ROAD AND SKY SEGMENTATION ALGORITHM USING GEOCONTEXT MODELING METHOD AND ADAPTIVE PATTERN MATCHING IN VEHICLE-BASED GEOVISION SYSTEM

摘要

PURPOSE: A GeoContext modeling in a vehicle based GeoVision and a partitioning method of a road and sky by using an adaptive pattern matching are provided to stably partition an image regardless of an adaptive change of a composition pattern of a road and sky depending on a current environment. CONSTITUTION: A color image is partitioned by a Mean Shift algorithm(002). An adjacent region list is constructed(003). An initial region is merged by the use of color information(004). A sky region and a road region are recognized by the comparison between a static pattern of database and a currently partitioned region(005). An adaptive pattern is extracted. An extracted adaptive pattern is compared to an adjacent region(006). The sky and road region is finally partitioned(007).