Use of uncrewed aerial vehicles to determine plant stand density

Abstract

Recently, the use of uncrewed aerial vehicles (UAV) in various branches of human activity has become more widespread and successful. Agriculture is no exception as the use of UAVs makes it possible to increase the economic efficiency of crop production. The research goal is to demonstrate the possibility of using middle price uncrewed aerial vehicles to control the quality of sowing, in particular the plant density. The plant density is one of the important indices of plant growing quality. Its accuracy may determine the plans for further plant care measures, yield and economic efficiency of crop production. It is proposed to use the DJI Air2S quadcopter-type UAV to photograph the fields. The sequence of setting up and using software for marking up photos, followed by training the YOLO11 neural network model is described. The YOLO11 model is used to solve the problem of determining the number of plants in a photo. The sequence of direct use of the trained YOLO11 neural network model for calculating the number of sunflower plants and weeds in an image is also described. Just eight images of the field surface with sunflower seedlings were used to train the neural network. With such a small number of photos, the determining accuracy of sunflower plants by the neural network model was 76%. This indicator is a fairly good result. But for real use, the accuracy of the neural network needs to be improved by further training on more photos. According to the experiment, the problem of correctly determining and counting the number of sunflower plants in photographs taken from a height of 5 m is quite solvable using the YOLO11 model.