Data Mining during Rover Traverse: from Images to Geologic Signatures

Soon robotic explorers will be able to produce more scientific data than can be transmitted or interpreted efficiently. We present a method for characterizing geology during rover traverse using autonomous data analysis techniques. The strategy detects discrete geologic features in images; distributions of these features constitute a signature that correlates with the geology of each site. These numerical profiles reveal subtle trends and boundaries in geologic units that facilitate targeted sample selection and efficient data analysis. We demonstrate the system’s use on field data collected during a field expedition to the Atacama Desert of Chile.