AUTHOR=Kirwan R. F. , Abbas F. , Atmosukarto I. , Loo A. W. Y. , Lim J. H. , Yeo S. 

TITLE=Scalable agritech growbox architecture

JOURNAL=Frontiers in the Internet of Things

VOLUME=Volume 2 - 2023

YEAR=2023

URL=https://www.frontiersin.org/journals/the-internet-of-things/articles/10.3389/friot.2023.1256163

DOI=10.3389/friot.2023.1256163

ISSN=2813-3110

ABSTRACT=Urban farming has gained significance in Singapore and presents opportunities for automation to enhance its efficiency and scalability. In collaboration with Archisen Pte Ltd (2021), a Singaporean urban farming company, this work presents an IoT-based approach for automated farming that incorporates an agnostic growbox and a web application dashboard for smart monitoring of crop growth. This approach offers an open-source and cost-effective solution for scalable urban farming architecture.The presented agnostic growbox system offers accessibility and scalability. Furthermore, the cost-effective and modular hardware components with open-source software make it customizable and more accessible than commercial growbox products. The authors expect that this approach will have broader applications within the urban farming space and facilitate easier and more efficient urban farming through automation.Additionally, the applied image analytics approach is scalable, time-efficient, and accurate for disease detection in urban farming. The detection of crop diseases presents a significant challenge in urban farming, as early detection can mitigate crop wastage. This paper proposes an integrated solution that includes an image analytics approach for efficient and accurate classification of crop disease phenotypes, specifically chlorosis and tip burn in lettuce crops. The authors' approach is hardware-and software-efficient, eliminating the need for large datasets of images to train models. The image analytics approach is compared favourably with a machine learning approach, evaluating the accuracy of categorization for the same dataset. The approach is also shown to maintain advantages over the time and cost involved with machine learning.