Prediksi Jarak dan Waktu Collision Antar Kapal dengan Algoritma Vision-Based Object Recognition

Parirak Tandiasik, Ahnastasya (2026) Prediksi Jarak dan Waktu Collision Antar Kapal dengan Algoritma Vision-Based Object Recognition. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5018221005-Undergraduate_Thesis.pdf Restricted to Repository staff only Download (5MB) | Request a copy

Abstract

Collision antar kapal merupakan salah satu risiko dalam aktivitas pelayaran yang dapat terjadi apabila keberadaan dan jarak objek di sekitar kapal tidak terpantau dengan baik. Pengamatan visual secara manual memiliki keterbatasan karena dipengaruhi oleh jarak, sudut pandang, pencahayaan, dan kondisi lingkungan. Penelitian ini bertujuan untuk merancang sistem prediksi jarak dan waktu collision antar kapal berbasis vision-based object recognition. Sistem yang dikembangkan menggabungkan deteksi objek menggunakan YOLOv12, stereo vision, perhitungan disparity map menggunakan StereoSGBM, rekonstruksi 3D, estimasi jarak, dan perhitungan Time to Collision. Input citra diperoleh dari kamera stereo, kemudian frame dipisahkan menjadi citra kiri dan citra kanan. Kedua citra diproses melalui rectification menggunakan parameter kalibrasi kamera stereo, lalu digunakan untuk membentuk disparity map dan data kedalaman. Deteksi objek kapal dilakukan menggunakan model YOLOv12, sedangkan titik tengah bounding box digunakan sebagai acuan pengambilan nilai kedalaman dari data points_3D. Nilai jarak yang diperoleh kemudian digunakan untuk menghitung Time to Collision berdasarkan perubahan jarak objek terhadap waktu. Hasil implementasi menunjukkan bahwa sistem mampu menampilkan bounding box objek kapal, kelas objek, ID objek, estimasi jarak, nilai Time to Collision, FPS, dan disparity map secara langsung. Hasil pengujian pada beberapa varian kapal menunjukkan selisih jarak antara jarak aktual dan estimasi sistem berada pada rentang 0,00 m sampai 0,04 m. Dengan demikian, sistem ini dapat digunakan sebagai dasar pengembangan sistem bantu visual untuk mendeteksi kapal, memperkirakan jarak, dan memberikan informasi waktu potensi collision.
=================================================================================================================================
Ship collision is one of the risks in maritime activities that may occur when the presence and distance of surrounding objects are not properly monitored. Manual visual observation has limitations because it is affected by distance, viewing angle, lighting, and environmental conditions. This research aims to design a system for predicting the distance and collision time between vessels based on vision-based object recognition. The developed system combines object detection using YOLOv12, stereo vision, disparity map calculation using StereoSGBM, 3D reconstruction, distance estimation, and Time to Collision calculation. Image input is obtained from a stereo camera, and the captured frame is divided into left and right images. Both images are processed through rectification using stereo camera calibration parameters, then used to generate a disparity map and depth data. Ship object detection is performed using the YOLOv12 model, while the center point of the bounding box is used as a reference for obtaining depth values from points_3D data. The obtained distance value is then used to calculate Time to Collision based on changes in object distance over time. The implementation results show that the system is able to display ship object bounding boxes, object classes, object IDs, distance estimation, Time to Collision values, FPS, and disparity map in real time. Testing on several ship variants shows that the difference between the actual distance and the system estimation ranges from 0.00 m to 0.04 m. Therefore, this system can be used as a basis for developing a visual assistance system to detect ships, estimate distance, and provide information on potential collision time.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	collision kapal, YOLOv12, stereo vision, disparity map, Time to Collision, ship collision, YOLOv12, stereo vision, disparity map, Time to Collision.
Subjects:	V Naval Science > V Naval Science (General) V Naval Science > VK > VK200 Merchant marine--Safety measures V Naval Science > VK > VK555 Navigation. V Naval Science > VM Naval architecture. Shipbuilding. Marine engineering > VM293 Shipping--Indonesia--Safety measures
Divisions:	Faculty of Marine Technology (MARTECH) > Naval Architecture and Shipbuilding Engineering > 36201-(S1) Undergraduate Thesis
Depositing User:	Ahnastasya Parirak Tandiasik
Date Deposited:	01 Aug 2026 05:24
Last Modified:	01 Aug 2026 05:24
URI:	http://repository.its.ac.id/id/eprint/141425

Actions (login required)

View Item