Safe spacecraft landing on planetary and small body surfaces is of primary concern. Estimation of landing risk is a critical task when evaluating and certifying potential landing sites. Such analyses require the detection and mapping of all potential landing hazards such as rocks and boulders, craters, slopes, and terrain roughness.
The availability of very high-resolution orbital imagery from the HiRISE camera onboard the Mars Reconnaissance Orbiter, and from the Narrow Angle Camera (NAC) onboard the Lunar Reconnaissance Orbiter, makes it possible to detect and map boulders automatically. Computer vision techniques have been applied to implement algorithms that detect rocks from images. At NAC nominal resolution, it is possible to fully resolve boulders 2.5 m or larger, and detect boulders as small as 1.2 m. With knowledge of the Sun angles, the shadows detected from the images are used to generate rock models describing the location, width, and height of boulders. A mapping toolkit computes statistical plots and maps of rock density and abundance.
These algorithms are tailored to the lunar environment, including two major adaptations: automatic handling of shadow contrast variations and image blur, and a capability to distinguish small fresh craters (which cast shadows as well) from large boulders. The algorithms run on desktop computers and can process a NAC image in 30 minutes with multiple diagnostics output.
All current and future landed missions would benefit from detailed analyses of potential landing sites, provided orbital imagery with sub-meter resolution is available. If not available, hazard and avoidance systems are needed. The rock detection algorithms described were originally designed for such a task and would be readily applicable.
This work was done by Andres Huertas and Yang Cheng of Caltech for NASA’s Jet Propulsion Laboratory.