Cornell University
Object-Level Signed Distance Functions and their Utility for 3D Phenotyping
Abstract
dc:description.abstractRepresenting plant morphology is crucial for plant phenotyping, as it is a bottleneck in improving the efficiency of breeding programs, understanding plant-environment interactions, and managing agricultural systems. However, capturing accurate 3D reconstructions of plants poses significant technical challenges. Nevertheless, recent advancements in implicit neural representations offer promising learning-based methods for reconstructing precise 3D representations of crops in the field. In this thesis, I provide a comprehensive review of 3D reconstruction techniques and their utility in creating photorealistic representations for 3D plant phenotyping. We propose a technique called ObjectCarver which decomposes a scene composed of multiple objects into the constituent geometries via separate signed distance fields. Lastly, I demonstrate the usage of signed distance functions in 3D plant phenotyping, extracting traits like leaf length, leaf width, and leaf area on three varietals of hemp plants.
Degree
thesis:*- Name thesis:degree_name
- M.S., Computer Science
- Level thesis:degree_level
- Master of Science
- Discipline thesis:degree_discipline
- Computer Science
- Grantor
- Cornell University
- Year dc:date.issued
- 2024
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Moon, Hyun
- Chair dc:contributor.chair
-
- Petersen, Kirstin
- Committee members dc:contributor.committeemember
-
- Bhattacharjee, Tapomayukh
- Jiang, Yu
Rights
dc:rights- Statement dc:rights
-
- Attribution 4.0 International
- Licence dc:rights.uri
- Language dc:language.iso
- en
Identifiers
dc:identifier.*- Dc Identifier Bibid
- 16611723
- Dc Identifier Other
-
ProQuest Submission ID: 12268
ProQuest Publication ID: 31489575