Date of Award

Fall 2020

Project Type

Thesis

Program or Major

Electrical and Computer Engineering

Degree Name

Master of Science

First Advisor

Richard A Messner

Second Advisor

Wayne Smith

Third Advisor

John LaCourse

Abstract

Use of image segmentation has caused agriculture advancement in species identification, chlorophyll measurements, plant growth and disease detection. Most methods require some level of manual segmentation as autonomous image segmentation is a difficult task. Methods with the highest segmentation precision use a priori knowledge obtained from user input which is time consuming and subjective. This research focuses on providing current segmentation methods a pre-processing model that autonomously extracts an internal and external contour of the leaf. The model converts the uniform Cartesian images to non-uniformly sampled images in log polar space. A recursive path following algorithm was designed to map out the leaf’s edge boundary. This boundary is shifted inward and outward to create two contours; one that lies within the foreground and one within the background. The image database consists of 918 leaves from multiple plants and different background mediums. The model successfully created contours for 714 of the leaves. Results of the autonomously created contours being used in lieu of user-input contours for a current segmentation algorithm are presented.

Download

Share

COinS