Patrol robots to perform specific service in environment travel the same path repeatedly based on map. For the safe navigation of patrol robots, environment recognition and precise localization should be necessary. In outdoor environment, it is generally using expensive equipment and many sensors for environmental recognition and localization. However, complex system configuration and increase in costs of robot are required. Furthermore, in urban road environment, the precision of GPS is reduced by high structures. In order to solve these problems, we propose localization algorithm using the curb feature. An onboard LRF is used to extract curb. The extracted curb is matched with line map and fused by Extend Kalman Filter with odometry and DGPS measurement. We verify the robustness of the proposed algorithm by experiment, although the large DGPS error occurs due to the buildings.