Date Recognition of Handwritten Chinese Documents Based on Object Detection and Character Classification

With the increasing degree of informatization in human society, paper documents are increasingly stored and processed in the form of electronic documents. People often want to extract key information such as dates in the process of digitizing paper documents. However, for text images with severe linking, extracting key information is very difficult. This paper proposes a date recognition framework for handwritten Chinese documents. First, YOLOv9 is trained to detect dates in images. Then single characters are segmented according to the pixels and character edges of the date area. And finally, Support Vector Machine (SVM) and Convolutional Neural Network (CNN) are used to classify handwritten Arabic numerals and handwritten Chinese characters and recognize handwritten numerals. At the same time, the date area in the text is located and recognized according to the date writing habits in Chinese documents. The method in this paper was experimented on 150 different handwritten Chinese text images with dates, and the recognition accuracy of dates in images reached 91.3%. The method proposed in this paper effectively achieves the localization and recognition of dates in handwritten Chinese character documents.