-
- GC32D-02: Harnessing Machine Learning and Super-Resolution Sentinel-2 Imagery to Accurately Map the Weed Kochia
-
NOLA CC
Author(s):Generic 'disconnected' Message
Steven Shirtliffe, University of Saskatchewan (First Author, Presenting Author)
Thuan Ha, University of Saskatchewan
Kwabena Nketia, University of Saskatchewan
Japiyot Sandhu, University of Saskatchewan
Sarah van Steenbergen, University of Saskatchewan
Hansanee Fernando, University of Saskatchewan
Kochia is a fast-spreading weed that reduces crop yield and is almost impossible to control because it resists many herbicides. However, it usually occurs in dense patches which may allow it to be mapped with remote sensing. While drones can map kochia accurately, they only cover small areas. This study used advanced satellite images and computer models to track kochia across large parts of Saskatchewan from 2020 to 2024. The satellite data was enhanced with a super-resolution procedure to increase the ground resolution from 10m to 1m. The results were validated using field visits and high-resolution images from drones and dedicated high-resolution satellites. The system was built using Google Earth Engine, so it can be used anywhere in the Canadian Prairies. New tools were also created to help tell kochia apart from other plants. The method was very accurate—over 98%—and could help farmers and researchers monitor and manage kochia using patch based methods.
Scientific DisciplineSuggested ItinerariesNeighborhoodType
Enter Note
Go to previous page in this tab
Session
