Rahmawati, - (2024) PERANCANGAN MODEL SISTEM DETEKSI DAN ESTIMASI JARAK SPEED BUMP UNTUK KENDARAAN OTONOM MENGGUNAKAN ALGORITMA DEEP LEARNING. S1 thesis, Universitas Pendidikan Indonesia.
![]() |
Text
S_TEKOM_2008187_Title.pdf Download (579kB) |
![]() |
Text
S_TEKOM_2008187_Chapter1.pdf Download (78kB) |
![]() |
Text
S_TEKOM_2008187_Chapter2.pdf Restricted to Staf Perpustakaan Download (575kB) | Request a copy |
![]() |
Text
S_TEKOM_2008187_Chapter3.pdf Download (322kB) |
![]() |
Text
S_TEKOM_2008187_Chapter4.pdf Restricted to Staf Perpustakaan Download (2MB) | Request a copy |
![]() |
Text
S_TEKOM_2008187_Chapter5.pdf Download (61kB) |
![]() |
Text
S_TEKOM_2008187_Appendix.pdf Restricted to Staf Perpustakaan Download (5MB) | Request a copy |
Abstract
Sistem deteksi speed bump merupakan komponen penting yang harus dimiliki oleh kendaraan otonom level 3 dan 4, sesuai instruksi dari US National Highway Traffic Safety Administration (NHTSA). Berbagai metode telah dikembangkan untuk sistem deteksi speed bump, termasuk penggunaan sensor ultrasonik dan algoritma deep learning seperti CNN, YOLOv2, dan YOLOv5. Namun, metode-metode ini masih memiliki beberapa kekurangan, seperti jarak deteksi yang terlalu dekat, kecepatan pemrosesan model yang lambat, dan nilai mAP yang kurang dari 90%. Oleh karena itu, penelitian ini dilakukan untuk merancang model sistem deteksi dan estimasi jarak speed bump untuk kendaraan otonom menggunakan algoritma deep learning. Metode penelitian yang digunakan dalam penelitian ini adalah AI Project Cycle. Dataset yang digunakan merupakan dataset kustom yang disusun berdasarkan aturan speed bump yang berlaku di Indonesia. Model yang dikembangkan menggunakan arsitektur YOLOv8n dan metode pengukuran jarak menggunakan konsep stereo vision dengan kamera stereo. Model hasil pelatihan memiliki mAP sebesar 92,5%. Model diimplementasikan pada Jetson Nano dengan dukungan GPU dengan kecepatan komputasi mencapai 3-4 FPS. Hasil pengujian menunjukkan model dapat mendeteksi dan mengukur jarak speed bump pada siang hari dengan jarak maksimal 20 meter, namun menjadi terbatas ketika malam hari. Rata-rata nilai error pengukuran estimasi jarak mencapai 0,41-0,891 meter. Intensitas cahaya terbukti mempengaruhi kinerja model dalam mendeteksi dan mengukur jarak speed bump. ---------- A speed bump detection system is an important component for level 3 and 4 autonomous vehicles, as instructed by the US National Highway Traffic Safety Administration (NHTSA). Various methods have been developed for speed bump detection systems, including the use of ultrasonic sensors and deep learning algorithms such as CNN, YOLOv2, and YOLOv5. However, these methods still have some shortcomings, such as too close detection distance, slow model processing speed, and mAP value less than 90%. Therefore, this research was conducted to design a speed bump detection and distance estimation system model for autonomous vehicles using deep learning algorithms. The research method used in this research is AI Project Cycle. The dataset used is a custom dataset compiled based on the speed bump rules applicable in Indonesia. The model developed uses the YOLOv8n architecture and the distance measurement method uses the concept of stereo vision with a stereo camera. The trained model has a mAP of 92.5%. The model is implemented on Jetson Nano with GPU support with computational speed reaching 3-4 FPS. The test results show that the model can detect and measure the distance of speed bumps during the day with a maximum distance of 20 meters, but it becomes limited at night. The average error value of distance estimation measurement reaches 0.41-0.891 meters. Light intensity is proven to affect the model's performance in detecting and measuring speed bump distance.
Item Type: | Thesis (S1) |
---|---|
Additional Information: | Link Google Scholar: https://scholar.google.com/citations?user=1cCoEYUAAAAJ&hl=en ID SINTA Dosen Pembimbing: Munawir: 6745899 Muhammad Taufik Dwi Putra: 6745726 |
Uncontrolled Keywords: | Deep Learning, Deteksi Objek, Estimasi Jarak, Speed Bump, YOLOv8, Distance Estimation, Deep Learning, Object Detection, Speed Bump, YOLOv8 |
Subjects: | Q Science > QA Mathematics > QA75 Electronic computers. Computer science T Technology > T Technology (General) |
Divisions: | UPI Kampus cibiru > S1 Teknik Komputer |
Depositing User: | Rahmawati |
Date Deposited: | 23 Aug 2024 04:23 |
Last Modified: | 23 Aug 2024 04:23 |
URI: | http://repository.upi.edu/id/eprint/120112 |
Actions (login required)
![]() |
View Item |