Mohanad Albughdadi
Google Cloud ML Engineer
Artificial Intelligence
Earth Observation
Python Expert
Field Boundary Delineation in Sentinel-2 Images
A DL based approach for field
boundary delineation
Precision agriculture services aim at optimizing the management of the crop and increasing in plot profitability through different biophysical and vegetation parameters. One of the first steps in order to provide these maps is to define parcels’ boundaries. Nonetheless, generating these boundaries manually is a non-trivial task especially when the number of parcels is large.
Recent advances in machine learning and more specifically deep learning have fostered the computer vision field and its application. Deep learning algorithms allow machines to understand images, extract their features and hence perform various computer vision tasks such as semantic and instance segmentation as well as object detection.
Super Resolved Sentinel-2 Images
A DL based single image super resolution approach
Sentinel-2 images are characterized by a fine spectral resolution. However, for some precision agriculture applications their spatial resolution is not sufficient enough especially for vineyards. The super resolution method aims at enhancing the spatial resolution of Sentinel-2 images based on deep learning algorithms and very high resolution satellite images.
Detection of Fire Risk Zones Using BDORTHO Images
DL and ML applied to the detection of forest areas and typology classification
The objective of this use case is to update the forest fire hazard zoning in multiple French departments. Two layers are produced. The first layer is a binary mask to delimit the forest areas. From the detected forest mask, a typology mask will be produced to classify different forest types. This layer is important for fire hazard maps.
Massive LAI Estimation from Sentinel-2
DL approach for biophysical parameters fast estimation
LAI provides an absolute quantification of the biomass of vegetation allowing an overview of the development status of a plant. The estimation of this parameter requires sophisticated and complex algorithms that are computationally intensive. A deep learning method was developed to estimate this parameter over ile-de-france and occitanie for multiple years. The proposed method takes around 18 seconds to estimate the LAI for a Sentinel-2 tile compared to 15 minutes using the SNAP software developed by ESA.
Detection of Swimming Pools Using BDORTHO Images
DL applied to the detection of swimming pools in multiple French communes
The objective of this use case is to provide a database with the geolocalization of swimming pools in multiple French communes for circular economy puroposes.