| Title |
Development of an Algorithm for Train Approach Detection Based on Optical Flow Estimation AI Model |
| Authors |
김상암(Sang-Ahm Kim) ; 송은주(Eun-Ju Song) |
| DOI |
https://doi.org/10.5370/KIEE.2024.73.10.1794 |
| Keywords |
Railway Safety; Artificial Intelligence; Optical Flow; Train Access Information; Object Detection |
| Abstract |
This paper proposes an AI-based train approach detection algorithm designed for safety assistance systems aimed at reducing the increasing trend of accidents involving railway trackside workers. The proposed algorithm estimates optical flow using past and present images with a time difference, then detects trains in the current image using an object detection AI. It further utilizes radar data to acquire information on moving objects, combining these data to determine train approach. To validate the accuracy and reliability of the proposed algorithm, both laboratory and field tests were conducted, achieving a 100% detection rate in both daytime and nighttime conditions. The portable worker safety system incorporating this algorithm is expected to enhance the safety of trackside workers and contribute to the efficiency of railway maintenance operations. |