基于数字孪生与视觉驱动的重载铁路沿线障碍物识别

Obstacle Recognition Along Heavy-Duty Railway Lines Based on Digital Twin and Visual Drive

摘要:

铁路沿线轨道上障碍物若没有及时发现和清理,极有可能导致列车脱轨、碰撞,造成重大人员伤亡与经济损失。在此背景下,提出基于数字孪生与视觉驱动的多模态融合识别方法。激光雷达扫描与无人机倾斜摄影技术获取的数据,通过三维建模技术构建高精度铁路沿线实体对应的无障碍背景数字孪生模型,结合视觉传感器实时采集图像,根据SSIM的计算进行虚实场景匹配,实现障碍物的精准识别。结果表明:该方法的静态和动态障碍物识别效果。可以适应复杂铁路环境,识别结果可以为智能铁路安全监测提供了价值的参考意见。

Abstract:

If obstacles on the railway tracks are not detected and cleared in a timely manner, it is highly likely to cause train derailment, collision, and significant casualties and economic losses. In this context, a multimodal fusion recognition method based on digital twin and visual drive is proposed. The data obtained by laser radar scanning and drone oblique photography technology is used to construct high-precision digital twin models of obstacle free backgrounds corresponding to entities along the railway line through 3D modeling technology. Combined with real-time image acquisition by visual sensors, virtual real scene matching is performed based on SSIM calculation to achieve accurate recognition of obstacles. The results indicate that the method has good static and dynamic obstacle recognition performance. It can adapt to complex railway environments, and the recognition results can provide valuable reference opinions for intelligent railway safety monitoring.