New methods for mars rover localization based on integration of ground and orbital imagery

Localization of the Mars Exploration Rovers (MER) is currently realized with their odometers, Inertial Measurement Units (IMU), and navigation cameras (Navcam) and panoramic cameras (Pancam). Location errors are inevitably accumulated in long-range rover traverse by only using these ground sensors. Two new rover localization methods are presented based on integration of ground (rover) and orbital imagery. These methods can localize the rover in orbital imagery, thus removing the accumulated localization errors and enhancing the security and effectiveness of rover operations. The first method extracts rocks from ground and high resolution orbital images, matches rock distribution patterns, and consequently localizes the rover in the orbital image. The second method generates an orthophoto from ground images and matches the orthophoto and the orbital image with scale-invariant feature transform (SIFT). The proposed methods are verified using Navcam images from multiple positions of Spirit rover and Opportunity rover and HiRISE orbital images that cover the two landing sites. The results show that the first method works in the regions where there are many rocks on the Mars surface, while the second method is suitable for the areas where there are outcrops in the soil. The accuracy of these localization methods is better than one pixel of HiRISE image, which is 0.25m. Copyright © (2009) by the Asian Association on Remote Sensing.