Research On Detection, Recognition And Mapping Of Road Marking

Road environment is the carrier of urban traffic. Many areas, such as vehicle navigation, intelligent transportation, and image maps, need to expend the perception of the road environment, in order to know traffic and understand traffic. In this demand, the paper researches the information and proposes range of approaches based on car system to increase the obtained information, including road lane recognition, road sign recognition and road mapping.In order to detect and identify all kinds of road lanes, the paper presents the approach of lane identification based on rotation histogram. The approach uses hough transform to get the main direction of road lane and uses histogram analysis to get the color, lines and dotted information. Experiment shows that this method can detect and identify all kinds of traffic lane information with high accuracy rate.In order to detect and identify road sign information with rotation invariance and scale invariance, this paper presents the approach based on hu moment invariants and projection histogram. The approach uses the 7 invariants as the main information, and uses the histogram statistics on the direction of contour segments. Experiment shows that the methods can handle many kinds of road sign with high accuracy rate.In order to expend the definition of road orthophoto map, the paper proposes the approach using the road-vehicle system. It uses vehicle-mounted camera to get road image information and uses position device to get vehicle location information. Then it uses inverse perspective transformation to produce road orthophoto fragment and uses offline interpolation algorithm to handle the position information and splice the fragments together. At last it uses wavelet transform and weighted median algorithm to fuse the fragments to a complete one. Experiment shows the approach can get road map with high-definition.