Development of an Unmanned Aerial Vehicle incorporating a Convolutional Neural Network for Weed Detection

Abstract

This paper introduces the development of an  Unmanned Aerial Vehicle (UAV) as an efficient  asset that incorporates a high-performing pre trained architecture for weed classification and  detection in real-time. The UAV includes a  modular sprayer and embedded systems. Weeds provide a continuous threat to the cultivation of  agricultural plants and reduce yield, as well as increase waste. To address the issue, an  autonomous system and intelligent sprayer are described to spot weeds or undesirable plants and  spray herbicides on desired places. The  effectiveness of five various pre-trained  architectures (AlexNet, GoogleNet, VGG19,  DenseNet201, and ResNet50) for image  classification and prediction are investigated, and  the top performing model was determined. The  best-performing trained model, ResNet 50,  achieved a training accuracy of 98.57 percent and  a validation accuracy of 100 percent at 25 epochs.

https://doi.org/10.59200/ICONIC.2022.012