Workshop Description: Segmentation is the process of partitioning a digital image into multiple homogeneous regions by grouping pixels based on similarity in terms of intensity, texture, or color. Segmentation is an indispensable prerequisite for computing plant phenotypes. The quality of segmentation is critical to accurate phenotype computation and subsequent phenotype-genotype mapping. The workshop will have two components: (A) a presentation of segmentation techniques used for plant phenotyping in both controlled and field-based phenotyping platforms; and (B) an interactive session to introduce a software tool named iPlantSeg+ to facilitate image segmentation and compute a set of generic phenotypes.
A.The presentation will include segmenting foreground from images captured by visible light cameras as well as less explored imaging modalities, e.g., infrared, near-infrared, fluorescent, and hyperspectral. It will demonstrate techniques to segment the whole plant or its structures for holistic (e.g., plant height, plant aspect ratio) and component phenotyping (e.g., leaf length, leaf curvature, stem angle). A set of traditional segmentation methods (e.g., frame differencing, color thresholding, spectral difference, graph-based) and a set of learning-based segmentation methods (e.g., clustering-based, neural network-based) will be presented using plant images.
B. This interactive session will introduce the iPlantSeg+ tool to compute phenotypes followed by segmentation. It provides a menu of both interactive and automated segmentation methods, thus allowing the users to select an algorithm and mode best suited for a particular application. It performs accurate segmentation of plants with thin structures in the presence of cluttered backgrounds and stores the phenotypes in a database.
**Please indicate interest in participating in this session via your registration record. All those that indicate interest during registration will receive a zoom link for the virtual workshop no later than July 28th from pbabstracts@aspb.org
Learning Objectives:
Upon completion, participants will be able to gain a detailed understanding of traditional and learning-based segmentation techniques used for plant phenotyping in both controlled and field-based phenotyping platforms.
Upon completion, participants will be able to conduct accurate segmentation of plants with thin structures in the presence of cluttered backgrounds and illumination variation using their own image datasets.
Upon completion, participants will be able to compute generic plant phenotypes using a GUI-based tool and store them in a database without an understanding of the underlying programming details.