This research is concentrated on vision-based localization under an environment with monotonous characteristics. Initially, the mobile robot wanders the unknown environment using GVG algorithm and laser scanner. During the initial wandering, a Topological map containing image database generated according to the robot map standardization document.

The generated image map is used for position recognition of the mobile robot or human. When the image of a random location is recognized, the geometric information of the image is extracted from the comparison of line histogram and vanishing point of the input image and image database. From the result of image processing, the candidates of exact positions are recommended from the database. The exact position among the candidates is extracted by the particle filter. This algorithm is applied to android smartphone prototype application