Imaging Science Ph.D. Defense: Advancing the Utility of Unmanned Aerial Systems-Based (UAS) Imaging Techniques in Broadacre Agriculture: A Multimodal Case Study on Table Beets

Imaging Science Ph.D. Defense
Advancing the Utility of Unmanned Aerial Systems-Based (UAS) Imaging Techniques in Broadacre Agriculture: A Multimodal Case Study on Table Beets

Mohammad Shahriar Saif
Imaging Science Ph.D. Candidate
Rochester Institute of Technology                                                  

Register for Zoom here

Abstract:
Efficient and sustainable agricultural production systems are becoming increasingly critical to meet global food demands. However, novel approaches are required to address challenges related to rapid, large-area monitoring, both of which put a strain on current labor-intensive practices. Unmanned aerial systems (UAS), offering high-throughput, spatially-explicit data acquisition, are gaining traction for agricultural monitoring in this context. Subterranean specialty crops, such as table beet, present additional unique challenges and remain underexplored, given their priority insofar as global staple crops are concerned. Furthermore, there is a need to evaluate the efficacy of different sensor configurations for predicting key field agronomic traits. To address these gaps, we conducted flight campaigns over two growing seasons (2021 / 2022) at Cornell AgriTech, Geneva, NY, collecting five-band multispectral imagery (475, 560, 668, 717, and 840 nm), hyperspectral imagery (400–1000 nm), and light detection and ranging (LiDAR) data. These datasets were utilized to predict harvest root yield (R²test = 0.81; RMSEtest = 0.58 kg m⁻²), in situ disease severity (R²test = 0.82; RMSEtest = 9.31%), and plant population density (R²test = 0.77; RMSEtest = 29%). Our findings demonstrate that lightweight multispectral sensors can offer a practical alternative to heavier, more complex sensing systems. Additionally, we demonstrate that advanced modeling approaches can facilitate end-of-season yield prediction, in situ disease severity assessment, and plant counting even at relatively coarse spatial resolutions, paving the way for the commercial viability of UAS-based solutions for subterranean crop management.

Intended Audience:
All are Welcome!

To request an interpreter, please visit myaccess.rit.edu