TY - GEN
T1 - Bok Choy Growth Monitoring Using IoT Technology and a Recurrent Segmentation Model
AU - Kang, Chenchen
AU - Mu, Xinyang
AU - Seffrin, Aline Novaski
AU - Di Gioia, Francesco
AU - He, Long
N1 - Publisher Copyright:
© 2024 ASABE Annual International Meeting. All rights reserved.
PY - 2024
Y1 - 2024
N2 - Hydroponics is a subset of hydroculture for growing plants without soil but by using water-based mineral nutrient solutions in a controlled environment. Traditionally, crop monitoring in hydroponics is carried out manually, which makes it unfeasible to frequently collect data to represent the dynamics of plant growth during the growing period. This study developed an integrated IoT and computer vision system specifically tailored for Controlled Environment Agriculture (CEA) systems, enabling continuous monitoring and analysis of plant growth. The core innovation of this research is the implementation of a recurrent image segmentation model that processes sequential image data to accurately track temporal changes in plant development. This vision system is supported by an IoT framework designed to capture high-resolution imagery at predetermined temporal frequencies. Applied to bok choy, this integrated system successfully segments individual plants and tracks leaf coverage area throughout their growth cycle. The quantitative analysis of Intersection over Union (IoU) scores among the segmentation approaches shows that the recurrent model begins with an IoU score of 0.986 during the early growth stages and maintains a robust performance, achieving a score of 0.893 at later stages. The findings demonstrate that this approach significantly enhances the precision of monitoring practices and facilitates effective decision-making for nutrient management and growth optimization in CEA systems.
AB - Hydroponics is a subset of hydroculture for growing plants without soil but by using water-based mineral nutrient solutions in a controlled environment. Traditionally, crop monitoring in hydroponics is carried out manually, which makes it unfeasible to frequently collect data to represent the dynamics of plant growth during the growing period. This study developed an integrated IoT and computer vision system specifically tailored for Controlled Environment Agriculture (CEA) systems, enabling continuous monitoring and analysis of plant growth. The core innovation of this research is the implementation of a recurrent image segmentation model that processes sequential image data to accurately track temporal changes in plant development. This vision system is supported by an IoT framework designed to capture high-resolution imagery at predetermined temporal frequencies. Applied to bok choy, this integrated system successfully segments individual plants and tracks leaf coverage area throughout their growth cycle. The quantitative analysis of Intersection over Union (IoU) scores among the segmentation approaches shows that the recurrent model begins with an IoU score of 0.986 during the early growth stages and maintains a robust performance, achieving a score of 0.893 at later stages. The findings demonstrate that this approach significantly enhances the precision of monitoring practices and facilitates effective decision-making for nutrient management and growth optimization in CEA systems.
UR - http://www.scopus.com/inward/record.url?scp=85206094082&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85206094082&partnerID=8YFLogxK
U2 - 10.13031/aim.202401198
DO - 10.13031/aim.202401198
M3 - Conference contribution
AN - SCOPUS:85206094082
T3 - 2024 ASABE Annual International Meeting
BT - 2024 ASABE Annual International Meeting
PB - American Society of Agricultural and Biological Engineers
T2 - 2024 American Society of Agricultural and Biological Engineers Annual International Meeting, ASABE 2024
Y2 - 28 July 2024 through 31 July 2024
ER -