Graduation Semester and Year
Fall 2025
Language
English
Document Type
Dissertation
Degree Name
Doctor of Philosophy in Computer Science
Department
Computer Science and Engineering
First Advisor
Dr. William J. Beksi
Abstract
Advancements in computer vision, facilitated by the proliferation of sensing technologies and increased computational power, have resulted in major innovations across many fields. Agricultural automation has notably benefited from these developments. However, the uncontrolled and dynamic characteristics of agricultural environments limits the effectiveness of conventional computer vision techniques. To overcome these obstacles, this dissertation develops the following novel computer vision methods to advance infield agricultural automation. First, a motion-model-based algorithm for tracking and counting cotton bolls via video sequences is presented. Next, an appearancebased re-identification technique is introduced to tackle the challenge of crop self-similarity and improve multi-object tracking performance across various crop types. Then, a neural radiance field-based 3D segmentation approach is created to address issues such as occlusion and incorrect associations in image- and video-based tracking, enabling accurate crop segmentation and counting. Finally, to meet the demand for efficient storage and transmission of large-scale 3D data, a learning-based point cloud compression architecture is designed to operate on raw 3D point clouds across a wide range of bitrates using a single trained model. All the proposed methods are evaluated on multiple crop varieties (e.g., apples, pears, and cotton) and demonstrate state-ofthe-art performance. Furthermore, this dissertation creates the first publicly available cotton-tracking dataset, TextCot22, to facilitate future research.
Keywords
Compute vision, Machine learning, Precision agriculture, Object tracking, Counting, Data compression
Disciplines
Artificial Intelligence and Robotics | Computer Sciences
License
This work is licensed under a Creative Commons Attribution 4.0 International License.
Recommended Citation
Muzaddid, Md Ahmed Al, "Computer Vision Methods for Advancing Infield Agricultural Automation" (2025). Computer Science and Engineering Dissertations. 427.
https://mavmatrix.uta.edu/cse_dissertations/427