计算机工程与应用 ›› 2018, Vol. 54 ›› Issue (13): 137-141.DOI: 10.3778/j.issn.1002-8331.1703-0104

• 模式识别与人工智能 • 上一篇    下一篇

有砟轨道扣件缺失识别算法的研究

张新峰,王明玉,张春梅   

  1. 北京工业大学 信息学部,北京 100124
  • 出版日期:2018-07-01 发布日期:2018-07-17

Research on ballasted track fastener loss recognition algorithm

ZHANG Xinfeng, WANG Mingyu, ZHANG Chunmei   

  1. Faculty of Information Technology, Beijing University of Technology, Beijing 100124, China
  • Online:2018-07-01 Published:2018-07-17

摘要: 针对有砟轨道扣件缺失问题,提出了基于fast PCA 和bag of words的两级分类扣件图像缺失识别算法。考虑到算法的可移植性和拍摄角度的不同,图像中钢轨和轨枕的方向并不一定是垂直的,单扣件定位识别具有局限性,提出采用双扣件图像作为样本。通过第一级分类器判断扣件是否缺失,进而得到扣件的具体位置。通过第二级分类器判断缺失扣件的类型,进一步得到扣件缺失的数目。该识别算法可达到较好的识别效果,有效地解决了拍摄角度不同的情况下有砟轨道缺失扣件的检测问题。

关键词: 扣件缺失识别, 词袋模型, fast PCA, 两级分类算法

Abstract: In view of ballasted track fastener missing problem, the algorithm for two level classification of fastener image recognition based on fast PCA and bag of words is proposed. Considering the portability of the algorithm and different shooting angles, rails and sleepers in the image direction are not vertical and single fastener positioning identification has limitations, the double fastener image sample for identification is put forward. Whether fasteners are absent can be detected by the first level classifier, and then the positions of the fasteners are obtained. The types of missing fasteners can be further judged by the second level classifier, and the more accurate number of missing fasteners can be acquired. The proposed recognition algorithm can achieve good recognition effect, effectively solve the detection of ballasted track missing fasteners under different shooting angles.

Key words: fastener identification, bag of words, fast PCA, two level classification algorithm