Skripsi

ESTIMASI DAN PENGUKURAN DIMENSI LUBANG JALAN SECARA REAL-TIME DENGAN ALGORITMA DEEP LEARNING DAN SENSOR ULTRASONIK

---

Pothole detection systems typically feature real-time measurement of pothole length, width, and depth, achieved through the integration of a distance sensor for depth and a camera for width and length measurement. However, previous research using this methodology focused primarily on object detection rather than comprehensive dimension estimation. Consequently, this study aimed to integrate both detection and real-time dimension estimation features. The dataset was collected directly from video recordings using an action camera on roads in Palembang City, South Sumatra (Jl. Putri Kembang Dadar and Jl. Masjid Algazali). The research utilized the YOLO V11s-Seg algorithm, comparing it against EfficientDet-D0 and EfficientDet-D2. Training results indicated the superior capability of YOLO V11s-Seg in segmentation and dimension estimation with a mAP-50 of 76%, significantly outperforming EfficientDet-D0 (0.2%) and EfficientDet-D2 (3.6%). In subsequent testing, EfficientDet-D0 showed substantial errors in width and height (1,205.2% and 3,315.2%, respectively), while YOLO V11s-Seg was far more accurate, achieving testing errors of 95.64% (width) and 34.53% (height). Furthermore, in real-time testing, the YOLO V11s-Seg system demonstrated better precision with width and height errors of 32.96% and 14.41%, respectively. Finally, the depth measurement using the ultrasonic sensor yielded a separate percentage error of 16.7%.

---

Availability

Detail Information

Series Title

-

Call Number

T1867922025

Publisher

Indralaya : Prodi Teknik Elektro, Fakultas Teknik Universitas Sriwijaya., 2025

Collation

xiii, 115 hlm.; ilus.; tab.; 29 cm.

Language

Indonesia

ISBN/ISSN

-

Classification

625.807

Content Type

Text

Media Type

unmediated

Carrier Type

other (computer)

Edition

-

Subject(s)

Prodi Teknik Elektro
Teknik Pengukuran Jalan

Specific Detail Info

-

Statement of Responsibility

MI

Other version/related

File Attachment

Comments

---

You must be logged in to post a comment