Contact:
Tyson Hilmer
University of Hawaii
(808) 956-4103
hilmer[at]hawaii.edu

Latest image from Ipan, Guam:

Site:
The experiment site is at Togcha Bay near Ipan, Guam. Togcha Bay is characterized by a shallow and uniform reef composed of macroalgae and turf overlying a pavement substrate.  The entire nearshore section of the reef, approximately 500 m in cross-shore distance, is regularly exposed at low tide.  The island of Guam was chosen because of the frequency of typhoon wave and wind driven events.  The uniform profile of Ipan Reef is ideal for estimation of the effects of bathymetry and friction.

Some images from installation.

A CDIP buoy is located just offshore of this site.

Setup:
The video acquisition equipment consists of a Sony SSC-E473 color video camera with a Fujinon 2.7 mm auto-iris lens.  The equipment was installed atop an 8 m high concrete building overlooking the southern half of Togcha Bay on August 20th, 2005.  Power and data signal are transferred via co-axial cable to an indoor logging computer.  A Matrox Meteor II frame grabber converts the analog video to digital input. Image aquisition and initial pre-processing are performed using automated Matlab software (image aquisition and image processing toolboxes).

Data Products:
Full frame RGB images are captured in uncompressed format every 15 min (top-left).  A single line of pixels spanning the image, hereafter pixel-line, is captured at 2 Hz. The pixel lines are then stacked left to right, creating a "time-stack" (top-center). Thus, the horizontal axis corresponds to time. The vertical axis is a proxy for distance, although the scaling is not linear due to the view angle. After rectification, features with a constant velocity form a line of constant slope.

Some timestacks from a storm event.

Recently I've added a third data product, quite similar to the timestack. The pixel-line is captured at 20 Hz, and the variance is calculated over 0.5 s intervals. This "v-stack" (top-right) is concatenated at the same frequency (2 Hz) and spatial resolution of the timestack. V-stacks can be thought of as a form of motion-detection, as features changing in time are emphasized. The sampling frequency and variance interval can be tuned to the temporal span of the feature of interest.

The images are then rectified to real-world coordinates using the equations outlined in Holland et al (1997)*. The Littoral Dynamics Teams of the NRL have an excellent website outlining their techniques. A fairly knowledgeable website on camera calibration can be found here.

Broken wave heights, offshore and secondary swell wavelengths and velocities, and wave refractions are successfully resolved using the timstack technique.  A novel feature-detection algorithm was developed for accurately and autonomously generating a time series of broken wave heights.Initial comparisons between video measurements of breaking wave height show a strong correspondence with in-situ pressure sensor data.

My senior thesis on this work.

Current Work:
Quantification of other features from video images, currently working on the spatial and temporal extent of bores.
Comparison to in-situ pressure sensor and ADCP data.

* Holland, K.T., Holman, R.A., Lippmann, T.C., & J. Stanley. 1997. Practical use of video imagery in nearshore oceanographic field studies. IEEE J. Oceanic Engineering. Vol 22. No. 1.

Surveyed ground control points (GCP-s) used in calibration of the camera model.  Colored dots are corresponding image coordinates (u, v).  Vertical black line coincides with image pixels sampled for timestacks.


Spatial coordinates of ground control points (red) and image pixels (black) as calculated from rectification equations.  Pixel footprints are on the order of 1 cm within 100 m of the camera, and increase to order 10 m within 500 m offshore.

Rectification of image to horizontal and vertical planes.  The horizontal plane extends approximately 600 m cross-shore.  Note difference in vertical scale.


Comparison of camera model to geo-referenced tagged image file with 4 m resolution (DOC, 2004).  The visual agreement is quite good, although a quantitative comparison has yet to be made.

Some timestacks from a storm
on 8.30.05

More images from this storm