Table area identification model generation method and device and table positioning method and device

Abstract

The invention discloses a table area identification model generation method and device and a table positioning method and device, and the method comprises the steps: obtaining a plurality of documentpages comprising at least one table, and converting the document pages into an image which serves as a sample image; obtaining the distance between each first unit area in the sample image and each table title, and obtaining whether each first unit area in each sample image is an identifier of any one table area or not; and taking the sample image, the obtained distance and the obtained identifieras a training sample set, iteratively training a pre-constructed deep learning network model, and generating a table region recognition model for recognizing whether the table region is a table region or not. As a large number of training samples are adopted to train the deep learning network model, the generated table area identification model is a stable model, the success rate of table identification is improved, and the smooth proceeding of the subsequent table positioning process is ensured.

Description

technical field [0001] This article relates to image processing technology, especially a method and device for generating a form area recognition model and a form positioning method and device. Background technique [0002] Information analysis often needs to extract information such as text, tables, and pictures from documents, such as Portable Document Format (PDF), and to achieve information extraction, it must first realize identification and positioning, because the storage of data in PDF uses It is an unstructured data storage method, so the identification of information is more complicated, especially the identification of tables. [0003] In related technologies, table recognition is often based on the fact that the table and its non-table fonts are not the same or that the character stream of the table is aligned in columns. [0004] However, this method can only be carried out based on special circumstances. When the form does not have the above characteristics, i...

AI Technical Summary

Problems solved by technology

[0004] However, this method can only be carried out based on special circumstances. When the form does not have the above characteristics, it still cannot be recognized, so the success rate of form recognition is very low, which further affects the positioning of the form

Images