HR: 1340h
AN: B43B-0287    [Abstracts]
TI: Crown Features Extraction from Low Altitude AVIRIS Data
AU: * Ogunjemiyo, S O
EM: sogunjemiyo@csufresno.edu
AF: California State University, Fresno, Department of Geography 2555 East San Ramon Avenue M/S SB69, Fresno, CA 93740
AU: Roberts, D
EM: dar@geog.ucsb.edu
AF: University of California, Santa Barbara, Department of Geography, Santa Barbara, CA 93106
AU: Ustin, S
EM: susan@cstars.ucdavis.edu
AF: University of California, Davis, Department of Land, Air and Water Resources , Davis, CA 95616
AB: Automated tree recognition and crown delineations are computer-assisted procedures for identifying individual trees and segmenting their crown boundaries on digital imagery. The success of the procedures is dependent on the quality of the image data and the physiognomy of the stand as evidence by previous studies, which have all used data with spatial resolution less than 1 m and average crown diameter to pixel size ratio greater than 4. In this study we explored the prospect of identifying individual tree species and extracting crown features from low altitude AVIRIS (Airborne Visible/Infrared Imaging Spectrometer) data with spatial resolution of 4 m. The test site is a Douglas-fir and Western hemlock dominated old-growth conifer forest in the Pacific Northwest with average crown diameter of 12 m, which translates to a crown diameter pixel ratio less than 4 m; the lowest value ever used in similar studies. The analysis was carried out using AVIRIS reflectance imagery in the NIR band centered at 885 nm wavelength. The analysis required spatial filtering of the reflectance imagery followed by application of a tree identification algorithm based on maximum filter technique. For every identified tree location a crown polygon was delineated by applying crown segmentation algorithm. Each polygon boundary was characterized by a loop connecting pixels that were geometrically determined to define the crown boundary. Crown features were extracted based on the area covered by the polygons, and they include crown diameters, average distance between crowns, species spectral, pixel brightness at the identified tree locations, average brightness of pixels enclosed by the crown boundary and within crown variation in pixel brightness. Comparison of the results with ground reference data showed a high correlation between the two datasets and highlights the potential of low altitude AVIRIS data to provide the means to improve forest management and practices and estimates of critical plant-variables that are required by major components of ecosystem and climate models.
DE: 0400 BIOGEOSCIENCES
DE: 1620 Climate dynamics (0429, 3309)
DE: 1640 Remote sensing (1855)
DE: 1851 Plant ecology (0476)
DE: 3322 Land/atmosphere interactions (1218, 1631, 1843)
SC: Biogeosciences [B]
MN: Fall Meeting 2005