Abstract:As people increasingly aspire to intelligent and convenient lifestyles, the demand for text extraction in complex scenarios is growing. OCR(Optical Character Recognition) provides two methods for text extraction: text detection combined with text recognition and end-to-end recognition. However, due to the diversity and complexity of scene text, there is currently no text detection and recognition application on the market that performs exceptionally well in all scenarios. A common solution to this problem is to construct targeted models and collect targeted datasets for specific usage scenarios. In this paper, based on deep learning methods, we focus on the tasks of label recognition and book information retrieval for library books. We construct our own dataset, design, and build a text detection and recognition network. The model is deployed in an Android application to enable scanning and recognition of book labels. Additionally, it can retrieve relevant information about books corresponding to the labels by querying a database.

Research and Development of Book Query and Management System Based on OCR.