Approaches and challenges in machine learning for monitoring agricultural products and predicting plant physiological responses to biotic and abiotic stresses

Abstract

The world's population and the subsequent demand for food are increasing at an unprecedented rate, presenting significant challenges to sustainable food production. The impact of abiotic and biotic stresses on agricultural productivity is one of the major obstacles threatening food security. As a potential solution to these challenges, advancements in machine learning (ML) and deep learning (DL) based systems analyzing have emerged as promising solutions for improving crop yields, as well as mitigating plant stresses with high accuracy and efficiency. Furthermore, the increasing availability of sensor technologies and communication networks in the agriculture sector has led to the widespread adoption of ML for yield prediction and plant phenotyping, particularly on a large scale. The application of ML in conjunction with high-throughput imaging and genomic data is examined for early detection of physiological stress indicators and acceleration of crop improvement programs. This review highlights the latest technologies and approaches that are currently employed in ML and DL to effectively detect biotic and abiotic plant stresses. Despite notable progress, limitations persist in areas such as data quality, model generalization across agro-ecological zones, and field-level deployment. Emerging directions—including automated ML (AutoML), quantum machine learning, and digital twin technologies—are discussed as promising solutions for advancing precision agriculture and enhancing crop resilience under changing climatic conditions. These cutting-edge technologies have the potential to significantly enhance the sustainable production of food by efficient crop management and address the challenges posed by the growing global population and climate change, while mitigating the impacts of environmental and biotic stressors on crop production.