Ticket Text Detection and Recognition Based on Deep Learning
Xiuxin Chen, Zhijing Lv, Dongdong Zhu, Chongchong Yu · 2019
In traditional OCR (Optical Character Recognition) technology, a series of complex preprocessing of the image is needed in the process of detecting and recognizing of the ticket texts. This paper proposes a method based on deep learning in order to detect and recognize the ticket texts automatically. This method consists of two parts: ticket texts detection and recognition. First, CTPN text detection model is used for detect the horizontal texts of the ticket. Then, cropping the detected text image by its coordinate position, resizing them and changing to gray. After that, improved CRNN model which adding self-attention in feature extraction layer is used for text recognition. Experiments show that the proposed method can automatically detect and recognize the texts in the ticket. It not only solves the problem that the traditional method takes a lot of time and waste too much manpower to extract features, but also has good generalization ability and scope of application.