This paper describes an approach in order to detect the entrance of buildings for navigation of robot. We focus on the problem of entrance detection using visual information. Entrances not only have a variety of color but also are placed in various environments. We tried to detect the entrance that is opened and closed or blocked by cars and so on with a variety of colors of doors. We adopt the fuzzy system to analyze features whether they belong to entrance then we apply region growing algorithm to the candidates of entrance to cover the various environment and colors of entrance. In the building, entrance is the main components with wall region and windows. We think that the regions which are not wall and windows are considered to the entrance candidates. From windows, we can easily limit the candidates of entrance and detect entrance as employing the height of a floor or number of floors. Before entrance detection, we detect the building surface and extract wall region and windows within the building surface. After the process of window detection, we learned approach method from a few training examples with additional assumptions about structure characteristics of entrance. The approach is tested on numerous examples and it shows a good performance.