Monitoring the cumulative germination of tomato seeds using artificial intelligence

Abstract

Monitoring seed germination is a common task in research activities for various crops under different treatments. To determine the impact of a seed treatment, it is essential to conduct germination trials, which typically last around ten days. One of the main inputs obtained from these trials is the graphs of the germination rate per day. To evaluate the germination status, direct observation in the laboratory is necessary, verifying the number of germinated seeds throughout each trial. This visual inspection requires adequate training and is time-consuming, as well as prone to human error. This proposal focuses on an embedded system based on a Raspberry Pi 4, located inside a germinator, where it captures images periodically. Through image processing, primarily based on the use of custom-trained YOLO machine learning models, these images are analyzed, tracking each tray and the germination status of its seeds. Furthermore, the development of a local web server with a graphical interface is proposed, allowing scientists to audit and correct the system's results. This system would essentially transform the germinator into a smart germinator. This work is part of a research project evaluating the effect of magnetic fields on tomato crops, within the framework of the strategic research group on sustainable agri-food production at the Technological University of Uruguay.