the TIMEX or SIGMA images that reveal these features.
Click on the
figure below to view a 3-month movie of daily TIMEX images. Note the
change with time of the shape of the bright line that denotes the shoreline
location and the bright forms offshore that denote the location of submerged
sand bars. Shoreline attachment of submerged sand bars to small seaward
projections of the shoreline can be observed half way through this movie
clip.
A short, 3-month movie clip is useful in the analysis of
storm effects; the increase in the bright features offshore are an indication of
high waves breaking on the sand bars. The absence of bright features is an
indication of little or no wave breaking, not an absence of sand bars.
Looking at the features near the beginning and end of a storm can reveal
storm-induced changes in the sand bar morphology that can affect subsequent
shoreline changes. A year-long movie clip can reveal seasonal affects on
both the shoreline and the offshore sand bars.
(Click on image to run movie)
Higher resolution animations are found in the ftp directory (~90MB each). These are the files RectMovie5nc.avi and RectMovie2nc.avi.
These are but a few examples as to how ABMS images can be analyzed and presented to reveal quantitative information about the physical features of the nearshore environment. Some of this information can be obtained using standard surveying methods. However, the high spatial and temporal resolution of the ABMS surveying provides information about the nearshore environment never before available to engineers and coastal managers. For more information about ABMS please contact info@planetargus.com. For more information regarding the ABMS application at Lake Worth Inlet, FL, contact William Curtis at William.R.Curtis@erdc.usace.army.mil or Kent Hathaway at Kent.K.Hathway@erdc.usace.army.mil.