For the IVS to work, the camera needs time to see the object and process it and decide if it matches the criteria. For LPR and tighter zooms for a camera, that can be problematic for some field of views. The car is in the field of view for such a short period of time that it will probably struggle more times than not. Especially at night if all it sees are head/taillights and a plate it will miss most of them.
I am usually a fan of the Dahua AI, but for LPR this is an unique situation that you would be better off with motion detection and a min object size instead of trying the vehicle AI. You will obviously probably get some false triggers, but it beats the alternative of IVS missing a lot of them.