Application of Vision-Language Model to Pedestrians Behavior and Scene Understanding in Autonomous Driving