1. Transformer text recognition with deep learning algorithm
- Author
-
Ye Chen, Wenjiao Xu, Zhengyu Yang, Hongchun Shu, Zhihu Hong, and Mingshuai Dong
- Subjects
Computer Networks and Communications ,business.industry ,Computer science ,Network on ,Deep learning ,020206 networking & telecommunications ,02 engineering and technology ,Text recognition ,Electric power system ,0202 electrical engineering, electronic engineering, information engineering ,020201 artificial intelligence & image processing ,Test performance ,Artificial intelligence ,Sensitivity (control systems) ,business ,Algorithm ,Nameplate ,Transformer (machine learning model) - Abstract
The transformer is a vital equipment in the power system, which is used in large quantities and replaced frequently in industrial projects. Therefore, it is essential to find an efficient automatic detection and recognition method for the text information of the transformer nameplate. At present, the text information of the transformer nameplate is collected manually, which is inefficient. On the other hand, the complex text features of transformer nameplates are a challenge to the existing text recognition algorithms. Therefore, we propose a two-stage network based on deep learning to recognize the nameplate text content automatically. At the same time, we establish a transformer nameplate dataset due to the particularity of the data characteristics of the transformer nameplate. The dataset is used to train our network to improve its sensitivity of the transformer nameplate information. The experimental results show that our model achieves a recognition accuracy of 71% in the transformer nameplate dataset. The test performance of our network on the transformer nameplate dataset is comparable with the state-of-the-art text recognition algorithms.
- Published
- 2021
- Full Text
- View/download PDF