Skip to content

History

mpalmsten edited this page Jan 25, 2017 · 9 revisions

← Previous topic: Mission | Next topic: Future →

#Historical Development of the Argus System

The move to optical remote sensing and the development of the Coastal Imaging Lab at Oregon State University was a response to some very practical problems of studying infragravity wave dynamics under stormy conditions on the Oregon Coast. With wave heights often exceeding 5 m and surf zone widths reaching 1 km, sampling with traditional in-situ sensors was viewed as impractical, if not life-threatening!

The first roots of our optical approaches came from the use of time-lapse photography to capture runup time series of infragravity swash over a longshore span of beach. Data were played back one frame at a time and the swash location manually digitized using a variety of pointer devices ranging from a toy train to the track of a drawer slide. Frequency-wavenumber spectra of the resulting data were used to detect and study very large-scale edge wave motions [e.g., Holman and Bowen, 1984].

The time-exposure images that have become the primary Argus product were discovered by accident, as part of a study to detect longshore-standing edge wave motions on a pocket beach. To supplement swash data being collected optically along the length of the beach, it was decided to take 10-minute time-exposure images of the beach and nearshore using a 35 mm camera with a 13-stop neutral density filter. The hope was that dominant standing wave patterns would be directly revealed by a visible nodal pattern in the swash zone. While the time-average swash results turned out to be uninformative, the time-exposure image did reveal a surprising offshore band of strong wave breaking that was felt to be related to an offshore sand bar (Figure 1). This result rapidly became the focus of investigation until the link between sand bar location and its time-exposure signature was specifically demonstrated in ground-truth tests at Duck [Lippmann and Holman, 1989]. Further investigations have refined this relationship [e.g., Aarninkhof et al., 1997; van Enckevort and Ruessink, 2001; Aarninkhof et al., 2003].

Figure 1. The original time exposure (lower) and corresponding snapshot (upper) taken at Short Sands Beach in 1981. The offshore band of white (preferred breaking) was unexpected but was surmised to be due to the presence of a submerged sand bar (later proven to be true). For scale, note the two people walking at mid-beach in the snapshot (very small black object - the beach is ~800 m long).

The beginning of regular, long-term sampling of sand bar variability via time-exposure images also began fortuitously at the end of the SuperDuck field experiment at Duck, North Carolina, in 1986 [Crowson et al., 1988]. In closing down experiment sampling, it was decided to leave one of our video cameras on the newly constructed observation tower (built due to the tireless efforts of Col. Grumm, USACE). A VCR was programmed to collect daily 15-minute recordings of surf zone waves. Tapes were subsequently sent to OSU for post-processing to create daily digital, 512 x 480 pixel, time exposures, as above, using a recently acquired digital image processing system. It rapidly became apparent that our preconceptions of the nature and time scale of sand bar variability at Duck were completely naive [Lippmann and Holman, 1990]. The value of and need for long-term, low-cost measurements of sand bar variability became obvious. Argus was created to fill this need. The data collection effort at Duck from 1986 to 1993 is sometimes retrospectively called Argus 0, the pre-cursor to subsequent automated systems.

##Argus I The value of regular timex sampling was obvious but the tedious processes of videotape collection, shipping and post-processing motivated automation of timex creation. The first automated Argus Station in 1992 was based on a Dipix image processing board hosted in a DOS computer, deployed at Yaquina Head, Oregon, and connected by modem back to OSU. Analog black-and-white video signals from two cameras were digitized at 3.3 Hz and averaged into digital timex images of 640 x 480 pixels each. Collections were scheduled hourly, with data transmission to OSU programmed to occur nightly. Automation of data collection and return enormously simplified the sampling process and allowed continual (daylight) monitoring of sites of scientific interest without intervention. This continuous monitoring capability lead Paul O'Neil, an engineer responsible for early development, to give these computer stations the name Argus Stations after the hundred-eyed dog that constantly watched over Io in Greek mythology.

