Knowledge Model Based Intelligent Visual Identification of Catenary Defects
As high speed rails increase in mileage and operational intensity, the use of onboard video to conduct normalized and intelligent inspections of the service status of key catenary facilities under the condition of limited operation and maintenance skylights is an important demand problem that needs...
Saved in:
| Main Authors: | , , , , , |
|---|---|
| Format: | Article |
| Language: | zho |
| Published: |
Editorial Office of Control and Information Technology
2021-01-01
|
| Series: | Kongzhi Yu Xinxi Jishu |
| Subjects: | |
| Online Access: | http://ctet.csrzic.com/thesisDetails#10.13889/j.issn.2096-5427.2021.06.012 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | As high speed rails increase in mileage and operational intensity, the use of onboard video to conduct normalized and intelligent inspections of the service status of key catenary facilities under the condition of limited operation and maintenance skylights is an important demand problem that needs to be solved urgently. Based on the development process of the C3 system, this paper sorts out the specificity of catenary video inspection task such as full scene, high perspective, micro target, weak change and few samples, and points out the strengthening effect of expert knowledge and typical cases on data model. Considering the completeness of hanger fault samples, the accuracy of defect recognition and the high speed of data screening as the research objects, an intelligent visual identification method for catenary defects based on knowledge model is proposed. In the proposed method, numerical unbiased simulation of hanger defects is performed based on expert experience and typical cases, abnormality recognition of hanger is achieved based on multi-scale visual attention, and rapid state screening is carried out based on forest parameters. Field experimental data shows that the use of the proposed visual identification method can increase the identification accuracy by at least 3%. |
|---|---|
| ISSN: | 2096-5427 |