Modernization of the Automated Control Approach for Crop Development Indicators Using UAV Photography

The paper describes the results of the modernization of the approach to determining indicators of crop development using computer vision. A comparative analysis of the algorithm using a bundle of YOLOv5 neural networks for detection and FPN for semantic segmentation and an algorithm based on the YOLOv8 architecture for instance segmentation is carried out. High accuracy was achieved (F1 of 0.993 and <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$m A P &gt; 0.85$</tex>). To expand the capabilities of the system, the training data sets were supplemented with sugar beet.