Precision Identification System for Sowing Germination Trays

Over the past decade, the implementation of artificial intelligence algorithms in computer vision has revolutionized the scope of autonomous systems in industry and daily life by standardizing detection and monitoring tasks. However, these models often require substantial computational power, which is frequently unavailable in small-scale agricultural production systems. This underscores the need to develop systems capable of interacting with complex environments while using minimal computational resources. The method proposed in this paper is part of the vision module of a seeder system for smallscale productions of sowing germination trays in greenhouses environment. The method generated is able to classify the plots of a germination tray and determinate the viability of each one for sowing only with a computer vision program which was developed with the OpenCV library, and is divided in three steps: Tray detection, sorted row and plot detection. The results demonstrated the method's effectiveness; however, some limitations were identified, such as the requirement for the germination tray to remain in a static position for some time, and issues with noise in dirty testing environments. The key contributions of this paper are the development of a low computational consumption computer vision program to identify viable plots in a germination tray and the generation of a semiautomatic algorithm for sowing germination trays. Additionally, the developed algorithms are capable of managing other robotic tasks such as transplanting and removal of non-viable seedlings.