This pilot project was carried out for Natural Resources Canada. Geolocation identified all individual trees and hedges, as well as residential, single-family and multi-family houses, from the “Building” class data provided by the client. As part of this mandate, we also had to acquire annotations (training data or ground truth).
The work was carried out using RGBI optical satellite images at a 50 cm resolution.
The production of these training datasets was used to develop and use supervised machine learning techniques, statistical techniques in which the machine develops a model based on this training data.
For further information about our services in cartography, aerial imagery or LiDAR surveys, please contact us at 418-688-3308 or info@geolocation.ca.