VEHICLE PARKING VIOLATION DETECTION SYSTEM IN SPECIAL PARKING AREAS FOR LARGE VEHICLES USING YOLOv10 & FASTER R-CNN (Case Study: Undiksha Central Campus)

Fajar, Putu Bagus Muhammad (2025) VEHICLE PARKING VIOLATION DETECTION SYSTEM IN SPECIAL PARKING AREAS FOR LARGE VEHICLES USING YOLOv10 & FASTER R-CNN (Case Study: Undiksha Central Campus). Undergraduate thesis, Universitas Pendidikan Ganesha.

[img] Text (COVER)
2115101019-COVER.pdf
Download (2MB)
[img] Text (ABSTRAK)
2115101019-ABSTRAK.pdf
Download (364kB)
[img] Text (BAB 1 PENDAHULUAN)
2115101019-BAB 1 PENDAHULUAN.pdf
Download (318kB)
[img] Text (BAB 2 KAJIAN TEORI)
2115101019-BAB 2 KAJIAN TEORI.pdf
Restricted to Repository staff only
Download (687kB) | Request a copy
[img] Text (BAB 3 METODELOGI PENELITIAN)
2115101019-BAB 3 METODELOGI PENELITIAN.pdf
Restricted to Repository staff only
Download (201kB) | Request a copy
[img] Text (BAB 4 HASIL DAN PEMBAHASAN)
2115101019-BAB 4 HASIL DAN PEMBAHASAN.pdf
Restricted to Repository staff only
Download (1MB) | Request a copy
[img] Text (BAB 5 PENUTUP)
2115101019-BAB 5 PENUTUP.pdf
Restricted to Repository staff only
Download (31kB) | Request a copy
[img] Text (DAFTAR PUSTAKA)
2115101019-DAFTAR PUSTAKA.pdf
Download (169kB)
[img] Text (LAMPIRAN)
2115101019-LAMPIRAN.pdf
Download (442kB)

Abstract

The increasing number of motor vehicles, such as motorcycles and cars, has made the availability of parking spaces a crucial need, especially in public areas such as the campus of Universitas Pendidikan Ganesha. Parking regulations, including zones designated for large vehicles, aim to maintain order. However, violations still occur, particularly by two-wheeled vehicles parking in restricted areas. This study proposes a surveillance system based on CCTV utilizing object detection algorithms to automatically identify parking violations. Two object detection models are used: YOLOv10 (in five variants: nano, small, balance, medium, and large) and Faster R-CNN. YOLOv10 is selected for its real-time detection capability with fewer parameters, while Faster R-CNN is chosen for its high accuracy as a two-stage detector. The dataset consists of 2,268 filtered images from the COCO Dataset, split with an 8:2 ratio for training and testing. Additionally, nine simulation videos are used to evaluate detection performance on three key objects: cars, motorcycles, and persons. Training results show that the small version of YOLOv10 (YOLOv10s) performs best among all YOLOv10 variants with a mAP of 0.539, recall of 0.535, and inference speed of 11.2 ms. In contrast, Faster R-CNN achieves a mAP of 0.422 and a higher recall of 0.688, with a slower speed of 88.09 ms. In video testing, Faster R-CNN demonstrated parking violation detection with 88% accuracy and YOLOv10s at 84% accuracy. This advantage is attributed to its more “persistent” detection behavior, as indicated by its higher recall score. Meanwhile, YOLOv10s fails to detect small, occluded objects in one of the videos. Thus, despite having slower inference speed, Faster R-CNN proves to be more reliable in detecting parking violations compared to YOLOv10. classes.

Item Type: Thesis (Undergraduate)
Uncontrolled Keywords: Object Detection, Deep Learning, YOLOv10, Faster R-CNN, Parking Violation Detection
Subjects: Q Science > QA Mathematics > QA75 Electronic computers. Computer science
T Technology > T Technology (General)
Divisions: Fakultas Teknik dan Kejuruan > Jurusan Teknik Informatika > Program Studi Ilmu Komputer (S1)
Depositing User: Putu Bagus Muhammad Fajar
Date Deposited: 10 Aug 2025 06:06
Last Modified: 10 Aug 2025 06:06
URI: http://repo.undiksha.ac.id/id/eprint/26731

Actions (login required)

View Item View Item