System architecture for an intersection assistant fusing image, map, and GPS information

摘要

Street intersections account for a large portion of road accidents everywhere around the world. Therefore, a driver assistance system supporting the driver to negotiate intersections is highly desirable and has a tremendous potential of reducing the number of collisions at intersections. What makes intersection assistance more challenging than other traffic scenarios such as highways and rural roads? On one hand, the environment all around the vehicle must be monitored. A very large field of view for the local sensors is required. We decided to use both omnidirectional cameras for a panoramic view and active cameras to inspect smaller areas around us at a high resolution. On the other hand, intersections come in all shapes, sizes, and signal regulations. Hence, some prior knowledge of the signal mode and the intersection geometry is very helpful. This can be achieved via GPS in conjunction with an attributed digital map. We propose a system architecture for intersection assistance using image, map, and GPS information to issue the typical intersection warnings such as cross traffic warning or traffic light warning.