The main function of powered lower limb prosthesis is to help the amputees to achieve independent and comfortable walking. In order to make the powered lower limb prosthesis coordinate with the user, it is necessary to recognize the motion intention of the user. By knowing the environment information in front of the user, and taking it as a prior knowledge of motion intention recognition, the recognition accuracy of motion intention will be improved. In order to provide environment information in advance for the powered lower limb prosthesis, a wearable terrain recognition system was designed. The 2D lidar installed in the waist were used to collect the terrain data in front, then extract linear feature of the collected data using agglomerative hierarchical clustering algorithm, finally the terrain was recognized using finite automaton. In the experiment, four terrains were tested, including level ground to up/down stair and up/down slope. Results show that the recognition accuracy of the system to four terrains reaches to 95.8%, and it can obtain the information of terrain parameter including angle of slope, number of steps, stair height and stair depth which can′t be obtained from the traditional motion intention recognition methods. It proves that the system is effective and feasible for powered lower limb prosthesis.