How is computer vision applied in autonomous driving, and what are the key challenges in this domain?