An autonomous mobile robot needs to always obtain a self-position while it moves towards a target position. Conventional methods based upon recognition of landmarks by using image processing have risk of misrecognition by disturbance. It is difficult that the robot obtain the self-position precisely by using external area sensors such as GPS (Global Positioning System), ultrasonic range sensor, and so on. In this paper, we propose an estimation algorithm which obtains the self-position by using a ground image. We developed an application program based on the proposed algorithm, and verified a feasibility of the proposed method.