-
- B33I-1991: High-Resolution Inland Wetland Vegetation Classification with a Foundational CNN Framework Using WorldView-2
-
Board 1991‚ Hall EFG (Poster Hall)NOLA CC
Author(s):Generic 'disconnected' Message
Fiona Benzi, Florida Atlantic University (First Author, Presenting Author)
Caiyun Zhang, Florida Atlantic University
Inland wetlands are essential ecosystems that help manage floods, improve water quality, and support biodiversity. However, it’s difficult to accurately map and classify the different plant communities found in wetlands because of how similar they can look in satellite images. We developed an AI model that uses satellite images to better identify and classify wetland vegetation in Central Florida. The model was trained using high-resolution satellite imagery and expert-labeled maps from managed wetland areas in the Upper St. Johns River Basin. It works in two steps: first, it groups vegetation into broad types, and then it uses sub-models designed to classify specific plant communities. The model uses both image patterns and carefully selected indicies to solve common problems like look-alike vegetation, overlapping species between communities, and shadows. Early results show the model can support long-term monitoring and mapping of wetlands. Once established, the model can be applied repeatedly to detect shifts over time, making it a valuable tool for understanding ecological change and guiding land management.
Scientific DisciplineSuggested ItinerariesNeighborhoodType
Enter Note
Go to previous page in this tab
Session
