Vision-based On Board Obstacle Detection System (VOODS) pada Sarana Perkeretaapian Menggunakan Pantauan Kamera

Diva Trisna; Sutrisno; Joko Hariyono

Authors

Diva Trisna Program Studi Teknik Elektro, Pascasarjana Universitas Sebelas Maret, Surakarta
Sutrisno Program Studi Teknik Elektro, Pascasarjana Universitas Sebelas Maret, Surakarta
Joko Hariyono Program Studi Teknik Elektro, Pascasarjana Universitas Sebelas Maret, Surakarta

Keywords:

Vision-based On Board Obstacle Detection System, YOLOv8 Segmentation, Obstacle Detection, Railway Safety, Rail Boundary System

Abstract

Urban railway industries face various safety challenges, including potential collisions, vehicles crossing rail tracks, and foreign objects on the track. Conventional inspection systems relying on train operators have limitations under restricted visibility and high speeds. This study develops a camera-based Vision-based On Board Obstacle Detection System (VOODS) using the YOLOv8 segmentation algorithm capable of detecting and classifying obstacles in real-time on railway tracks. The system utilizes existing CCTV cameras on trains, Ethernet network as communication channel, and a mini PC as the processing unit. A dataset was collected from the KRL Jabodetabek operational environment under morning and daytime lighting conditions, comprising 300 images (240 training and 60 validation) across four classes (Rail, Train, Human, Car). Annotation was performed using the LabelMe (Anaconda) platform, considering data confidentiality since the entire dataset is stored on the researcher’s local PC. The yolov8m-seg model was trained via Google Colaboratory with epochs=150, imgsz=1024, and patience=30. Danger zone determination is performed dynamically using the rail segmentation mask as Region of Interest (ROI); a warning subsystem then checks overlap between rail mask and object mask using bitwise AND operation to determine on-track presence. This approach refers to the clearance profile defined in Indonesian Ministerial Regulation PM No. 60 Year 2012. Evaluation results show a Box mAP@0.5 of 0.959 across all classes. Per-class performance: Rail 0.995, Train 0.989, Human 0.933, and Car 0.918. Confusion Matrix analysis shows a True Positive Rate of 1.00 for Rail, 0.96 for Train and Car, and 0.88 for Human. The warning subsystem was successfully tested on operational video and can trigger alerts when a human enters the rail area. The high Recall value is positive because, in railway safety context, Recall is more critical than Precision. The developed VOODS contributes significantly to improving operational safety of urban railway transportation in Indonesia

Downloads

Download data is not yet available.

References

Z. Zhao, J. Kang, Z. Sun, T. Ye, and B. Wu, “A real-time and high-accuracy railway obstacle detection method using lightweight CNN and improved transformer,” Measurement, vol. 238, p. 115380, Jul. 2024, doi: 10.1016/j.measurement.2024.115380.

D. Li and J. Zhang, “Intelligent road segmentation and obstacle detection for autonomous railway vehicle,” Advances in Mechanical Engineering, vol. 16, no. 1, p. 16878132231225312, Jul. 2024, doi: 10.1177/16878132231225312.

X.-Y. Xu, S.-M. Wang, W.-Q. Liu, and Y.-Q. Ni, “Advancements in Obstacle Intrusion Detection Methods for Rail Transit: A Comprehensive Review,” IEEE Transactions on Instrumentation and Measurement, vol. 74, pp. 1–34, Jul. 2025, doi: 10.1109/TIM.2025.3612624.

D. He et al., “Improved Mask R-CNN for obstacle detection of rail transit,” Measurement, vol. 190, p. 110728, Jul. 2022, doi: 10.1016/j.measurement.2022.110728.

D. He et al., “Urban rail transit obstacle detection based on Improved R-CNN,” Measurement, vol. 196, p. 111277, 2022, doi: 10.1016/j.measurement.2022.111277.

D. He, Z. Zou, Y. Chen, B. Liu, X. Yao, and S. Shan, “Obstacle detection of rail transit based on deep learning,” Measurement, vol. 176, p. 109241, Jul. 2021, doi: 10.1016/j.measurement.2021.109241.

F. Guo, Y. Qian, Y. Wu, Z. Leng, and H. Yu, “Automatic railroad track components inspection using real-time instance segmentation,” Computer-Aided Civil and Infrastructure Engineering, vol. 36, pp. 362–377, Mar. 2021, doi: 10.1111/mice.12625.

M. Zhang et al., “LDCD-Net: lightweight dual-channel detection network for obstacles detection on rail transit,” Signal Image Video Process., vol. 19, no. 4, p. 281, Jul. 2025, doi: 10.1007/S11760-025-03845-X.

