Convolutional Neural Network for Studying Plant Nutrient Deficiencies

Rishav Bose, Henrik Hautop Lund

Research output: Chapter in Book/Report/Conference proceedingArticle in proceedingsResearchpeer-review

37 Downloads (Pure)

Abstract

We discuss the development of a vision-based plant phenotyping system based on a novel type of robotic system called a food computer. The food computer used in this project is called the GrowBot. It has a host of sensors to help analyse the growth chamber including a Raspberry Pi camera. The project revolved around developing a system to segment the plant canopy from its background and analyse nutrient deficiencies from the images taken by the camera. The pilot project investigated how a segmentation model called U-Net could be used to study the images. One of the drawbacks of many existing vision-based plant phenotyping systems is that their convolutional neural networks (CNNs) were trained to analyse very ideal images of individual leaves. This pilot project tried to address that issue, while at the same time explored how to train the neural networks to learn segmentation from a small image dataset.
Original languageEnglish
Title of host publicationProceedings of International Conference on Artificial Life and Robotics
Volume27
PublisherALife Robotics Co, Ltd.
Publication date2022
Pages25-29
ISBN (Print)978-4-9908350-7-1
DOIs
Publication statusPublished - 2022
Event2022 International Conference on Artificial Life and Robotics - Virtual Event
Duration: 20 Jan 202223 Jan 2022

Conference

Conference2022 International Conference on Artificial Life and Robotics
LocationVirtual Event
Period20/01/202223/01/2022

Keywords

  • Convolutional Neural Networks
  • Cyber Agriculture
  • Food Computing
  • Image Segmentation
  • Plant Health Monitoring
  • U-Ne

Fingerprint

Dive into the research topics of 'Convolutional Neural Network for Studying Plant Nutrient Deficiencies'. Together they form a unique fingerprint.

Cite this