Integrated Smart Parking System Using Internet of Things (IoT) and Digital Image Processing with Faster Region Based Convolutional Neural Network (Faster R-CNN) and Optical Character Recognition (OCR)

Parking is an integral part of driving, especially in urban areas. However, inefficient parking systems often cause drivers to spend a significant amount of time just to find a parking spot, even up to 10 hours per month in major cities like Jakarta. Besides wasting time, prolonged parking searches also increase fuel consumption and CO2 emissions. This study aims to develop smart parking application integrated with Internet of Things and Computer Vision. The system uses ultrasonic sensors connected to an ESP32 microcontroller to detect real-time parking slot availability, sending status updates to Firebase. Additionally, Faster Region Based Convolutional Neural Network (Faster R-CNN) and Easy OCR technologies are used to detect and identify vehicle license plates and parking slot markers, ensuring consistency between reservations and actual parking slot usage. The research findings indicate that the sensors system has the fastest measurement time of 106.3 milliseconds and data transmission time of 73.6 milliseconds. The Faster R-CNN model shows and accuracy of 86% for detecting license plates and parking slot markers, while EasyOCR achieves 100% accuracy in recognizing the characters of license plates and parking slot markers. The average total time for inference and character extraction is 2.7 seconds. This integrated approach provides significant solution for both users and parking vendor.