The AIKO cloudy_CHARLES is an onboard imagery processing software for cloud segmentation. With CHARLES, the quality of data is assessed directly onboard and only the relevant frames are sent to the ground.
The AIKO cloudy_CHARLES is a software tool that enables onboard data processing of Earth Observation (EO) data in the visible spectrum. Using state-of-the-art Machine Learning (ML) techniques, cloudy_CHARLES identifies clouds in the image frames soon after they are acquired.
The cloudy_CHARLES helps in discarding the data onboard spacecraft that are deemed unprofitable or unusable due to the presence of heavy cloud coverage. This saves valuable bandwidth by avoiding the downlink of poor-quality data, or by prioritizing the downlink of frames that are less affected by the presence of clouds.
Since cloudy_CHARLES applies a mission-agnostic segmentation technique to identify clouds in the images, it is possible to train its models to analyze data acquired by any optical payload operating in the visible range. When choosing cloudy_CHARLES to augment the capabilities of a customer’s EO satellite, AIKO engineers will support you throughout the model training and customization process, ensuring a perfect fit with the specific optical payload used.
The ML algorithms that power cloudy_CHARLES have been successfully tested on CPUs (x86_64 and ARM-64, with TFLite and ONNX runtimes), GPUs (Nvidia Jetson Nano), VPUs (Intel Myriad), and TPUs (Google Coral).
While cloudy_CHARLES is a self-contained tool that can be deployed onboard to provide simple but efficient data filtering, it can be also installed in combination with orbital_OLIVER, AIKO’s software solution for onboard autonomy. When coupled with orbital_OLIVER, the cloud coverage information is not just used to sift out the low-quality data, but it provides real-time insights that can be used to optimally replan the observation schedule according to the meteorological conditions.
cloudy_CHARLES has been extensively validated on the ground on Sentinel data. The core machine learning models reached TRL9 in 2019, during the deployment on a Low-Earth Orbit EO mission by Satellogic.
Including the product customization phase, the lead time for the deployment of cloudy_CHARLES is about two months.
Disclaimer: satsearch is not responsible for any mistakes on this page, although we do our best to ensure correctness. Please report any mistakes to us.
Last updated: 2022-07-26