Parking Assistance Application using IoT Devices and Machine Learning on the Edge

Thumbnail Image

Date

2024-04-20

Journal Title

Journal ISSN

Volume Title

Publisher

Nazarbayev University School of Engineering and Digital Sciences

Abstract

Traditional parking management systems are inefficient in monitoring large scaled parking slot. This study addresses the growing issue of urban parking inefficiency, where drivers waste significant time searching for vacant spots. This problem is tackled by deploying edge devices in the parking lot, equipped with compact convolutional neural networks for real-time analysis. To evaluate performance, three models (mAlexNet224, mAlexNet32, and Mohan's model) were chosen from other research or adapted for this task. These models were then tested in a simulated parking lot environment on OpenMV or Raspberry Pi depending on the model's input size. While mAlexNet224 on Raspberry Pi achieved the best accuracy among the tested models, other lightweight models running on OpenMV also delivered satisfactory results, exceeding 90\% in accuracy. Based on the results of the study, mAlexNet32 and Mohan's model on OpenMV's camera appear well-suited for real-world deployment and the affordability of OpenMV devices further strengthens this case.

Description

Keywords

Parking, Model comparison, Weather conditions, OpenMV, RaspberryPI, Low-resource models, Type of access: Restricted

Citation

Amangeldina A., Baimukhanov B., Saduakhas R., and Tuleshov Z. (2024) Parking Assistance Application using IoT Devices and Machine Learning on the Edge. Nazarbayev University School of Engineering and Digital Sciences