Dissertation/Thesis Abstract

Geospatial Technology Applications to Strawberry, Grape and Citrus Production Systems
by Saraswat, Dharmendra, Ph.D., The Ohio State University, 2007, 301; 10835955
Abstract (Summary)

The objective of this study was to investigate the spatial and temporal variability in strawberry and grape production systems using aerial digital multispectral imagery, along with reference ground-truth data. Another objective was to identify and extract citrus trees from QuickBird satellite image for obtaining tree count.

Two images were acquired for strawberries in 2002 to study differences in crop growth. Two images for grapes were acquired, one year apart in 2004 and 2005 around the same crop growth stage (veraison). The ground-truth for both the crops included: 1) soil and plant information, 2) apparent soil electrical conductivity data, and 3) yield data. A QuickBird satellite image was acquired in 2004 to identify and extract citrus trees and to determine tree count. The reference tree count was acquired through a 30 cm spatial resolution digital ortho quadrangle quad (DOQQ).

The aerial and satellite images were geometrically corrected, reprojected and subsetted to obtain images of the study site. The images used for studying grape production were converted to a reflectance image using empirical line approach. The digital count value contained in the QuickBird image was converted to spectral radiance values before extracting citrus trees. The relationships between image data and reference ground-truth were determined for strawberries and grapes. Different pan sharpening methods were tried to enhance the spatial details on multispectral QuickBird satellite image. A methodology was developed and tested for extracting citrus trees from pan sharpened QuickBird image using commercially available image object analysis software.

The relationships between soil-plant variables and image data were found to be dependent on the growth stage in strawberries. The best linear and non-linear model estimated 42% and 56% of strawberry yields, respectively. Vegetation indices (NDVI, RVI and SAVI) derived from digital multispectral aerial image data provided a better estimate of leaf area index (r2 = 0.81, 0.78, and 0.58, respectively) than yield of grapes (r2 = 0.14, 0.15, and 0.11, respectively). The best net and gross accuracy of tree count was found to be 78.1.6% and 62.1% on QuickBird satellite image pan sharpened by Gram-Schmidt method.

Indexing (document details)
Advisor: Brown, Larry
School: The Ohio State University
Department: Food, Agricultural, and Biological Engineering
School Location: United States -- Ohio
Source: DAI-B 79/10(E), Dissertation Abstracts International
Subjects: Bioengineering
Keywords: Ec, Geospatial, Gis, Gps, Remote sensing, Yield variability
Publication Number: 10835955
ISBN: 978-0-355-97094-4
Copyright © 2020 ProQuest LLC. All rights reserved. Terms and Conditions Privacy Policy Cookie Policy