eCognition Segmentation of Crop Management Sub-units using Polarimetric SAR Data. (A03-bugden947702-oral)
Authors:
J.L. Bugden* - Agriculture & Agri-Food Canada E. Pattey - Agriculture & Agri-Food Canada H. McNairn - Canada Centre for Remote Sensing
Abstract:
The aim of this project is to retrieve biophysical crop and field descriptors and to delineate homogenous sub-units for crop management zones using remote sensing data. In this project, four methods of sub-unit delineation are being explored which include unsupervised classification, random block correlation, fuzzy k-means and eCognition segmentation. eCognition segmentation is based on both spectral and spatial heterogeneity and is well suited to low contrast data such as radar images. The analysis was done using Convair-580 polarimetric SAR data acquired over the Greenbelt Farm in Ottawa, Canada. Images were acquired on June 13th, June 26th and July 19th, 2001 and intensive field campaigns were conducted simultaneously on the ground (e.g., electrical conductivity, biomass,
nitrogen, chlorophyll, soil moisture, LAI, etc). The image quality and calibration accuracy were assessed using Polarimetric Workstation (CCRS). Preliminary data explorations show polarimetric SAR data to be sensitive to crop and field descriptors (e.g., nitrogen, senescence, tasselling, electrical conductivity, soil moisture, biomass). Segmentation of polarimetric SAR data into field sub-units can be achieved with eCognition software. These sub-units are representative of landscape features (e.g., water), soil texture variation and within-field variability over the growing season.
Speaker Information: Joni Bugden, Agriculture & Agri-Food Canada, c/o Dr. Elizabeth PatteyEastern Cereal & Oil, Ottawa, ON K1A0C6; Phone: 613-759-1523; E-mail: [email protected]
Session Information: Tuesday, November 4, 2003, 12:55 PM-4:15 PM Presentation Start: 3:00 PM
Keywords: remote sensing; management zones; polarimetric SAR; methodology