基于机器视觉和边云协同的道岔缺口检测方法

A Switch Gap Detection Method Based on Machine Vision and Edge and Cloud Collaboration

摘要:

针对目前道岔缺口检测精确度低、鲁棒性差、难以在边缘计算设备满足性能要求等问题，提出了一种基于机器视觉和边云协同的道岔缺口检测方法；采用YOLOv8算法对缺口和检测柱进行高效目标检测，并通过图像处理技术精确获取边缘信息，以计算缺口的偏移量；实验结果表明，该方法在正常和摄像头抖动情况下均展现出优越的检测性能，平均误差控制在0.1mm以内，满足铁路现场实际工程上的应用；采用边云协同的工作模式显著提升了数据处理效率，优化了资源使用，降低了对单一计算节点的依赖，单次处理平均耗时在59.16ms，相比单独依靠边缘设备平均耗时降低了48.99%。

Abstract:

Aiming at the problems of low accuracy, poor robustness and difficulty in meeting the performance requirements of edge computing equipment in turnout gap detection, a turnout gap detection method based on machine vision and edge and cloud collaboration is proposed; Using the YOLOv8 algorithm for efficient object detection of gaps and detection columns, and accurately obtaining edge information through image processing technology to calculate the offset of gaps; The experimental results show that this method exhibits superior detection performance under normal and camera shake conditions, with an average error controlled within 0.1mm, which meets the practical application in railway field engineering; The adoption of edge and cloud collaboration significantly improves data processing efficiency, optimizes resource utilization, and reduces dependence on a single computing node. The average processing time for a single transaction is 59.16ms, which is 48.99% lower than relying solely on edge devices.