ANALISIS PENGUJIAN KINERJA SISTEM PENGUKUR KECEPATAN KENDARAAN MENGGUNAKAN CAMERA VISION BERBASIS YOLO V5

Perkembangan teknologi computer vision membuka peluang dalam pengembangan sistem monitoring kecepatan kendaraan yang lebih modern dan fleksibel dibandingkan metode konvensional berbasis sensor fisik. Di lingkungan Politeknik Penerbangan Surabaya, kebutuhan akan sistem monitoring kecepatan kendaraan menjadi penting untuk mendukung keselamatan dan ketertiban area operasional kampus, khususnya pada jalur kendaraan di sekitar fasilitas pendidikan dan operasional. Namun, penelitian sebelumnya lebih banyak berfokus pada sistem pengukuran berbasis sensor atau hanya pada deteksi objek kendaraan, sedangkan analisis kinerja sistem pengukuran kecepatan berbasis camera vision menggunakan algoritma YOLOv5 masih belum banyak dibahas secara mendalam. Oleh karena itu, penelitian ini bertujuan untuk menganalisis kinerja sistem pengukuran kecepatan kendaraan berbasis camera vision menggunakan algoritma YOLOv5. Proses pengukuran dilakukan dengan mendeteksi dan melacak pergerakan kendaraan roda dua menggunakan metode centroid tracking, kemudian menghitung estimasi kecepatan berdasarkan perpindahan objek antar-frame video. Pengujian dilakukan pada variasi kecepatan 20 km/jam, 30 km/jam, dan 40 km/jam dengan membandingkan hasil pengukuran sistem terhadap speedometer kendaraan sebagai nilai acuan (ground truth). Parameter evaluasi yang digunakan meliputi persentase error, Mean Absolute Error (MAE), dan Root Mean Square Error (RMSE). Hasil pengujian menunjukkan bahwa sistem mampu melakukan pengukuran kecepatan kendaraan secara real-time dengan tingkat akurasi yang cukup baik. Sistem menghasilkan rata-rata persentase error sebesar 5,50%, nilai MAE sebesar 1,60 km/jam, dan RMSE sebesar 1,88 km/jam, sehingga masih berada dalam kategori baik dan dapat diterima. Meskipun terdapat keterbatasan pada perangkat dengan kemampuan komputasi rendah dalam menjalankan proses inferensi secara stabil serta peluang pengembangannya untuk mendukung sistem pengawasan kendaraan operasional di area apron bandara.

Kata kunci : camera vision, YOLOv5, pengukuran kecepatan, MAE, RMSE, monitoring kecepatan

===========================================================
The development of computer vision technology opens up opportunities for the development of a more modern and flexible vehicle speed monitoring system compared to conventional methods based on physical sensors. In the Surabaya Aviation Polytechnic environment, the need for a vehicle speed monitoring system is crucial to support the safety and order of the campus operational area, especially on vehicle routes around educational and operational facilities. However, previous studies have focused more on sensor-based measurement systems or only on vehicle object detection, while the performance analysis of a computer vision-based speed measurement system using the YOLOv5 algorithm has not been discussed in depth. Therefore, this study aims to analyze the performance of a computer vision-based vehicle speed measurement system using the YOLOv5 algorithm. The measurement process is carried out by detecting and tracking the movement of two-wheeled vehicles using the centroid tracking method, then calculating the speed estimate based on the object's displacement between video frames. Tests were conducted at speed variations of 20 km/h, 30 km/h, and 40 km/h by comparing the system's measurement results with the vehicle's speedometer as a reference value (ground truth). The evaluation parameters used include the percentage error, Mean Absolute Error (MAE), and Root Mean Square Error (RMSE). The test results show that the system is capable of measuring vehicle speed in real-time with a fairly good level of accuracy. The system yielded an average error percentage of 5.50%, an MAE value of 1.60 km/h, and an RMSE of 1.88 km/h, which are still in the good and acceptable category. Despite the limitations of devices with low computing capabilities in running the inference process stably, there are development opportunities to support operational vehicle monitoring systems in airport apron areas.

Keywords: camera vision, YOLOv5, speed measurement, MAE, RMSE, speed monitoring