Doorplate adaptive detection and recognition for indoor mobile robot self-localization

Self-localization is key to mobile robot navigation. Among many techniques for self-localization, landmark-based approaches can avoid complex and often memory demanding descriptions of robot surroundings. In this paper, doorplate is selected as visual landmark. For color images captured by robot, adaptive thresholding segmentation based on component color histograms and rival penalized competitive learning (RPCL) are employed to locate doorplate position in these images. Figures within doorplate are detected by edge detection operator and recognized with their geometric features extracted from results by Hilditich thinning algorithm. Experiments on mobile robot CASIA-1 show that this method is effective for indoor mobile robot self-localization