Predicting carob tree physiological parameters under different irrigation systems using Random Forest and Planet satellite images.

Introduction: In the context of climate change, monitoring the spatial and temporal variability of plant physiological parameters has become increasingly important. Remote spectral imaging and GIS software have shown effectiveness in mapping field variability. Additionally, the application of machine learning techniques, essential for processing large data volumes, has seen a significant rise in agricultural applications. This research was focused on carob tree, a drought-resistant tree crop spread through the Mediterranean basin. The study aimed to develop robust models to predict the net assimilation and stomatal conductance of carob trees and to use these models to analyze seasonal variability and the impact of different irrigation systems. Methods: Planet satellite images were acquired on the day of field data measurement. The reflectance values of Planet spectral bands were used as predictors to develop the models. The study employed the Random Forest modeling approach, and its performances were compared with that of traditional multiple linear regression. Results and discussion: The findings reveal that Random Forest, utilizing Planet spectral bands as predictors, achieved high accuracy in predicting net assimilation (R² = 0.81) and stomatal conductance (R² = 0.70), with the yellow and red spectral regions being particularly influential. Furthermore, the research indicates no significant difference in intrinsic water use efficiency between the various irrigation systems and rainfed conditions. This work highlighted the potential of combining satellite remote sensing and machine learning in precision agriculture, with the goal of the efficient monitoring of physiological parameters.

In-Field Automatic Identification of Pomegranates Using a Farmer Robot.

Ground vehicles equipped with vision-based perception systems can provide a rich source of information for precision agriculture tasks in orchards, including fruit detection and counting, phenotyping, plant growth and health monitoring. This paper presents a semi-supervised deep learning framework for automatic pomegranate detection using a farmer robot equipped with a consumer-grade camera. In contrast to standard deep-learning methods that require time-consuming and labor-intensive image labeling, the proposed system relies on a novel multi-stage transfer learning approach, whereby a pre-trained network is fine-tuned for the target task using images of fruits in controlled conditions, and then it is progressively extended to more complex scenarios towards accurate and efficient segmentation of field images. Results of experimental tests, performed in a commercial pomegranate orchard in southern Italy, are presented using the DeepLabv3+ (Resnet18) architecture, and they are compared with those that were obtained based on conventional manual image annotation. The proposed framework allows for accurate segmentation results, achieving an F1-score of 86.42% and IoU of 97.94%, while relieving the burden of manual labeling.