Localizing Active Objects from Egocentric Vision with Symbolic World Knowledge