You can check how to use the Finder mode or the single shot search mode in the documentation:
It is basically calling a different method to trigger the search. The single shot search, will just send one image in a query and result in a single scan. The finder mode, sends one query every 500ms + request round trip, depending on the network bandwidth and latency.
Regarding the scans. We count as a visual scan every image that is sent to the Recognition API. If the query is successful, then we count it as a recognition.