Control of lighting and camera positioning will be extremely helpful in this (outdoor) application. In order to acheive reliable face dection you're going to need to normalize the view and the lighting conditions under daylight, night, weather, etc.
The recognition problem is related, but a different challenge. Remember, recognition will never be 100% accurate. How many times do you as a human, with one of the most sophisticated pattern recognition machines in the planet in your head, have trouble recognizing somebody you know in person or in pictures?
You're going to need to specify your acceptable false positive/negative rates, as well as the system architecture with with all of your non functional requirements (performance, mounting and lighting, etc).