Development of a UAV-Based Multi-Dimensional Mapping Framework for Precise Frost Management in Apple Orchards
Open Access
- Author:
- Yuan, Wenan
- Graduate Program:
- Agricultural and Biological Engineering
- Degree:
- Doctor of Philosophy
- Document Type:
- Dissertation
- Date of Defense:
- January 25, 2022
- Committee Members:
- Long He, Major Field Member
Daeun Choi, Chair & Dissertation Advisor
Dimitrios Bolkas, Outside Unit & Field Member
Paul Heinemann, Major Field Member
Suat Irmak, Program Head/Chair - Keywords:
- camera
CNN
image registration
LiDAR
object detection
point cloud
RGB
temperature
thermal
camera
CNN
image registration
LiDAR
object detection
point cloud
RGB
temperature
thermal - Abstract:
- As one of the main causes of weather-related damages in agriculture, frost leads to significant economic losses for farmers worldwide. Yet, traditional frost protection and temperature assessment methods in orchards remain rudimentary. Unmanned aerial vehicles (UAVs) and airborne sensing instruments emerged in recent years as promising tools for assisting efficient and convenient crop monitoring and management in precision agriculture, which creates opportunities in revolutionizing orchard frost protection approaches. With an overarching goal of building an autonomous cyber-physical system (CPS) consisting of UAV-based sensing and unmanned ground vehicle (UGV)-based heating for precise frost management, in this dissertation, a multi-dimensional mapping framework was proposed to process UAV-based thermal imagery, RGB imagery, and light detection and ranging (LiDAR) point cloud data to extract growth stage, canopy temperature, and tree structural information of an apple orchard. A thermal image stitching algorithm was developed to create high-resolution orchard temperature maps. A convolutional neural network (CNN)-based classifier was developed for detecting apple flower buds in RGB images, whose robustness against artificial image distortions and training dataset attributes were also investigated in depth. A UAV-LiDAR system was developed for identifying orchard regions that were unsafe for UGV travelling. The final output of the mapping framework, georeferenced orchard navigation maps, indicated both orchard heating requirements and orchard open space regions, which can potentially serve as a guide for UGV path planning and heat treatment application during frost events in future studies.