Because time-exposure images paint a picture of beach morphology over only a swath of beach (the surf zone) that shifts cross-shore position with the tide, the idea was developed to compute daytimex images to average over tidal shifts. Variance images were also introduced at this time as a method to isolate breaking waves in the surf zone from bright but unchanging regions such as the sandy beach. The time-exposure and variance image products have considerably changed the way we think about sand bar dynamics and Large Scale Coastal Behavior (LSCB).

At around this same time, we began to explore the use of time series data from individual image pixels to extract information about the wave field that forced observed sand bar changes [Lippmann and Holman, 1991]. Sampled at 2 Hz, data could be collected from arrays of pixels in what eventually evolved into the current pixel instruments. However, the technology of the early DOS Argus computers could not keep up with the increasing sampling demands of pixel time series and a new generation of Argus Stations was needed.

##Argus II Argus II was designed around a computer host, the SGI O2 Unix workstation that had both a robust and flexible computing environment and a native ability to digitize video data. Argus II was based on color cameras whose frames were digitized into 640 x 480 pixel images that were of considerably better quality than those from Argus I. Moreover, pixel time series data could be collected robustly at 2 Hz due to the SGI's video capability.

The limitation of Argus II lay in the fact that the SGI had only one video input and could digitize only one video signal at any instant in time. Selection of input from among a suite of cameras was made by a computer-controlled video switch and digitization of two cameras was handled by rapid switching between them. However, the data were not quite synchronous (¼ s offset) and the finite switching time of the video switch meant that, at most, two camera inputs could be handled in this manner at any time. The lack of synchronicity introduced processing complications in pixel array data that spanned two cameras and meant that stereo analysis of surf zone waves was not possible due to wave motion in the intervening ¼ s. The limitation to two inputs implied that pixel arrays that might naturally span many cameras could only be sampled from a sub-array that was constrained to two selected cameras. Thus data collection from a typical 5-camera station involved a sequence of sub-collections, for example sampling cameras 1 and 2 for ten minutes, followed by 3 and 4, then finally camera 5. While a very capable system, by 2002 these problems limited the potential role of Argus II for nearshore sampling and a new generation was needed.

##Argus III The third and current generation of Argus was developed jointly by John Stanley of Oregon State University and Irv Elshoff of what was then called Delft Hydraulics. Argus III is based on digital video cameras with 1024 x 768, 1280 x 960, or even better, pixel resolution, a considerable improvement in quality from Argus II. All cameras at a site are connected directly to a host Linux computer as digital devices, using FireWire connections. Cameras are pulsed (usually at 2 Hz) by a common external trigger so that frames are truly synchronous and stereo analysis of moving targets (waves) and cBathy-type cross-spectral analyses could be used.

Pixel arrays can be designed to span any number of cameras. Pixel time series can be collected from all cameras at once and can be collected at the same time as other types of collections, such as time-exposure images. In fact, there is substantial flexibility in scheduling, including the potential for nesting short hourly pixel time series collections and timexes within a longer pixel time series collection such as might be needed to study low frequency wave motions. Much of this new capability came from the adoption of a robust (Linux) operating system as the system controller.

Examples of the four generations of Argus images are shown below from the FRF, the longest running Argus station. In addition to the great advantages of using increasingly robust operating systems, the image quality has also greatly improved.

Figure 2. Four generations of Argus snaps from the FRF at Duck, NC. The dune is much wider in early images than at present.

In 2013, the CIL began using better Firewire cameras with a pixel resolution of 2048 by 2448. These were first installed at Cape Disappointment overlooking the mouth of the Columbia River where sampling to extreme distances (close to the horizon) was required. These stations are not considered to be a new generation but clearly continue the evolution of image acquisition. Development continues with new GigE cameras now being installed. One of the purposes of this wiki is to host an open discussion of these developments.

Clone this wiki locally