Deep learning based plant segmentation from RGB images

Lameski, Petre and Zdravevski, Eftim and Kulakov, Andrea and Trajkovik, Vladimir (2017) Deep learning based plant segmentation from RGB images. In: PROCEEDINGS of the 14th Conference on Informatics and Information Technology. Faculty of Computer Science and Engineering, Ss. Cyril and Methodius University in Skopje, Macedonia, Skopje, Macedonia, pp. 49-51. ISBN 978-608-4699-07-1

[img]
Preview
Text
978-608-4699-07-1_pp49-51.pdf

Download (416kB) | Preview
Official URL: http://ciit.finki.ukim.mk

Abstract

Plant-ground segmentation is one of the most important steps in the process of plant classification. Plant segmentation is used for the process of weed detection in fields, vegetation coverage estimation from satellite images, detection of illness in plants, water level detection in plants, etc. There are quite a few approaches presented in the literature that use different types of color indexes, segmentation and classification techniques for the purpose of weed segmentation. In this paper we apply the SegNet deep learning architecture for plant weed segmentation. The images are taken from approximately 1m height under slightly varying lightning condition from the same field using a smart phone regular RGB camera with auto-focus. The presented results show that it is possible to successfully train a deep learning model based on a single image and obtain similar results to the best plant segmentation techniques available that use color indexes without expensive cameras.

Item Type: Book Section
Subjects: International Conference on Informatics and Information Technologies > Multimedia
?? CIIT_ SP ??
Depositing User: Vangel Ajanovski
Date Deposited: 29 Nov 2017 18:33
Last Modified: 29 Nov 2017 18:33
URI: http://eprints.finki.ukim.mk/id/eprint/11373

Actions (login required)

View Item View Item