Utilizing convolutional neural network (CNN) for orchard irrigation decision-making

Abstract

Efficient agricultural management often relies on farmers' experiential knowledge and demands considerable labor, particularly in regions with challenging terrains. To reduce these burdens, the adoption of smart technologies has garnered increasing attention. This study proposes a convolutional neural network (CNN)-based model as a decision-support tool for smart irrigation in orchard systems, focusing on persimmon cultivation in mountainous regions. Soil moisture data and corresponding leaf RGB images were collected over two growing seasons to train, test, and validate the CNN model for determining optimal irrigation timing. The model achieved over 80% accuracy in identifying water stress levels in persimmon trees based on leaf images. These findings indicate the potential of the developed model as a key component of a remote irrigation system. However, the model's performance limitations and challenges in adapting to diverse field conditions underscore the need for further research to enhance its robustness and applicability.

Keywords: Automated irrigation; Image processing; Machine learning; Persimmons; Remote sensing; Water stress.