Leopard: A new chapter in on-board deep learning-powered analysis of hyperspectral imagery

Hyperspectral imaging has become a mature technology which brings exciting possibilities in various Earth observation applications in a plethora of fields, including precision agriculture, forestry, event detection and tracking, and more. However, high dimensionality is an important obstacle in its effective transfer from a satellite back to Earth for further processing. To tackle it and enable faster adoption of the hyperspectral technology in practice, the on-board data processing has become critical, as it allows to substantially reduce the data dimensionality before transferring it, and to extract value from raw data. In this paper, we present Leopard - a CubeSat standard compliant Data Processing Unit (DPU) which enables mission designers to apply Artificial Intelligence solutions in space.