Publication

Abstract : Weed detection is a crucial challenge in modern agriculture because weeds reduce crop output, resource consumption, and farm productivity. The use of broadspectrum herbicides and other conventional weed management techniques results in herbicide-resistant weeds, environmental harm, and higher expenses. Precision agriculture has advanced recently, with a focus on increasing sustainability and efficiency. Nevertheless, aerial detection has trouble reliably differentiating weeds from crops, and existing detection techniques, particularly ground-based systems, have trouble covering huge farms. This project will be making use of Diminished Reality (DR) technology to detect weeds, which allows for accurate herbicide delivery by highlighting weeds and hiding crops. This project’s methodology plans on processing live aerial images of farmlands by combining deep learning models with DR technology. Precision herbicide spraying requires low latency processing, which the system’s performance will guarantee. The outcomes will show how well the system detects and separates weeds from crops. By focusing primarily on weed-infested areas, the suggested strategy aims to decrease the use of herbicides, limit environmental damage, and increase crop output. The goal of this study is to demonstrate an advancement over existing weed-detecting techniques.