En cours (2024‑2025)

CE PROJET EST FINANCÉ DANS LE CADRE DU PROGRAMME « PROJETS ÉMERGENTS DU RQRAD »

Reducing the use of herbicides is crucial for advancing environmental sustainability and aligning with consumer preferences for more sustainable agricultural practices. In this project, we propose an innovative aerial-ground collaborative mobile system for precision spraying. This system utilizes UAV technology to capture high-resolution images of entire fields. Leveraging superresolution techniques alongside weakly- and self-supervised learning methods, we aim to develop a deep learning network capable of accurately segmenting crops and weeds within both aerial and ground-level images. Then, the resulting weed map, complete with GPS coordinates, guides a ground-based vehicle platform to target weed-infested areas. Subsequently, an onboard vision system employs the developed weed-crop segmentation network to identify weeds from images collected at ground level. The pixel-level segmentation results serve as input for controlling individual nozzles, enabling precise and targeted herbicide spraying. Field experiments will be conducted to assess herbicide spraying saving rate compared to conventional uniform spraying methods.