Road geometry perception without accurate positioning and lane information

Road geometry perception indicates road direction and drivable area for Lane Keeping System Departure Warning in Advanced Driver Assistance Systems or routing decision-making Autonomous Driving. High-definition map with accurate positioning lane information is generally required. However, these two types of are not always available because failure mark degradation. In this paper, a estimation method proposed without requirement information. Due to the fact guides front vehicle, potential correlation between relative azimuth vehicle explored. The stable observable vehicles exploited, such as license plate location. A radar-camera fusion system using image processing, nonlinear then built Unscented Kalman Filter estimate curvature. Simulation results Carla demonstrate that enhances by 77.1%, compared baseline method. can provide reliable autonomous driving when unavailable, so it renders more robust safer future driving.