X. Zhang, M. Zhou, P. Qiu, Y. Huang, and J. Li, “Radar and vision fusion for the real-time obstacle detection and identification,” Industrial Robot: the international journal of robotics research and application, vol. 46, no. 3, pp. 391–395, Jul. 2019, doi: 10.1108/IR-06-2018-0113.

Z. Nan et al., “LiDAR-Camera joint obstacle detection algorithm for railway track area,” Expert Systems with Applications, vol. 275, p. 127089, Jul. 2025, doi: 10.1016/j.eswa.2025.127089.

A. Chernov, M. Butakova, A. Guda, and P. Shevchuk, “Development of Intelligent Obstacle Detection System on Railway Tracks for Yard Locomotives Using CNN,” 2020, pp. 33–43. doi: 10.1007/978-3-030-58462-7_3.

D. Ristic-Durrant, M. A. Haseeb, M. Franke, M. Banic, M. Simonovic, and D. Stamenković, “Artificial Intelligence for Obstacle Detection in Railways: Project SMART and Beyond,” 2020, pp. 44–55. doi: 10.1007/978-3-030-58462-7_4.

S. W. Ibrahim, “A comprehensive review on intelligent surveillance systems,” Communications in Science and Technology, vol. 1, no. 1, Jul. 2016, doi: 10.21924/cst.1.1.2016.7.

G. Jocher, A. Chaurasia, and J. Qiu, “Ultralytics YOLOv8,” Jul. 15, 2023.

J. Terven and D. Cordova-Esparza, “A Comprehensive Review of YOLO Architectures in Computer Vision: From YOLOv1 to YOLOv8 and YOLO-NAS,” Machine Learning and Knowledge Extraction, vol. 5, no. 4, pp. 1680–1716, Jul. 2023, doi: 10.3390/make5040083.

J. M. Molina, J. P. Llerena, L. Usero, and M. A. Patricio, “Advances in instance segmentation: Technologies, metrics and applications in computer vision,” Neurocomputing, vol. 625, p. 129584, Jul. 2025, doi: 10.1016/j.neucom.2025.129584.

L. Rey et al., “A Performance Analysis of You Only Look Once Models for Deployment on Constrained Computational Edge Devices in Drone Applications,” Electronics, vol. 14, no. 3, p. 638, Jul. 2025, doi: 10.3390/electronics14030638.

C. Shorten and T. M. Khoshgoftaar, “A survey on Image Data Augmentation for Deep Learning,” Journal of Big Data, vol. 6, no. 1, p. 60, Jul. 2019, doi: 10.1186/s40537-019-0197-0.

E. Pandilova, M. Petrov, V. Spasev, I. Dimitrovski, and I. Kitanovski, “Transfer Learning with Yolo for Object Detection in Remote Sensing,” B. Risteska Stojkoska and S. Janeska Sarkanjac, Eds., Springer Nature Switzerland, 2025, pp. 121–135. doi: 10.1007/978-3-031-86162-8_9.

J. Jonathan and D. Hermanto, “Penentuan Epochs Hasil Model Terbaik: Studi Kasus Algoritma YOLOv8,” Digital Transformation Technology, vol. 4, no. 2, pp. 792–798, Jul. 2024, doi: 10.47709/digitech.v4i2.4640.

L. Prechelt, “Early Stopping — But When?,” in Neural Networks: Tricks of the Trade: Second Edition, G. Montavon, G. B. Orr, and K.-R. Müller, Eds., Berlin, Heidelberg: Springer, 2012, pp. 53–67. doi: 10.1007/978-3-642-35289-8_5.

PUSDATIN, “PM 60 TAHUN 2012,” Jul. 15, 2026.

T.-Y. Lin, P. Goyal, R. Girshick, K. He, and P. Dollár, “Focal Loss for Dense Object Detection,” IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 42, no. 2, pp. 318–327, Jul. 2020, doi: 10.1109/TPAMI.2018.2858826.

F. Jupiter, E. S. Negara, Y. Kunang, and M. Herdiansyah, “Implementasi Algoritma CNN dan YOLO untuk Mendeteksi Jenis Kendaraan pada Jalan Raya,” Explore: Jurnal Sistem Informasi dan Telematika, vol. 14, p. 110, 2023, doi: 10.36448/jsit.v14i2.3259.

R. Padilla, W. L. Passos, T. L. B. Dias, S. L. Netto, and E. A. B. Da Silva, “A Comparative Analysis of Object Detection Metrics with a Companion Open-Source Toolkit,” Electronics, vol. 10, no. 3, p. 279, Jul. 2021, doi: 10.3390/electronics10030279.