Towards a Biosignatures Image Detection System for Planetary Exploration with UAVs

The search for life beyond Earth can benefit from orbiters and spacecraft with compact instruments able to identify potential biological signatures. One of the main challenges is the balance between a lower resolution and a wide field of view to discard uninteresting places while a high resolution - narrow field of view to collect data in higher detail. The recent flights of the Mars helicopter “Ingenuity” have shown UAVs are a viable platform to explore the surface of celestial objects in a wide and narrow approach using diverse remote sensing instruments. With data collected from real biosignatures in Western Australia, this work proposes an online UAV-based Artificial Intelligence detector using Convolutional Neural Networks (CNN) based on ResNet18 and YOLO models able to detect multiple potential biological signatures in near real-time. The system and pipeline presented allow the inclusion of new observations refined by scientists to increase the scientific exploration outcomes for remote-based operations.