Summary
International Conference on Emerging Technologies for Communications
2023
Session Number:P2
Session:
Number:P2-1
A Multi-Stage Solution for Scene Text Detection and Recognition
Chien-Hao Tseng, Meng-Wei Lin, Jyh-Horng Wu,
pp.-
Publication Date:2023/11/29
Online ISSN:2188-5079
Summary:
In this study, a text detection and recognition system was developed for surveillance images, focusing on Taiwanese street views. The developed method was verified using images captured by surveillance cameras under various weather conditions. This study has two main contributions. First, it established a multilayer deep learning architecture to realize multilingual sign text recognition and automatic text location identification in images captured at various angles while driving. Second, it employed a string boundary detection model to ensure accurate interpretation of the horizontally and vertically oriented text on Taiwanese signs. The verification results revealed that the proposed system achieved high accuracy in character recognition for dashcam images captured under different weather conditions. Accordingly, the proposed system can provide an intelligent solution for surveillance image analysis.