Negative obstacle sensing is one of the most difficult problems for unmanned ground vehicle in unstructured environments. The regular obstacle sensors, such as stereo vision, infrared detector and ladar, have their limited performances in unconstructed environments. Ultrawideband SAR (synthetic aperture radar) sensors have the ability to operate in all weather, all lighting and foliage covered conditions, which have been received widely. Sensing negative obstacle by ultrawideband SAR for unmanned ground vehicle was an effective way. Image geometry of negative obstacle was expounded. The simulation image of negative obstacle was obtained by simulation based on MATLAB, and the conclusion that the image feature of negative obstacle is the shadow area next to shine area is obtained. Moreover, a real data experiment is presented and the experimental result proves the same conclusion again.