It is very necessary for Autonomous Land Vehicle (ALV) to hove a reliable abiliiy of obstacle detection. Unlike srructural highway, environmems, where only vehicles ahead of the ALV are needed to consider, small obstacles are more ofien seen in semi-structural environments. In this paper we described a small obstacle detection qsrem for ALV under such ein.iroiiinenrs. The system consists of hvo 2D Laser Range Finders (LRE) which are responsible for obstacle recognizing and tracking respectively. For iop LRF. a D-S evidence riieon, based ousiacle recognition algorithm is used to distringuish the obstacles from rhe brushwood on the roadsides. Based on that. an obstacle tracking process ir sraricd bj, uillsing the data from borrom LRF. Besides the obstacles, the distribution of the brushwood on roadsides, which could be used,ior path planning. is also obtained. The experimenral results demonstrate our success.