AI for cloud detection
One of Φsat-2’s AI applications can automatically detect and discard any of the images acquired by the multispectral imager where clouds have obscured the view of Earth.
As a matter of fact, this application is far more capable then just being able to detect clouds in an image – it is able to classify them and provide insight into cloud distribution. This will give users more flexibility when it is time to deciding whether an image is actually usable or not, and the thresholds can be easily changed on request.
These specific capabilities are initially verified onboard as part of the application fine tuning phase and then run during the normal operations part of the mission.
What makes the Φsat-2 implementation of the cloud detection unique is its configuration as a service to support other onboard applications.
Cloud detection can be requested as part of the initial image pre-processing step and its output will represent the new input for the next in-line AI application. This will guarantee even better use of the limited onboard resources.
This application is developed by KP Labs in Poland.
