Development of an automated and artificial intelligence assisted pressure chamber for stem water potential determination

• Published in: Computers and electronics in agriculture Vol. 222; p. 109016
• Main Authors: Mucchiani, Caio, Karydis, Konstantinos
• Format: Journal Article
• Language: English
• Published: Elsevier B.V 01.07.2024
• Subjects: Agricultural automation; Agricultural robotics; Artificial intelligence assisted visual perception; Stem water potential; Water stress; Agricultural robotics; Agricultural automation; Artificial intelligence assisted visual perception; Water stress; Stem water potential
• ISSN: 0168-1699; 1872-7107
• DOI: 10.1016/j.compag.2024.109016

The pressure chamber (or Scholander chamber) is widely adopted for determining stem water potential (which is linked to plant water status) due to its simplicity, relative portability, and capacity to enable direct measurements. The method also serves as a reference when validating and calibrating other techniques. Despite its significant utility, the current form of the pressure chamber method is very labor-intensive, resulting in infrequent and spatially sparse sampling. Furthermore, the typical use of a compressed gas (usually nitrogen) cylinder to build up the pressure inside the chamber can cause safety issues (e.g., projectiles caused by pressure) and practical concerns (e.g., gas cylinder changes that may increase measurement time). In addition, the determination of the instance xylem water appears can vary depending on the experience of the user. For these reasons, automation and artificial intelligence (AI) technologies can be integrated to improve the current standard of practice in determining stem water potential. This work presents the development and testing of an automated pressure chamber that leverages machine vision to help determine the status of xylem wetness, a critical step toward full autonomy in stem water potential determination. The work contributes both to pneumatic actuation, whereby an air compressor and on-board electronics are employed to make the chamber fully controllable via software, and to visual perception, whereby a miniature camera and on-board electronics are integrated to provide easily visible, accessible, and real-time video feed on the excised end of a leaf’s stem. Further, an AI-based object detection algorithm is deployed to determine the xylem’s wetness status automatically. Several experiments with in-situ data collection demonstrate the efficiency of our system under both (semi-)manual and automatic (AI-assisted) modes of operation, thus confirming that our method can help enhance the current standard-of-practice pressure chamber method to determine stem water potential in a faster, more affordable, accurate, and repeatable manner. •Development of an automated and artificial intelligence-assisted pressure chamber for determining stem water potential.•Integration of both manual and automatic modes of operation considering live data stream and post-measurement recordings.•Evaluation of the proposed system with in-situ data collection and sample analysis to validate its efficacy in avocado tree crops.