Fitur Extraction untuk Mencari Kecepatan Relatif Berdasarkan Data Dashcam Kendaraan

Nafis, Akmal (2025) Fitur Extraction untuk Mencari Kecepatan Relatif Berdasarkan Data Dashcam Kendaraan. Project Report. [s.n.], [s.l.]. (Unpublished)

Text 5025211216-Project_Report.pdf - Accepted Version Download (893kB)

Abstract

Estimasi kecepatan kendaraan merupakan komponen fundamental dalam pengembangan sistem kendaraan otonom yang mempengaruhi kualitas pengambilan keputusan pada berbagai situasi dinamis. Data kecepatan yang akurat sangat essential untuk mendukung navigasi yang aman dan efisien. Meskipun sensor kecepatan konvensional dapat menyediakan informasi langsung, ketergantungan pada sensor tunggal berpotensi meningkatkan risiko kegagalan sistem. Penelitian ini mengimplementasikan pipeline terintegrasi yang menggabungkan sebagai fitur ekstraksi untuk estimasi menggunakan optical flow. Hasil implementasi menunjukkan bahwa sistem yang dikembangkan berhasil mengintegrasikan ketiga komponen dalam satu pipeline yang berfungsi efektif pada data video dashcam, dengan SerNet-Former memberikan segmentasi yang memuaskan, ProPainter melakukan inpainting yang sukses, dan RAFT menghasilkan estimasi optical flow yang lebih bersih dan konsisten
=================================================================================================================================
Estimating vehicle speed is a fundamental component in the development of autonomous driving systems, as it plays a critical role in how the system makes decisions in various dynamic situations. Having accurate speed data is essential to support safe and efficient navigation. Although conventional speed sensors are capable of providing direct speed information, relying on a single sensor can introduce potential risks, particularly if that sensor fails. This dependency can reduce the overall reliability of the autonomous system. In this research, we implement an integrated pipeline that uses optical flow as a feature extraction method for estimating vehicle speed. The goal is to develop a more robust approach that doesn’t rely solely on physical speed sensors, by leveraging computer vision techniques instead. The results of the implementation show that the developed system successfully integrates all three main components into one cohesive pipeline that works effectively with dashcam video data. The SerNet-Former model produces satisfying segmentation results, which help the system identify and isolate relevant objects and regions in the video frames. ProPainter is then used to perform inpainting, and it does this successfully by reconstructing occluded or missing parts of the image to maintain visual consistency. Finally, RAFT is used to estimate optical flow, and it produces cleaner and more consistent motion estimation results compared to traditional methods.

Item Type:	Monograph (Project Report)
Uncontrolled Keywords:	Speed Estimation, image segmentation, Inpainting, Optical Flow, SerNet-Former, ProPainter, RAFT, Estimasi Kecepatan, Segmentasi Citra, Inpainting, Optical Flow, SerNet-Former, ProPainter, RAFT,
Subjects:	Q Science > Q Science (General) > Q325.5 Machine learning. Support vector machines.
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Informatics Engineering > 55201-(S1) Undergraduate Thesis
Depositing User:	Akmal Nafis
Date Deposited:	23 Jun 2025 04:28
Last Modified:	23 Jun 2025 04:28
URI:	http://repository.its.ac.id/id/eprint/119225

Actions (login required)

View Item