Deep Learning for Super-Resolution of Time Series Satellite Images

We propose to use deep learning techniques to enhance the spatial resolution of time series satellite images. This improvement, also called “super-resolution” is becoming essential to compensate for relatively low resolution sensors on resource constrained environments such as SmallSats and CubeSats. Software approaches are increasingly considered in connection with smaller satellites for which size and power constraints limit the capabilities of the sensors. Recently, deep learning techniques have been used successfully for achieving super-resolution of single hyperspectral images; we are generalizing this approach to time series satellite multispectral or hyperspectral images.

Higher resolution satellite imaging data is often desirable or required for the understanding of the science or process being observed. There are however resource constraints that may limit the sensor capability such as size, cost, power, or limited transmission bandwidth. The trade-offs are not likely to change for future missions as the spatial resolution needs will continue to increase as fast as new sensor technology. The field of machine learning, and in particular deep learning, has the potential to mitigate trade-offs and drive significant advances in the processing of all types of science data.

This is early/mid-stage (TRL 2) — the most realistic path in is NASA SBIR/STTR, which funds small businesses and research institutions to develop technology aligned with NASA's needs (equity-free, phased funding). Check whether a current SBIR/STTR solicitation topic overlaps with this project's technology area, or contact the project directly (above) to ask.

None of these are guaranteed paths for this specific project — TechPort itself doesn't have an "apply" button. Reaching out to the contact(s) above with a specific question is usually the fastest way to find out what's actually open.