Autonomous Detection and Removal of Paddy Weeds using Computer Vision and Deep Learning

Abstract

Modern agriculture faces the critical challenge of efficiently managing weeds to ensure optimal crop yield and sustainability. Traditional weed management practices often fall short of addressing the complexities of diverse agricultural landscapes. This research aims to construct the weeder for the detection and removal of weeds unfold across two aspects: the design and development of the weeder and the construction of the deep learning model. The dataset is developed from the images of Monochoria vaginalis and Limnocharis flava weeds acquired from the paddy plots. YOLOv8, a renowned algorithm for real-time object detection, is employed to identify the weeds in paddy fields. The model is able to correctly predict weeds 99% of the time, and it can correctly predict non-weeds 100% of the time. The novelty of this research is the autonomous identification and removal of paddy weeds based on deep learning model operated in a custom built weeder robot