In Kajitori, we constructed an inexpensive people counting system using a network camera for the browsing room of the library. In using the system for a year, the counting of visitors to the browsing room has been done by a human because the number of visitors in a day is at most 150 or so. But now we have gotten enough data of the operation of the system to consider the automation of the counting. For the automation of the counting, we do not adopt a versatile approach but a simple one which is necessary and sufficient for our needs in the library. The result is that the method we developed is accurate enough except it cannot distinguish the library people.