Abstract
This paper proposes the self-localization method by an image evaluation which matches a present captured image to a field image captured previously. In conventional researches, the self-localization methods by using landmarks such as goal post or lines were proposed. However, in these methods, self-localization accuracy changes with places referred to as the landmarks. In order to reduce such accuracy change, we propose the self-localization method which uses the Higher-order Local Autocorrelation function (HLAC). HLAC has features of the location invariance and is easy to calculate. HLAC has been frequently used for object recognition in various fields. The experimental results show that an image evaluation using HLUC is available for self-localization in RoboCup.
