We found a match
Your institution may have access to this item. Find your institution then sign in to continue.
- Title
基于改进形状匹配的扣件 缺陷检测方法.
- Authors
刘贤华; 邱实; 胡文博; 王劲; 王卫东
- Abstract
Current fastener detection methods based on traditional image processing have problems in fastener positioning, such as inaccurate positioning and large limitations of the positioning methods, and the accuracy of fastener recognition is not high. However, the fastener detection method based on deep learning requires a large number of fastener samples as a training set.The training effect is difficult to guarantee. In view of the above shortcomings, this paper proposed a fastener defect detection method based on improved shape matching. One of the major advantages of this method was that it did not need to perform fastener positioning in advance, and the matching speed was fast and the recall rate was high. The main improvements of this method include the following three parts. 1) Use multi-template matching instead of single-template matching to increase the diversity of templates, thereby increasing the matching recall rate of fasteners. 2) Use shape matching based on HALCON to replace traditional template matching, which is convenient to improve the button. The edge detection efficiency of the parts and the detection robustness are enhanced. 3) The intelligent clipping of fastener pictures and the automatic update of the template library are proposed. Smart cutting can cut out a more neat and standardized data set, and the automatic update algorithm of the template library can be based on the matching the fastener data set dynamically updates the template library. In this paper, the method is verified experimentally with fastener images taken by the track integrated inspection vehicle. The results show that under the condition that the matching threshold is 0.75 and the number of template library fasteners is 32. The improved method for a single image matching time is only 0.18 s, and the detection recall rate reaches 98.15%. The improved method is efficient and intelligent, has high practicability, applicability, and feasibility.And it can meet the needs of daily maintenance and inspection of public works sections.
- Publication
Journal of Railway Science & Engineering, 2022, Issue 7, p1872
- ISSN
1672-7029
- Publication type
Article
- DOI
10.19713/j.cnki.43-1423/u.T20210806