A Method of Trackside Kilometer Post Identification Combined with YOLOv3 Model

In view of poor anti-interference ability and weak generalization of the present trackside identification method, it is unable to adapt to complex background fluctuation along railway. Therefore, an image recognition method based on YOLOv3 was proposed, which could still ensure good recognition accu...

Full description

Saved in:
Bibliographic Details
Main Authors: QIU Xinhua, WANG Wenkun, JI Yuwen, LI Jia
Format: Article
Language:zho
Published: Editorial Office of Control and Information Technology 2020-01-01
Series:Kongzhi Yu Xinxi Jishu
Subjects:
Online Access:http://ctet.csrzic.com/thesisDetails#10.13889/j.issn.2096-5427.2020.05.002
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:In view of poor anti-interference ability and weak generalization of the present trackside identification method, it is unable to adapt to complex background fluctuation along railway. Therefore, an image recognition method based on YOLOv3 was proposed, which could still ensure good recognition accuracy in the face of different illumination, complex background and different forms of image. Firstly, a convolutional neural network is used to learn image features and obtain network parameters of each layer. Transfer learning is adopted to obtain possible rectangular region of kilometer post based on the trained network parameters. Then, pattern recognition is used to extract, segment and recognize character region of kilometer scale, and finally output digital information of kilometer scale to the control system. Experimental results show that the identification time is about 0.04 s and the positioning accuracy is within 0.5 m by this method, which meets the accuracy and real-time requirements of position correction for rail flaw detection vehicle system.
ISSN:2096-5427