You are here:
A HYBRID HIGH RESOLUTION IMAGE CLASSIFICATION METHOD FOR MAPPING EELGRASS DISTRIBUTIONS IN YAQUINA BAY ESTUARY, OREGON
Citation:
Clinton, P J., D R. Young, AND D T. Specht. A HYBRID HIGH RESOLUTION IMAGE CLASSIFICATION METHOD FOR MAPPING EELGRASS DISTRIBUTIONS IN YAQUINA BAY ESTUARY, OREGON. Presented at 5th International Airborne Remote Sensing Conference, San Francisco, CA, September 17-20, 2001.
Description:
False-color infrared aerial photography of the Yaquina Bay Estuary, Oregon was acquired at extreme low tides and digitally orthorectified with a ground pixel resolution of 20 cm to provide data for intertidal vegetation mapping. Submerged, semi-exposed and exposed eelgrass meadows with a range of densities and mixtures with macroalgae were clearly visible providing a good, if generalized, basis for heads-up habitat delineation. Scattergram analysis of an initial supervised digital image classification showed promise for distinguishing eelgrass meadows from macroalgal beds on the basis of color band ratios. Normalized difference vegetation index (NDVI) derived algorithms were developed to classify eelgrass, macroalgae, and non-vegetated areas at the pixel level and assessed at an overall accuracy of ~70%. This method proved to be effective in distinguishing intertidal vegetation from non-vegetated areas but less so in separating vegetation classes. A three-band NDVI derived algorithm with waterbody removal and operator adjustable constants was developed to produce three band intertidal vegetation mosaiced images. The resulting ~2 gigabyte image with 20 cm ground pixels was reclassified into five statistically distinct classes using unsupervised isoclustering and gridded into a raster geographic information system (GIS). A blockmajority generalization filter was applied to the resulting grid which was re-tiled and vectorized. A trained photo-interpreter with orthophotography as a guide extracted eelgrass polygons using GIS selection tools to produce maps of eelgrass meadows. Some heads-up digitizing was required to include the submerged margins of the eelgrass meadows. This method combines vegetation classification in digital image processing, computer-aided drawing in raster-to-vector conversion and photo-interpreter experience to produce maps less generalized than manual methods and more accurate than digital processing.