摘要
研究视觉图像双轨轨道平行性准确检测问题。铁轨由于长期的磨损使得轨道不同的部位存在一定的角度和径向细微畸变,造成平行性误差。传统的图像的铁轨平行性检测方法,多是基于平行特征的有效匹配完成检测的。由于畸变误差极其细微,很容易把铁轨的角度和径向细微畸变当成一般的噪声误差混淆,而铁轨检测对精度要求极高,误差会降低轨道平行性检测的准确率。提出了一种图像误差弱化消除技术的双轨轨道平行性检测算法。利用差异矩阵变换方法,对采集到的高光谱铁轨图像进行空间变换,建立Arnold畸变配准误差消除模型,准确区分过滤畸变误差。实验表明,改进方法提高了双轨轨道平行性检测的准确率,为铁轨检测优化提供了参考。
Research the accurate parallelism detection of double track based on visual images. The certain angle and radial slight distortion exist in different tracks positions due to the long-term wear, caused by the errors in paral- lelism. An image error eliminating technology for double-track rail parallelism detection algorithm was presented. It made use of the difference matrix transformation method, and conducted space transformation of hyperspectral tracks images. Arnold distortion registration error elimination model was built to accurately distinguish filter distortion error. Experiments show that the method improves the double track parallelism detection accuracy, and provides a reference to the optimization of tracks inspection.
出处
《计算机仿真》
CSCD
北大核心
2012年第7期318-320,392,共4页
Computer Simulation
关键词
双轨轨道
平行性检测
差异图像配准
Double tracks
Parallelism detection
Fuzzy image registration
