Abstract: To solve the problems of low efficiency and poor accuracy of current railway turnout manual detection methods,a method based on deep learning to quickly identify turnout scene and detect turnout spacing was proposed.The railway point cloud information was obtained by linear array industrial camera scanning,the railway turnout scene recognition network connected by residual was designed,the optimal super parameters were searched with Tree-structured Parzen Estimator (TPE) algorithm,and the unbalanced number of samples was solved by focal loss function,so as to realize the accurate and fast recognition of railway turnout scene.Based on the recognized scene image of railway turnout,an edge extraction algorithm of turnout stock rail and switch rail was developed to accurately measure the inner distance between turnout stock rail and switch rail.The experimental results showed that the recognition accuracy of the proposed method reached 97.5%,the recognition time was within 0.02s,and the calculation error of turnout spacing was less than 0.2mm.Compared with the manual detection method,the detection efficiency and accuracy were greatly improved,which met the requirements of turnout detection.

Key words: railway turnouts, deep learning, scene recognition, edge extraction, distance measurement

摘要： 针对目前铁路道岔人工检测方法效率低、精度差的问题,提出一种基于深度学习的快速识别道岔场景及检测道岔间距的方法。采用线阵工业相机扫描获取铁路点云信息,设计残差连接的铁路道岔场景识别网络,采用树结构Parzen估计算法搜索最优超参数,采用Focal Loss损失函数解决样本数量不均衡问题,实现铁路道岔场景的准确快速识别。利用识别出的铁路道岔场景图像,开发了一种道岔基本轨与尖轨的边缘提取算法,准确测量道岔基本轨与尖轨的内侧间距。实验结果证明,该方法的识别准确率达到97.5%,识别时间在0.02 s内,道岔间距计算误差小于0.2 mm,相比人工检测方法,检测效率与检测精度均大幅提升,满足道岔检测的要求。

关键词: 铁路道岔, 深度学习, 场景识别, 边缘提取, 间距测量