Outdoor Illegal Parking Detection System Using Convolutional Neural Network on Raspberry Pi
This paper proposes a cost-effective vision-based outdoor illegal parking detection system, iConvPark, to automatize the detection of illegally parked vehicles by providing real-time notification regarding the occurrences and locations of illegal parking cases, thereby improving effectiveness of parking rules and regulations enforcement. The iConvPark is implemented on a Raspberry Pi with the use of Convolutional Neural Network as the classifier to identify illegally parked vehicles based on live parking lot image retrieved via an IP camera. The system has been implemented at a university parking lot to detect illegal parking events. Evaluation results show that our proposed system is capable of detecting illegally parked vehicles with precision rate of 1.00 and recall rate of 0.94, implying that the detection is robust against changes in light intensity and the presence of shadow effects under different weather conditions, attributed to the superiority offered by CNN.