| Abstract: | A color-based visual technique is described based on the mean shift image segmentation method providingrelevant information for robust localization of the visible road area in Urban Intelligent TransportationSystem (U-ITS). The traffic image sequences are firstly trained to extract the background and thensegmented into separated parts by the mean shift method as initialization, regions with the number ofpixels not less than a threshold and with more uniform surfaces with the "same" color compared to theirenvironment are filtered as recognized road area. The algorithm given in this paper can present road arearecognition with arbitrary shapes, which is fit for unstructured road applications in urban cities very well. |
