Image character identification model generation and vertical column character image identification method and device

Abstract

The embodiment of the invention discloses an image character identification model generation and vertical column character image identification method and device. The image character identification model generation method comprises the steps that a rotation line character training sample is generated, wherein the rotation line character training sample includes a rotation line character image and an expected character identification result corresponding to the rotation line character image, and each character unit in the rotation line character image is different from each character unit in a standard line character image for 90 degrees; and a set neural network is trained by using the rotation line character training sample so that an image character identification model is generated. According to the technical scheme, the technical defects that the existing line character image is difficult to be directly utilized to train a vertical column character image identification can be overcome so that efficient identification of the vertical column characters can be realized.

Description

technical field [0001] Embodiments of the present invention relate to character recognition technology, and in particular to a method and device for generating an image character recognition model and recognizing a vertical character image. Background technique [0002] With the popularization of smart phones and portable devices, the use of OCR (Optical Character Recognition, Optical Character Recognition) is more common, it can be used to reduce or replace cumbersome text input, users only need to take an image containing text, OCR technology The text in the image can be automatically recognized for subsequent processing (for example: retrieval and translation, etc.). [0003] The traditional OCR technology includes two categories: the first category is to over-segment the text line to obtain several candidate text areas, and then analyze each candidate text area according to the trained word recognition engine (for example: convolutional neural network, etc.) Identify an...

AI Technical Summary

Problems solved by technology

[0004] However, in an actual environment, the number of text lines is far greater than the number of vertical text. For the recognition of vertical text, it is difficult to directly use the existing text line images to train the model, so it is necessary to collect a large number of vertical text pictures to ensure the training performance of the recognition model, which will consume a lot of manpower and material resources

Images