The Chinese Journal of Nonferrous Metals
JOURNAL OF RAILWAY SCIENCE AND ENGINEERING
|第14卷 第2期 总第83期 2017年2月|
(兰州交通大学 自动化与电气工程学院，甘肃 兰州 730070)
摘 要: 针对现有铁路轨道检测识别算法的准确性和鲁棒性不高的问题，提出一种基于直线和双曲线相结合的分段曲线模型实现轨道线的检测、跟踪与验证。本算法首先依据轨道图像的边缘信息，通过多约束条件下的Hough变换初步检测轨道位置，确定轨道线消隐边界并标定近远景区域。然后，在近景区域，采用直线模型实现前方直轨拟合；在远景区域，融合轨间距离、轨道方向和像素灰度等先验知识构造边界置信度函数，设定可漂移窗口搜索算法完成特征点提取，以最小二乘法进行双曲线模型拟合。最后，依据模型切换及窗口搜索策略完成轨道线的跟踪。测试结果表明：该算法不仅较好地解决了弯轨描述问题，而且提高了检测的准确性和鲁棒性。
（School of Automation and Electrical Engineering, Lanzhou Jiaotong University, Lanzhou 730070,China）
Abstract:Aiming at the problem that the accuracy and robustness of the existing railway track detection and recognition algorithm are not high, a piecewise curve model based on the combination of linear and hyperbolic curves is proposed to realize the detection, tracking and verification of the railway line. Firstly, according to the edge information of the track image, this algorithm uses Hough transform under multiple constraint conditions to detect the position of the track, and determines the hidden position of the track, and calibrates the near and far area. Secondly, linear model is adopted to realize the fitting of front straight rail in the near region. In the far area, the track confidence function is constructed fusing the prior knowledge of the distance between the track, the track direction and pixel intensity and so on. At the same time, the drift window search rules are set up to complete the extraction of feature points, and the least squares method is used to complete the hyperbolic model. Finally, based on the model switching and the window searching strategy, the tracking of the railway line is completed. The test results show that the algorithm not only solves the problem of curved track description, but also improves the accuracy and robustness of the detection.
Key words: environmental understanding of the front of the train; rail detection; piecewise curve model; estimation of the vanishing point