Test paper handwritten English character recognition method and system based on deep learning

Abstract

The invention provides a test paper handwritten English character recognition method and device based on deep learning, and belongs to the technical field of image recognition. The method comprises the steps of acquiring a to-be-recognized test paper image; cutting the acquired image to obtain a word image in the test paper image, and identifying the word image by using a trained neural network model based on an attention mechanism to obtain a word identification result; wherein the step of cutting the acquired image specifically comprises the sub-steps of carrying out binarization operation on the test paper image, cutting text lines in the test paper image, and cutting English words in the text line image; according to the method, the English text line segmentation method based on dynamic line segmentation and the word sequence recognition method based on the attention mechanism are adopted, a good segmentation effect is achieved on the bent text, and the word recognition accuracy iseffectively improved.

Description

technical field [0001] The present disclosure relates to the technical field of image recognition, in particular to a method and system for recognizing handwritten English characters in test papers based on deep learning. Background technique [0002] The statements in this section merely provide background information related to the present disclosure and may not necessarily constitute prior art. [0003] The recognition of handwritten English characters in the test paper includes two steps. One is to cut out the English words in the test paper, and the second step is to recognize the cut out word pictures and transcribe them into word texts. [0004] Word segmentation is usually performed in two steps: first the segmentation of the text line, and then the segmentation of the words in the text line. [0005] (1) The basic idea of ​​​​segmentation of text lines is to find the starting point of segmentation first, and then find the dividing line with methods such as connecte...

AI Technical Summary

Problems solved by technology

This method extracts visual features, which are highly interpretable, but the process of extracting features is relatively complicated.

[0009] (2) The basic idea of ​​the recognition method based on statistical features is to extract the moment feature of the character or word image, perform K-L transform or Fourier transform on it to generate global features such as feature vectors, and transform the local grayscale of the image , projection, gradient and other local features for classification and recognition, but the interpretability of such features is weak

[0010] (3) The recognition method based on deep learning takes word recognition as a retrieval problem. The method of recognizing words through retrieval is also called Word Spotting. The general idea of ​​this method is to map word images and word texts to the same low-dimensional Space, use image features to search in the word text feature library during recognition, and select the word text with the highest matching degree as output. Although this method has high recognition accuracy, it cannot recognize words that are not in the word library

Images