Development of unified and dynamic geometric framework for modelling plant leaf spots

Loading...
Thumbnail Image
Other Title
Authors
Alshadli, Duaa
Borna, Kambiz
Lador, Cesar
Author ORCID Profiles (clickable)
Degree
Grantor
Date
2020-11
Supervisors
Type
Conference Contribution - Oral Presentation
Ngā Upoko Tukutuku (Māori subject headings)
Keyword
leaf spots
plant disease
dynamic geometry
image object
vector agents
image processing
modelling
ANZSRC Field of Research Code (2020)
Citation
Alshadli, D., Borna, K., & Lador, C. (2020, November). Development of Unified and Dynamic Geometric Framework for Modelling Plant Leaf Spots. Paper presented at the 2020 Symposium on Pattern Recognition and Applications (SPRA 2020), Rome, Italy.
Abstract
The use of computer vision and image processing techniques have proven to be effective in detecting and classifying plant diseases from symptoms, such as leaf spots or mosaic leaf patterns caused by pathogens. To identify these symptoms, these techniques typically utilize a static geometry specified by a human expert via pixels or image objects. Thus the results rely on generic parameters defined by the user before or after classification. In this paper, a dynamic geometry is proposed that can be applied to identify plant disease symptoms without setting any geometric parameters. The offered method consists of two primary phases established based on the notion of Vector Agents (VAs): construction of a unified geometry, and creation of a dynamic geometry. In the construction step, the method utilises a set of geometric rules to link the raster space to the vector space during the simulation process. These rules specify the growth direction of a leaf in the simulation space and the maximum length of each edge specified based on the image spatial resolution and size. The creation step includes four main geometric operators: 1) vertex displacement, 2) half-edge joining, 3) converging vertex displacement, and 4) edge remove. This enables the leaves to automatically change their geometry in the simulation space without setting any geometric parameters. This structure allows a classifier to model a wide range of leaf shapes in the image space. The proposed geometry method was tested to model different leaves and the results demonstrate its effectiveness in generating different leaf shapes.
Publisher
Link to ePress publication
DOI
Copyright holder
Authors
Copyright notice
All rights reserved
Copyright license
Available online at
This item appears in: