Files and links (1)
CC BY-NC V4.0, Open Access
Dissertation
PHENOMICS-DATA DRIVEN TOOLS FOR MACHINE LEARNING-ASSISTED DECISION SUPPORT IN AGRICULTURE
Washington State University
Doctor of Philosophy (PhD), Washington State University
01/2022
DOI:
https://doi.org/10.7273/000004433
Handle:
https://hdl.handle.net/2376/119017
Abstract
The advancements in sensor technologies – in omics research in agriculture – have increased the amount of generated data. One example of omics data is phenomics, the science of measuring phenotypes (i.e., crop and animal phenotype). Phenomics data are widely used in crop improvement research, precision crop, and livestock farming, allowing rapid, objective, and accurate measurement of phenotypes. In this dissertation, phenomics-data-driven tools were explored and developed to improve and optimize decision support in diverse agricultural applications.Aphanomyces root rot (ARR) disease resistance is an important trait to evaluate in pulse crops. Two image-based disease severity classification approaches were assessed to distinguish between three classes of ARR severity: resistant, intermediate, and susceptible, using Red-Green-Blue images of lentil roots. One approach was through hand-crafted features extracted from the images that were used as an input dataset for a generalized linear model with elastic net regularization. The second approach entailed building a convolutional neural network as an end-to-end feature learning and classification approach. The resistant class was accurately classified with both methods, with an accuracy of 0.92-0.96, while the former approach performed better for the other two classes. In another study, the scalability of field-based remote sensing using high-resolution satellite imagery and the synergy with unmanned aerial systems (UAS) sensing were evaluated to estimate the yield of field pea at the breeding plot level. The assessment was performed using feature fusion (satellite and UAS data) and image fusion (pan-sharpening of satellite data using UAS data). We found that the performance of satellite-imagery data-based random forest models was comparable to UAS-imagery data-based models at a later growth stage (after canopy closure). In the final study, health summary measures were developed to inform dairy cattle’s health status and disease severity. An approach to derive these metrics – objectively and using pathophysiological data – was established to evaluate four common diseases (ketosis, hypocalcemia, metritis, mastitis) in dairy cattle. The approach was found to show potential application in indicating health status based on comparisons with disability weight class indicators reported in the literature.
Metrics
3 File views/ downloads
20 Record Views
Details
- Title
- PHENOMICS-DATA DRIVEN TOOLS FOR MACHINE LEARNING-ASSISTED DECISION SUPPORT IN AGRICULTURE
- Creators
- Afef Marzougui
- Contributors
- Sindhuja Sankaran (Advisor)Lav R. Khot (Committee Member)Rebecca J. McGee (Committee Member)Weidong Chen (Committee Member)Brian H. Clowers (Committee Member)
- Awarding Institution
- Washington State University
- Academic Unit
- Department of Biological Systems Engineering
- Theses and Dissertations
- Doctor of Philosophy (PhD), Washington State University
- Publisher
- Washington State University
- Number of pages
- 165
- Identifiers
- 99900883436501842
- Language
- English
- Resource Type
- Dissertation