Vacant parking space detector for UTAR Kampar campus using YOLOv4

• Main Author: Lee, Vincent Wen Sheng
• Format: Final Year Project / Dissertation / Thesis
• Published: 2023
• Subjects: T Technology (General); TD Environmental technology. Sanitary engineering; TH Building construction
• Online Access: http://eprints.utar.edu.my/6000/1/fyp_IA_2023_LVWS.pdf; http://eprints.utar.edu.my/6000/

This project aims to develop a custom YOLOv4 detector model that can accurately detect vacant and occupied parking spaces on UTAR Kampar Campus. The motivation behind this project isto addressthe parking issuesfaced by students and staff on campus, which include difficulties in finding parking spots, illegal parking, and wasted time and fuel in search of parking. To gather data and identify the extent of the problem, a questionnaire survey was conducted among 10 random students on campus. Based on the results, up to 90% of the surveyors find it difficult to find parking on campus, and 80% have been late to or missed classes due to parking issues, impacting their academic performance. To improve the accuracy of the custom detector model, the project focuses on data preprocessing and augmentation, which involves collecting and labelling images of parking lots in various conditions, including lighting, weather, and vehicle types. The accuracy of the bounding box predictions is targeted to be above 70% for the whole image. If the targeted accuracy is not achieved, the data preprocessing process will start over again with the addition of new data sets. The proposed custom YOLOv4 detector model will benefit students and staff by providing the latest information on parking lot availability, reducing the time spent searching for available parking spots, promoting eco-friendliness by reducing fuel consumption and carbon emissions, and reducing instances of illegal parking that can lead to fines and contribute to congestion in the parking lot. Overall, this project presents a promising solution to the parking issues faced by UTAR Kampar Campus, with the potential for future expansion and application in other campuses or public areas. The result of this project is able to accurately predict the bounding box of vacant and occupied parking lots in UTAR Kampar Campus.