Files in this item

FilesDescriptionFormat

application/pdf

application/pdfCHOUDHURI-THESIS-2019.pdf (13MB)Restricted to U of Illinois
(no description provided)PDF

Description

Title:Autonomous robotic system for high throughput plant phenotyping (width estimation) in agricultural fields
Author(s):Choudhuri, Anwesa
Advisor(s):Chowdhary, Girish
Department / Program:Aerospace Engineering
Discipline:Aerospace Engineering
Degree Granting Institution:University of Illinois at Urbana-Champaign
Degree:M.S.
Genre:Thesis
Subject(s):Robotics, Deep Learning, Computer Vision, Agriculture
Abstract:We present an autonomous robotic system for the estimation of crop stem width in highly cluttered and variable agricultural fields. Stem width is an important phenotype (observable trait) needed by breeders and plant-biologists to measure plant growth. However, its manual measurement is cumbersome, inaccurate, and inefficient. There is an immense need to automate such a task in order to increase the productivity of plants in future. The presented system aims to achieve this goal. It navigates autonomously through every row of an agricultural field under the plant canopy. This navigation is based on deep optical flow on videos collected by the robot and lane estimates from a low cost LiDAR sensor. The phenotyping is performed using deep learning or a sequence of image processing steps to eliminate background. Width is estimated based on the robot velocity from wheel encoders, and validated by the lane estimates from the LiDAR. This system has been tested and exhaustively validated against available hand-measurements on biomass sorghum (Sorghum bicolor) in real experimental fields. Experiments indicate that this system is also effective for other kinds of crops, like corn. The width estimation match on sorghum is 93.5% (using optical flow) and 92.38% (using LiDAR lane estimates) when compared against manual measurements by trained agronomists. Thus, our results clearly establish the feasibility of using small robots for stem-width estimation under the canopy in realistic field settings. Furthermore, the techniques presented here can be utilized for automating other important phenotypic measurements.
Issue Date:2019-07-19
Type:Text
URI:http://hdl.handle.net/2142/105844
Rights Information:Copyright 2019 Anwesa Choudhuri
Date Available in IDEALS:2019-11-26
Date Deposited:2019-08


This item appears in the following Collection(s)

Item Statistics