Files in this item

FilesDescriptionFormat

application/pdf

application/pdf3131058.pdf (16MB)Restricted to U of Illinois
(no description provided)PDF

Description

Title:Hyperspectral Imagery for Precision Agriculture
Author(s):Yao, Haibo
Doctoral Committee Chair(s):Lei Tian
Department / Program:Agricultural Engineering
Discipline:Agricultural Engineering
Degree Granting Institution:University of Illinois at Urbana-Champaign
Degree:Ph.D.
Genre:Dissertation
Subject(s):Engineering, Agricultural
Abstract:The main contribution of this dissertation to hyperspectral image processing is that a hybrid feature selection and feature extraction approach was proposed and implemented for image feature reduction. The general procedure was to select a subset of the original image bands and transform the image band subset to a new image space. The band subset can be regarded as the significant image bands for a given application. The first implementation of the hybrid feature reduction approach was the GA-SPCA algorithm (Genetic Algorithm-based Selective Principal Component Analysis). In the initial experiments, the coefficients of correlation between an image and ground reference data were greater when using GA-SPCA than using standard PCA. To investigate the GA-SPCA algorithm further, a spatial-spectral feature extraction procedure was developed for soil nutrient classification, using a maximum likelihood classifier (SPCA+ML). In this case, feature extraction was based on class separability, measured using class distance. The second implementation of the hybrid feature reduction approach was the development of an automatic vegetation index generation algorithm---EAVI (Evolutionary Algorithm based Vegetation Index generation). The EAVI algorithm was applied to two precision farming applications---finding the temporal relationship between images and corn yield and estimating corn kernel grain quality. It was found that images from the beginning of full canopy coverage to the corn ear formation period provided the best and most stable results for corn yield estimation. The significant image bands were in the red edge region at 700 nm and in the NIR (near infrared) region at 826 nm.
Issue Date:2004
Type:Text
Language:English
Description:250 p.
Thesis (Ph.D.)--University of Illinois at Urbana-Champaign, 2004.
URI:http://hdl.handle.net/2142/86056
Other Identifier(s):(MiAaPQ)AAI3131058
Date Available in IDEALS:2015-09-28
Date Deposited:2004


This item appears in the following Collection(s)

Item Statistics