基于计算机视觉的车载钢轨光带异常检测系统研制

Research and manufacture of on-board detection system for abnormal rail light band based on computer vision

基于计算机视觉方法对钢轨光带检测系统进行了研究。首先,试验对比了面阵高速相机+面阵光源、线阵相机+近红外光源和线阵相机+LED线光源3种图像采集方式的效果。其次选取线阵相机+LED线光源方案进行图像采集,获取了可清晰反映轨面光带分布特征的图像数据。最后,设计了鲁棒、高效的光带检测方法,从图像中准确地提取出光带边缘的形状,并通过像素距离与物理距离的换算,得到了光带宽度和中心位置,从而实现了对光带异常的检测。

Detection system for abnormal rail light band was studied based on computer vision. Firstly,three types of image capture methods were compared by tests,which included area array high-speed camera ＋ array light source,linear array camera ＋ near-infrared light source and linear array camera ＋ LED line source. Next,the image capture type of linear array camera ＋ LED line source was chosen. It could acquire clear im ages which could show the distribution characteristics of rail light band. Finally,a robust efficient inspection method was designed,which could extract the edge shape of rail light band,and obtain its width and central position by distance conversion from image pixels to actual positions,thus achieve the goal of detecting abnormal rail light band.