| 摘要 |
PROBLEM TO BE SOLVED: To improve the accuracy of lane recognition by sorting out valid edge points based on at least one of coordinate deviation and angle deviation between an edge point of a lane and a road model. SOLUTION: Edge points with a density change of a threshold or more are detected within an image of a traveling road (step S5), and the detected edge points are compared with a generated road model, thereby sorting valid edge points (step S7), and the road model is updated based on the sorted valid edge point (step S8). In the sorting of valid edge point, a straight line perpendicular to the road model through each edge point is drawn to determine distance d from the edge point to a crossing point P with the road model, and a one with the distance d smaller than a coordinate threshold Dth is sorted as the valid edge point. Otherwise, an angle θd between a tangent vector of the road model and a vector of each edge point at the crossing point P is determined, and one with the angle θd smaller than an angle threshold θth is sorted as the valid edge point. COPYRIGHT: (C)2010,JPO&INPIT
|
