I don't really believe visual processing at extreme speeds will be able to compare to human vision to the degree required anytime soon. Your other idea sounds good (I think that's basically what CBTC is), but the New York City subway, for example, has been upgrading to such a system for nearly three decades now, so I'm not optimistic about the possibility of a widespread application of that technology.