Repository logo
 

Automatic Counting of Canola Flowers from In-Field Time-Lapse Images

Date

2018-05-17

Journal Title

Journal ISSN

Volume Title

Publisher

ORCID

0000-0002-9610-6664

Type

Thesis

Degree Level

Masters

Abstract

The combination of plant phenotyping and computer techniques has gained popularity amongst breeders and computer scientists. The recent evolution of the latter has allowed High-Throughput Phenotyping (HTP) to play a significant role in filling the genotype-to-phenotype gap. While most of the related work in HTP is performed in controlled environments, such as greenhouses, that allow automatic devices to capture the data reliably, research in in-field phenotyping is not as robust due to environmental confounds (i.e., fog or sun-reflections). The usage of high temporal density data has not been exploited to the same degree as high spatial resolution information. However, many phenotypes (e.g., canola flowering) have a temporal component. In this document, we present an image-processing-based method that attempts to detect and count flowers of canola during the early flowering stage on in-field time-lapse images. This approach can be used to analyze the evolution of the flower density of canola plants over short periods of time during the first days of flowering thanks to the availability of high temporal resolution images. We used images extracted during Summer 2016 to generate ground truth, tune the flower detection method and count the flowers during the first days of the flowering period. We provide an overview and a discussion about additional steps that might be needed to overcome the impact of sunlight reflection on canola leaves in the detection of flowers.

Description

Keywords

canola flower detection, image processing, high-throughput phenotyping

Citation

Degree

Master of Science (M.Sc.)

Department

Computer Science

Program

Computer Science

Part Of

item.page.relation.ispartofseries

DOI

item.page.identifier.pmid

item.page.identifier.pmcid