Sistem Pemantauan Kecepatan Kendaraan Berbasis Pengolahan Citra Digital dengan Model YOLOv8 pada CCTV di Lingkungan Institut Teknologi Sepuluh Nopember

Rahmatullah, Moh. Sulthan Arief (2025) Sistem Pemantauan Kecepatan Kendaraan Berbasis Pengolahan Citra Digital dengan Model YOLOv8 pada CCTV di Lingkungan Institut Teknologi Sepuluh Nopember. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5027211045-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (20MB) | Request a copy

Abstract

Pelanggaran batas kecepatan di lingkungan kampus seperti ITS menjadi risiko keselamatan seiring meningkatnya volume kendaraan, sementara sistem pemantauan manual memiliki keterbatasan operasional. Penelitian ini bertujuan mengembangkan dan mengevaluasi sistem pemantauan kecepatan kendaraan otomatis secara real-time berbasis model deep learning YOLOv8 menggunakan infrastruktur CCTV. Metodologi penelitian mencakup akuisisi data ground truth kecepatan, pelacakan objek kendaraan dengan BYTETracker, transformasi perspektif untuk pengukuran jarak akurat, estimasi kecepatan dari pergerakan objek, dan penghalusan nilai kecepatan menggunakan Exponential Moving Average (EMA). Berbagai varian model YOLOv8 (nano hingga extra-large) dianalisis kinerjanya pada input video rekaman dan uji coba langsung streaming RTSP, serta dalam kondisi pencahayaan cukup dan minim. Hasil penelitian menunjukkan trade-off signifikan antara akurasi dan efisiensi komputasi. Model besar seperti YOLOv8x mencapai akurasi tertinggi dengan Mean Absolute Error (MAE) serendah 0,14 km/jam, namun dengan beban komputasi tinggi (utilisasi GPU 70,4% dan FPS 16). Sebaliknya, model kecil seperti YOLOv8n sangat efisien (GPU 14,69% dan FPS 34), namun akurasinya sangat rendah dan tidak stabil (MAE hingga 20,94 km/jam), meskipun tidak menghasilkan deteksi palsu. Di antara keduanya, YOLOv8m menawarkan keseimbangan terbaik. Sistem berkinerja sangat baik pada pencahayaan ideal, namun gagal total pada kondisi minim cahaya (malam hari) akibat efek silau lampu kendaraan. Disimpulkan bahwa sistem ini dapat diimplementasikan, namun pemilihan model optimal dan penanganan kondisi pencahayaan buruk menjadi krusial untuk penerapan praktis.
=====================================================================================================================================
The increasing volume of vehicles at Institut Teknologi Sepuluh Nopember (ITS) has raised safety concerns due to speed limit violations, while manual monitoring systems face operational limitations. This study develops and evaluates a real-time automatic vehicle speed monitoring system based on the YOLOv8 deep learning model, utilizing existing CCTV infrastructure. The methodology includes ground truth data acquisition, vehicle tracking with BYTETracker, perspective transformation for accurate distance measurement, speed estimation from object movement, and value smoothing using an Exponential Moving Average (EMA). Various YOLOv8 model variants (from nano to extra-large) were analyzed, revealing a significant tradeoff between accuracy and computational efficiency. The large model, YOLOv8x, achieved the highest accuracy with a Mean Absolute Error (MAE) as low as 0.14 km/h, but at a high computational cost (70.4% GPU utilization and 16 FPS). Conversely, the smallest model, YOLOv8n, was highly efficient (14.69% GPU and 34 FPS) but suffered from very low and unstable accuracy (MAE up to 20.94 km/h), although no false detections were observed. YOLOv8m was identified as the most balanced option. System performance was excellent in ideal lighting but failed completely in low-light (night) conditions due to vehicle headlight glare. The study concludes that the system is viable, but the selection of an optimal model and addressing poor lighting conditions are crucial for practical deployment.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	CCTV, Deteksi Objek, Estimasi Kecepatan, YOLOv8, object Detection, Speed Estimation
Subjects:	T Technology > T Technology (General) > T174 Technological forecasting T Technology > T Technology (General) > T57.5 Data Processing T Technology > T Technology (General) > T57.62 Simulation
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Information Technology > 59201-(S1) Undergraduate Thesis
Depositing User:	Moh. Sulthan Arief Rahmatullah
Date Deposited:	17 Jul 2025 05:52
Last Modified:	17 Jul 2025 07:21
URI:	http://repository.its.ac.id/id/eprint/119863

Actions (login required)

View Item