Development of a Camera-Based Parking Monitoring System with an Automatic Parking Spot Identification

This paper presents the development of a camera-based sensor system for the real-time monitoring of an off-street parking area, whereby each parking spot of this area is separately observed. In contrast to commercially available ground sensors, the position of a spot has not to be predefined, as it is identified automatically after a few parking processes during the system setup phase. Therefore, the system is also suitable for parking areas with variable positions and sizes of spots, for instance on road edges (on-street parking) or unsurfaced off-street parking areas without colored marking. The algorithms of the system are efficiently designed for the specific purpose of parking monitoring, so no high-level hardware like a graphics processing unit (GPU) is required. Instead, the system components are a small single-board computer (Raspberry Pi 4 B) and an USB (Universal Serial Bus) cable camera with a standard resolution (at least 640 x 480 pixels) as well as a zoom lens and an infrared emitter for spot focusing and night vision. To evaluate the reliability of the developed monitoring system, a field test with two different ground sensors on the same parking spot was conducted for several weeks. The experimental results, which are presented in the end of this paper, show that the system works reliably in different situations and usually reacts faster than the proven ground sensors.