There are several ways to do it.
OnBecameVisible and OnBecameInvisible functions in any script are called when game object becomes visible/invisible to any camera.
OnWillRenderObject (new in Unity 2.0) is called on any script when object will be rendered this frame for some camera. Inside this function, Camera.current is the camera that will render the object. If object will be rendered by several cameras, OnWillRenderObject will be called multiple times.
And finally, GeometryUtility class allows you to test visibility manually (by doing bounding box vs. camera frustum intersection test).
Unity does not do occlusion culling, i.e. it renders everything in camera's frustum. Of course there's nothing stopping you from doing some sort of occlusion culling yourself, targeted at your specific game.
But if you want to test occlusion, you'll have to do it yourself somehow. Occlusion is a complex topic though...