Table or text line processing method and device, bill processing method and device and storage medium

Abstract

The invention belongs to the technical field of document recognition, and particularly relates to a table or text line processing method, a bill processing method and device and a storage medium. The method mainly comprises the following steps: acquiring points on two opposite boundaries of a to-be-processed object, and respectively counting the points into a first boundary point set and a second boundary point set; obtaining a first boundary curve according to the first boundary point set, and obtaining a second boundary curve according to the second boundary point set; source control points with the corresponding number are extracted from the first boundary curve and the second boundary curve respectively; aligning the source control points to obtain corresponding target control points; obtaining an interpolation function according to each source control point, the target control point and the minimum energy function; and processing the text line region corresponding to the to-be-processed object by applying the interpolation function. The thin-plate spline interpolation technology is applied to processing of curved text lines or curved tables, and the problem that the correction effect is limited by training samples when a machine learning method is used for correcting curved texts in the prior art is effectively solved.

Description

technical field [0001] The invention belongs to the technical field of document identification, and in particular relates to a form or text line processing method, a bill processing method, a device and a storage medium. Background technique [0002] Over the past few years, computer vision research has largely focused on Convolutional Neural Networks (often shortened to ConvNets or CNNs). Since convolutional neural networks can use local operations to perform hierarchical abstraction of representations, convolutional neural networks have achieved better performance on a wide range of classification and regression tasks. The success of convolutional architectures in the field of computer vision is mainly driven by a key design idea: first, CNN takes advantage of the 2D structure of the image, and since pixels in adjacent regions are usually highly correlated, CNN does not need to use all One-to-one connections between pixel units (which most neural networks do), instead gro...

AI Technical Summary

Problems solved by technology

[0005] The present invention proposes a form or text line processing method, bill processing method, device, and storage medium. The thin plate spline interpolation technology is used in the processing of curved text lines or curved forms, which effectively solves the problem of using machines in the prior art. When the learning method corrects curved text, the correction effect is limited by the training samples

Images