Layout Detection and Table Recognition – Recent Challenges in Digitizing Historical Documents and Handwritten Tabular Data

摘要: In this paper, we discuss the computer-aided processing of handwritten tabular records historical weather data. The observationes meteorologicae, which are housed by Regensburg University Library, one oldest collections data in Europe. Starting 1771, meteorological was consistently documented a standardized form over almost 60 years several writers. structure, as well unconstrained textual layout comments and use characters, propose various challenges text recognition. We present customized strategy to digitize combining state-of-the-art methods for OCR fit collection. Since recognition documents still poses major challenges, provide lessons learned from experimental testing during first project stages. Our results show that deep learning can be used detection. However, they less efficient structures. Furthermore, tailored approach had developed characters manual creation ground truth system achieved an accuracy rate 82% heterogeneous handwriting 87% tables.