Sistem Pencatatan Pelanggaran Kendaraan Overdimension Berbasis Multi Kamera Menggunakan Deep Learning

Suprantiyo, Javier Janeti (2025) Sistem Pencatatan Pelanggaran Kendaraan Overdimension Berbasis Multi Kamera Menggunakan Deep Learning. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5024211002-Undergraduate_Thesis.pdf Restricted to Repository staff only Download (18MB) | Request a copy

Abstract

Truk merupakan moda utama angkutan barang di Indonesia, namun tingginya permintaan sering memicu pelanggaran kelebihan dimensi (overdimension), yang merusak infrastruktur dan membahayakan keselamatan. Penelitian ini mengembangkan sistem deteksi otomatis berbasis deep learning dengan pendekatan multi-kamera untuk mendeteksi pelanggaran overdimension secara real-time. Model YOLOv8n-segmentasi digunakan untuk klasifikasi jenis dan bagian truk, serta pengukuran dimensi kendaraan. Hasil pelatihan model menunjukkan mAP50 untuk bounding box sebesar 0,90 dan mAP50 untuk mask segmentation sebesar 0,78. Sistem diuji pada perangkat Coral Dev Board dengan rata-rata FPS sebesar 9,19 untuk video pertama dan 9,14 untuk video kedua dalam skema satu perangkat dua input, serta 14,87 FPS dan 14,60 FPS dalam skema dua perangkat dua input. Sistem multi-kamera mampu mencapai tingkat konsistensi pelacakan objek sebesar 90% dan pencocokan fitur antar kamera dengan tingkat keberhasilan 96,67%. Dengan dukungan pelacakan objek dan pencocokan antar kamera, sistem mampu mendeteksi pelanggaran secara objektif dan ditampilkan melalui platform pemantauan berbasis web.
==================================================================================================================================
Trucks are the primary mode of freight transportation in Indonesia, but high demand often leads to overdimension violations, which damage infrastructure and pose safety risks. This study develops an automated detection system based on deep learning with a multi-camera approach to detect overdimension violations in real-time. The YOLOv8n-segmentation model is used to classify truck types and parts, as well as to measure vehicle dimensions. The model training results show an mAP50 for bounding boxes of 0,9 and an mAP50 for mask segmentation of 0,78. The system was tested on a Coral Dev Board with an average FPS of 9.19 for the first video and 9.14 for the second video in a single device, two-input configuration, and 14.87 FPS and 14.60 FPS in a two-device, two-input configuration. The multi-camera system achieved a tracking consistency of 90% and feature matching accuracy of 96.67%. With object tracking and cross-camera matching, the system can objectively detect violations, displayed through a web-based monitoring platform.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Deteksi Overdimension, Segmentasi, Multi Kamera, Situs Pemantauan, Pelacakan Objek, Pencocokan Fitur, Overdimension Detection, Segmentation, Multi-Camera, Monitoring Website, Object Tracking, Feature Matching.
Subjects:	Q Science > QA Mathematics > QA336 Artificial Intelligence T Technology > T Technology (General) > T57.5 Data Processing T Technology > TE Highway engineering. Roads and pavements > TE228.3 Intelligent transportation systems. T Technology > TK Electrical engineering. Electronics Nuclear engineering > TK3070 Automatic control T Technology > TK Electrical engineering. Electronics Nuclear engineering > TK5105.888 Web sites--Design. Web site development. T Technology > TK Electrical engineering. Electronics Nuclear engineering > TK6592.A9 Automatic tracking.
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Computer Engineering > 90243-(S1) Undergraduate Thesis
Depositing User:	Javier Janeti Suprantiyo
Date Deposited:	04 Jun 2025 04:25
Last Modified:	04 Jun 2025 04:25
URI:	http://repository.its.ac.id/id/eprint/119115

Actions (login required)

View Item