Geological Layer Detection and Candidate Science Target Identification

Geologic layers provide valuable information about the history of a planetary region. We have developed an approach to identifying candidate science targets within layered geologic deposits onboard an in-situ spacecraft, such as the Mars Exploration Rover (MER) Mission or the Mars Science Laboratory (MSL) Mission. The approach includes automated detection of layers in a scene and an algorithm for selecting science targets for further study that sample across the span of identified layers. Identifying targets in representative observed layers enables rapid collection of targeted remote sensing data on valuable layer targets without requiring multiple ground communication cycles. This type of automated image processing and target selection could enable increasingly informed, onboard decisions about when and where to take follow-up measurements and could easily be integrated into existing flight software that prioritizes science targets for follow-up measurements. The approach is divided into two elements. First, the presence of layers in a scene is ascertained. To determine the presence and location of layers within the image, the statistical properties of image regions are used. A supervised (trained on known examples) and unsupervised method have been developed. Performance assessment criteria include detection and false alarm rates, as well as computational requirements and run time. Upon detecting the presence of layering, candidate science targets are efficiently selected and prioritized to enable surveying the layers. The approach involves estimating the slope of the layering, selecting a transect perpendicular to the direction of the layers and then determining relatively homogeneous regions along the transect. We show results on images collected by the MER Mission rovers using both Pancam and Navcam imagery as well as scenes from a field experiment in the Mojave Desert.