Customized obstacle detection system for High-Speed Railways: A novel approach toward intelligent rail transportation

Abstract

With the rapid advancement of rail transportation technology, particularly in high-speed rail, efficient and accurate obstacle detection is a crucial research focus. Traditional methods often depend on extensive datasets and complex computations, necessitating high-performance GPUs, which escalate hardware costs and power consumption. Moreover, these approaches may struggle with real-time performance and robustness.

To address these challenges, we propose a novel approach termed the “Customized Obstacle Detection System (CODS)” for high-speed railways. CODS swiftly and precisely identifies non-track elements by analyzing discrepancies between real-time sensor data and a predefined background model of an obstacle-free track. The proposed system is composed of three main components: constructing a prototypical rail environment, analyzing discrepancies to detect obstacles, and implementing a self-supervised mapping update with distributed storage.

Experimental results demonstrate that CODS significantly enhances obstacle detection, achieving a 10% increase in detection mean average precision and a 75% improvement in detection speed under various railway conditions. This research offers a robust, efficient solution for obstacle detection, contributing to the development of intelligent rail transportation systems.