This paper describes a localization technique that permits mobile robots to accurately determine their positions and even to respond to unforeseen obstacles such as pedestrians. The effectiveness of the method is confirmed in trials using a model robot. Generally, robots are taught the layout of operational environment in order to localize self-position. This is normally done through the generation of an environmental map that is created by running the robot through every accessible area of the environment beforehand. This also means that such robots are unable to localize their positions in unknown environments. In this research, we propose a method that allows robots to perform self-position localization in unknown environments through the use of low-resolution photographs available through online services such as Google Maps. This research involves the application of a space observation model to a framework of self-position presumptions that do not rely on the extraction of strict edge information from aerial photographs. The validity of the proposed technique was verified through a presumptive self-position experiment conducted in an unknown environment using a map created from an aerial photograph.