Files and links (1)
CC BY V4.0, Open Access
Dissertation
Speed up the Precision Agriculture with Artificial Intelligence Empowered Phenomics
Washington State University
Doctor of Philosophy (PhD), Washington State University
2023
DOI:
https://doi.org/10.7273/000006306
Abstract
Timely and large-scale monitoring of crop behavior in response to biotic and abiotic stresses is critical for crop breeding and management. One example is the Alfalfa biomass breeding project. As the most widely cultivated forage legume, improvements in Alfalfa genetics have been limited for complex economically traits such as biomass. One of the major bottlenecks is the labor-intensive phenotyping burden for biomass selection. Another example is the fight against wheat stripe rust. As one of the most devastating diseases of wheat, development of resistant cultivars and application of fungicides are major approaches to controlling stripe rust. However, visual scoring is time-consuming, labor-intensive, and subject to human errors. Unmanned aerial vehicles (UAVs) provide a high-throughput option with flexibility in revisiting times and can capture high-resolution images, which makes them widely used in observing crop behavior in various crop fields. Analyzing these images highlights the urgent need for efficient pipelines and algorithms. Deep Learning (DL) has the potential to process images and videos captured from UAVs to enable high-throughput phenotyping for timely monitoring. However, applying DL technologies in image classification and object detection is still faced with obstacles, including image preprocessing, and labeling as well as model development and deployment. In this thesis, the first chapter is a literature review. The second chapter presents a pipeline for estimating alfalfa biomass with UAV-based multispectral imagery. The third and fourth chapters each demonstrate an affordable high-throughput field detection method using UAV for detecting wheat stripe rust, and a label image software, ROOSTER, to support this method. The fifth chapter releases a software called Ladder, which supports efficient labeling of training datasets, training the object detection models, and deploying models for new images. The last chapter is an advanced high throughput assessment of wheat stripe rust severity at single-row precision. Overall, three pipeline and software tools are introduced for applying DL and computer vision in UAV-based imagery, enhancing in-field phenotyping efficiency for crop breeding and management.
Metrics
6 File views/ downloads
53 Record Views
Details
- Title
- Speed up the Precision Agriculture with Artificial Intelligence Empowered Phenomics
- Creators
- Zhou Tang
- Contributors
- Zhiwu Zhang (Advisor)Xianming Chen (Committee Member)Xianran Li (Committee Member)Michael O. Pumphrey (Committee Member)Sindhuja Sankaran (Committee Member)
- Awarding Institution
- Washington State University
- Academic Unit
- Department of Crop and Soil Sciences
- Theses and Dissertations
- Doctor of Philosophy (PhD), Washington State University
- Publisher
- Washington State University
- Number of pages
- 190
- Identifiers
- 99901086721801842
- Language
- English
- Resource Type
- Dissertation