基于线激光扫描成像技术的钢轨炉号识别算法

Rail Number Recognition Algorithm Based on Line Laser Scanning Imaging Technology

为了解决人工识别钢轨炉号效率低下,费时费力等问题,采用非接触式线激光扫描成像的方法对钢轨炉号进行识别,结合C#和HALCON混合编程开发了字符识别系统。该系统以线激光传感器扫描的点云数据为基础重构钢轨廓面图像,通过对图像进行中值滤波、阈值分割、图像形态学处理、字符切分等处理操作获得字符图像,然后使用多层神经网络分类器对字符进行了识别。运行结果表明炉号字符一次识别率可达96%,解决了钢轨因污损、锈蚀导致的炉号误读问题,实现了高速钢轨焊前检测过程自动化。

To solve the problem of low efficiency,time-consuming and labor-consuming in the process of therail number identification by manual,this paper proposes a method of non-contact line laser scanning andimaging to identify the rail number,and develops an identification system by using hybrid programmingwith C# language and HALCON. The line laser sensor was used to scan and reconstruct the profile of the railsurface, and the character image was pretreated through the process of median filtering, thresholdsegmentation,image morphological processing,and character segmentation,and the characters of the railnumber were recognized by the Multilayer Perceptron Classifier. The results indicated that the recognition rate ofthe rail number character was up to 96%,which solved the problem of character recognition error due tocorrosion and fouling,and the automatic detection of high-speed rail welding was realized.