Abstract
To develop a practical indoor mobile robot system, invisible marks made of photochromic material are provided to the floor of an indoor environment. Invisible mark is usually colorless and transparent: however, once irradiated by ultraviolet rays from the robot side, it turns red (or green). Thus, by doing color binarization, this mark is easily extracted. By using these floor marks, the area where object exists can be easily extracted, by detecting the occluded marks. In the following image processing, this area is used for template matching based on feature such as SIFT, finally the kind of object can be recognized. Moreover, information profitable for the object recognition can be obtained from how to hide the mark.