Automatically extracting rocks from Martian surface images is valuable for hazard avoidance and rover localization in rover missions. To extract rocks, data field on image is proposed to model the interaction between two pixels of an image under the imagery characteristics of Martian surface. Foreground rocks are firstly grouped into clusters by locating the centers and edges in data field after they are differed from background information via binarizing partition. Then target rocks are discovered for the Rover to keep healthy paths. The experiment shows the data field on image is practical and potential.
展开▼
