Skripsi

DETEKSI DAN ESTIMASI DIMENSI LUBANG JALAN SECARA REAL-TIME DENGAN ALGORITMA MASK R-CNN

---

Pothole detection on roads features measurement of the length and width of potholes and can detect the depth of potholes in real-time. This is possible due to the presence of distance sensors to detect pothole depth and cameras to detect the length and width of potholes on the road. However, previous research has focused only on object detection without considering the estimation of pothole dimensions. Therefore, this study combines detection features with dimension estimation in real-time. The dataset used comprises pothole data collected by recording videos using an action camera along the following roads in Palembang, South Sumatra: Jl. Kaca Piring, Jl. Swakarya I & II, Jl. Dwikora II, Jl. Ariodillah, and from Jl. Kp Bali to Jl. Sungai Kundur. This research uses the Mask R-CNN algorithm, performing three training processes with 700, 800, and 900 iterations. The best model was obtained at 800 iterations with the lowest loss of 0.6198. The measurement of pothole length and width with Mask R-CNN applies the influence of depth information to increase measurement accuracy. Thus, testing with the test dataset, video simulation, and real-time simulation yielded smaller errors. Testing with the test dataset showed width and height errors of 30.49% and 28.75%, respectively. Video simulation testing showed width and height errors of 32.26% and 18.035%, respectively. Finally, real-time detection and dimension estimation of potholes showed width and height errors of 29.53% and 68.04%, respectively. It can be concluded that the Mask R-CNN model can detect and estimate pothole dimensions using the test dataset in image, video simulation, and real-time formats. In addition to measuring pothole length and width with Mask R-CNN, pothole depth measurement was also conducted using an ultrasonic sensor mounted on a motorcycle. The measurement results from the sensor were compared to the actual pothole dimensions, yielding an error percentage of 6.27%.

---

Availability

Detail Information

Series Title

-

Call Number

T1514092024

Publisher

Indralaya : Prodi Teknik Elektro, Fakultas Teknik Universitas Sriwijaya., 2024

Collation

xiv, 79 hlm.; ilus.; 29 cm

Language

Indonesia

ISBN/ISSN

-

Classification

006.370 7

Content Type

Text

Media Type

unmediated

Carrier Type

-

Edition

-

Subject(s)

Prodi Teknik Elektro
Mask R-CNN

Specific Detail Info

-

Statement of Responsibility

KA

Other version/related

File Attachment

Comments

---

You must be logged in to post a comment