About ROV Frame Grabbed Images

History of the video frame grab system

In the beginning there was no facility for routine digital image frame capture on ROV expeditions. With the 1996-1997 VIMS/Vicki project came the purchase of SGI workstations capable of real-time digital image capture and processing. At first we simply plugged the camera feed into the system and used the system tool CosmoCapture to grab images and save them to disk. A script was written to transfer these images to another SGI system in the digital video lab where they could then be viewed from internal web pages. The script converted the original SGI RGB images (about 1 MB each) to 100% quality JPEG images (about 300 KB each) and removed the originals. The 100% JPEG images could be viewed in Netscape, yet preserved most of the detail within the image. This system was in place from about October 1996 to the summer of 1997.

We learned much from using this system, and the ease of retrieving the images was appreciated. There were several deficiencies with this system, namely:

There is no confirmation that images are actually being saved to disk

No ancillary information is saved with the images

Naming of files is error-prone

Image files are not organized in much of any fashion making their administration difficult

Difficult to capture other video sources, such as from the sonar or sit cam

Starting around July 1997 (at year day 216) we integrated the digital image frame capture system with the VIMS/Vicki software and the image file names changed to that of the frame code of when they were captured. The advantages of this are as follows:

The above inadequacies in the initial system are addressed
The fact that an image has be snapped can be entered into the VIMS database
Smaller (about 50 KB) 75% quality JPEG preview images with overlain text are created
The original SGI RGB images are compressed and saved, these can be easily converted to other formats for import to other programs with no loss of detail.
Annotation text appears on the image, which may encourage real-time use of the system for annotating video
Tape timecode is now saved with each image to aid in finding events on the tapes
The web interface to the images was improved

In December 1998, beginning with year day 363, we started saving the original images in PNG format rather than compressing the original RGB files. This change was made after making sure that there would be no loss of detail in the conversion from RGB to PNG. The advantages of this are as follows:

A savings of disk space. Each gzipped RGB file consumed about 750 KB, an equivalent PNG consumes about 500 KB.
No server-side conversion script is required to convert PNG files so that they can be viewed inside a web browser.
Integration with the VIMS query facility is easier with the PNG format.

In February 1999, beginning with year day 035, unneeded engineering-type information was removed from the text overlay put on to the jpeg preview image. With this, a new file (ending in '.overlay') is created with the 4 lines of overlay text. The original '.comment' file, with its 9 lines of information, is still kept in case there is any processing depending on the name/value pairs contained therein.

In the Summer of 1999 we begin to run out of disk space on our 9 GB drive on lepas for shore-side image storage. We are now in the process of moving the older image directories to CDROM. These will be network accessible and the URLs for each image will stay the same as we'll be able to NFS automount the CDROM into the web directory hierarchy.

What follows are details on the installation of the system on both vessels and links to the source code (Note: Much of this is out of date as the installation of the nav processor on the Point Lobos has changed things a bit.). This web page is not meant to be an instruction manual for how to use the VIMS/Vicki system, for more details on how one should use it, please see the Sample Collection web page or the Video Lab pages.

Ventana image grabbing system

Please see this document Installation of Vicki and image capture on the Pt. Lobos (caution: documentation is out of date)

The images are saved on host algae in directory /video2/Ventana/stills. There are subdirectories for the years and year-days for each exedition. This directory is mapped to the CanyonHead web server as http://mww.mbari.org/Ventana.

Tiburon image grabbing system

Please see this document Installation of Vicki and image capture on the Western Flyer

The images are saved on host algae in directory /video2/Tiburon/stills. There are subdirectories for the years and year-days for each exedition. This directory is mapped to the CanyonHead web server as http://mww.mbari.org/Tiburon.

The software

For ease of implementation existing packages were used to the greatest extent feasible. The major tools used include SmallTalk for the Vicki program, Perl for the vfcd and vifd daemons and the vfc2web script, ImageMagick for the image processing, cgi-bin scripts and web pages for the user access. Here are the development locations of the source code for the major components (please refer to the above documents for details on how these fit together).

vfcd: The Video Frame Capture daemon (development) is located on unix in /u/mccann/Media/vfcd
vifd: The Vims Interchange Format daemon (development) is located on unix in /u/mccann/Media/vifd
vfc2web: Video Frame Captures to Web (production) is located on unix in /u/mccann/Media/vfc2web
Vicki: The SmallTalk vicki code is developed on a Windows 95 PC, but a copy of the source (files ending in .st) is available on unix in /u/mccann/vims. The version of vicki as of this writing is 160, for what it's worth here is the vr160 source.
fmd, vtcc, vcrd, vtcd, libvicki.sl, libvicki.dll, etc.: These are supporting programs for vicki that Gene Gulielmo wrote. There are copies of the source in calypso:/users/gene, I have my own for development a porting purposes on unix in /u/mccann/vims/vicki.

Back to:

Last updated: 16 August 1999, Mike McCann