2009 年 129 巻 11 号 p. 2074-2082
This paper describes a reliable method developed for a view-based navigation which uses approximate planes of image pixel brightness in which edge-based information has been modified with a least-square algorithm. The system developed in this work have shown not only the robustness of this method to ambient lighting conditions due to indoor illumination or daylight from windows, but also the high capability of its noise rejection for images taken in recording runs. This method has been found to be highly tolerated in larger ambient illumination change than the used conventional method where brightness-based pixel values are utilized. Based on this scheme, the experiment has been successfully made in the navigation test even if large ambient lighting condition was given to the robot by means of the change in the number of the luminescent lamps or in the night and the day.