AMANDA, RIO (2025) DETEKSI KOMPONEN BAWAH KENDARAAN PADA BAGIAN SISTEM PEMBUANGAN MENGGUNAKAN ALGORITMA YOLO. Diploma thesis, POLITEKNIK KESELAMATAN TRANSPORTASI JALAN.
|
Text (ABSTRAK)
22033102-KKW-ABSTRAK.pdf Download (863kB) | Preview |
|
|
Text (BAB 1)
22033102-KKW-BAB _1.pdf Download (512kB) | Preview |
|
![[img]](http://eprints.pktj.ac.id/style/images/fileicons/text.png) |
Text (BAB 2)
22033102-KKW-BAB_2.pdf Restricted to Registered users only Download (897kB) | Request a copy |
|
|
Text (BAB 3)
22033102-KKW-BAB_3.pdf Restricted to Registered users only Download (1MB) | Request a copy |
|
|
Text (BAB 4)
22033102-KKW-BAB_4.pdf Restricted to Registered users only Download (1MB) | Request a copy |
|
|
Text (BAB 5)
22033102-KKW-BAB_5.pdf Download (481kB) | Preview |
|
|
Text (LAMPIRAN)
22033102-KKW-LAMPIRAN.pdf Restricted to Registered users only Download (2MB) | Request a copy |
Abstract
The exhaust system of motor vehicles plays a crucial role in reducing harmful emissions and ensuring driving safety and comfort. Components such as the catalytic converter, exhaust pipe, muffler, and tailpipe must be thoroughly inspected during technical vehicle examinations, particularly during undercarriage inspections. However, conventional manual inspections are time-consuming, require specific expertise, and are prone to human error. Therefore, an automated object detection system is needed to enhance the efficiency and accuracy of exhaust system inspections. This study aims to develop an automated detection system for under-vehicle exhaust components using the YOLO (You Only Look Once) algorithm. The research utilized primary data collected through underbody photographs of light commercial vehicles (JBB < 3,500 kg), specifically Daihatsu Gran Max, Mitsubishi L300, and Suzuki Carry. Images were annotated using the Roboflow platform and used to train three object detection models: YOLOv8, YOLOv9, and YOLOv11. The dataset was organized into ten versions, each increasing in size from 100 to 1,000 images, to examine the effect of dataset size on model performance. Model performance was evaluated using accuracy metrics including mean Average Precision (mAP), precision, recall, and F1-score. The results indicate that YOLOv11 outperformed YOLOv8 and YOLOv9 in both detection accuracy and training efficiency. The model effectively identified exhaust components and was capable of detecting potential damage, such as cracks or leaks, that may not be visible to the naked eye. The implementation of this model offers a modern solution for undercarriage inspections, allowing faster and more reliable assessments while reducing dependency on manual methods. This research contributes to the advancement of AI-based vehicle inspection systems and supports improved road safety and environmental sustainability through the application of automated visual detection technology. Keywords: Object Detection, Exhaust System, YOLO, Motor Vehicle, Undercarriage Inspection, Artificial Intelligence, Model Evaluation, Image Annotation
| Item Type: | Thesis (Diploma) |
|---|---|
| Uncontrolled Keywords: | Object Detection, Exhaust System, YOLO, Motor Vehicle, Undercarriage Inspection, Artificial Intelligence, Model Evaluation, Image Annotation |
| Subjects: | T Technology > T Technology (General) T Technology > TL Motor vehicles. Aeronautics. Astronautics |
| Divisions: | Teknologi Otomotif > Teknologi Otomotif |
| Depositing User: | 22033102 22033102 |
| Date Deposited: | 04 Aug 2025 03:01 |
| Last Modified: | 04 Aug 2025 03:01 |
| URI: | http://eprints.pktj.ac.id/id/eprint/3892 |
Actions (login required)
 |
View Item |
