A Machine Vision Detection of Unauthorized On-Street Roadside Parking in Restricted Zone: An Experimental Simulated Barangay-Environment

Abstract

The study developed a cost-effective framework for unauthorized parking detection using a machine-vision based deep learning method. The system was introduced on a Raspberry Pi 4b using the MobileNet SSD algorithm to detect vehicles illegally parked based on the live feed received from a Pi camera. The system was introduced to monitor unauthorized parking on a specific barangay simulated-roadside-parking lot. Results of the assessment indicate that the study was capable of identifying illegally parked vehicles with an overall performance rate of 96.16% and 98.93% respectively for legally and illegally parked vehicles, with a combined test resulting in 97.56%. The study showed that the detection was robust to changes in light intensity and the presence of shadow effects in varying environmental conditions, due to the deep learning strength.