Project: Texture in Ocean Floor Images and Characterization of Habitat

Print version
Text Size: Change text to small (default) Change text to medium Change text to large
Enlarge Image
HabCam image of the ocean floor. Georges Bank, Coastal Massachusetts. (S. Gallager, N. Vine, R. Taylor)

Enlarge Image
HabCam image 2. Georges Bank, Coastal Massachusetts. (S. Gallager, N. Vine, R. Taylor)

Enlarge Image
HabCam image of the ocean floor. Georges Bank, Coastal Massachusetts. (S. Gallager, N. Vine, R. Taylor).

Joint work with Scott Gallager, Norman Vine, and Richard Taylor, WHOI

The understanding of ocean floor habitat is critical not only to fisheries management, but also to any prediction of how habitats will reorganize in response to factors like environmental damage or global climate change.  Although the ocean floor habitat has been studied using many sensing modalities, high resolution digital imagery has provided the most detailed information.  The Woods Hole Oceanographic Institution (Gallager) has been funded by NOAA to undertake ten cruises during 2005-2007 to map the Georges Bank region of Coastal Massachusetts.  The survey will produce hundreds of terrabytes of data in the form of digital images of the ocean floor of exceptional quality.  Faced with such a large volume of data, it is essential that we process information about habitat and substrate automatically and rapidly.  We need automated tools that not only analyze and classify the myriad varieties of habitats we consider important to bottom dwelling organisms, but also visualize those habitats at many scales.

We will be using a multiscale mathematical approach to the automated analysis, classification, and visualization of textures in ocean floor habitat.  Our incoming image data will be segmented at multiple scales using a combination of directional wavelet-based methods and a computational geometric method for representing images in terms of polygons that conform to the boundaries of the structures.  The multi-scale features will then be used to identify both biological textures and background textures.  Markov Chain Monte Carlo techniques will be used to determine the sampling widths required of the data.  The accumulated data will be interpolated to produce density profiles of the nonsampled areas, along with uncertainty characterizations.  These will be visualized at many scales ranging from that of the raw images to that of Georges Bank itself.  Research will be undertaken towards improving identification and classification algorithms, and improving the visualization package to encode spatio-temporal information, larval densities, and ocean currents.


  • Taylor, R., N. Vine, S. Tiwari, A. Girard, and S. Gallager.  2006.  "High Resolution Underwater Imaging and Image Processing for Identifying Essential Fish Habitat."  AGU/Ocean Sciences February 2006 Meeting, Honolulu, HI
  • Gallager, S., S. Tiwari, H. Singh, J. Howland, N. Vine, R. Taylor, and P. Rago.  2005.  "High Resolution Underwater Imaging for Characterization of Habitat."  In Special Session on Using Video Technology for Fisheries Applications, American Fisheries Society, 135th Annual Meeting, Anchorage, AK, September 2005


WHOI logo

rss headlines

Last updated April 17, 2006
© Woods Hole Oceanographic Institution. All rights reserved