Character recognition method based on an attention mechanism and linkage time classification loss

Abstract

The invention discloses a character recognition method based on an attention mechanism and linkage time classification loss. The character recognition method comprises the following steps that S1, a data set is collected; s2, carrying out preprocessing such as scale zooming, gray level conversion and pixel normalization on the image sample; s3; processing the tag sequence of the sample, includingfilling, coding and word embedding; s4, constructing a convolutional neural network, and performing feature extraction on the text image processed in the step S3; s5, encoding the features extracted in the step S4 by using a stacked bidirectional recurrent neural network to obtain encoded features; s6, inputting the coding features obtained in the S5 into a connection time classification model tocalculate a prediction probability; and S7, calculating weights of different coding characteristics by using an attention mechanism to obtain coded semantic vectors.

Description

technical field [0001] The invention belongs to the field of optical character recognition, in particular to a character recognition method based on attention mechanism and connectionist time classification. Background technique [0002] With the popularization of intelligence and mobile terminals, the semantic information of natural scene images plays an increasingly important role in the fields of automatic driving, intelligent transportation, and visual assistance. [0003] To solve the problem of text image recognition in natural scenes, Strokelets obtains stroke features in text by clustering image blocks, uses HOG features to detect characters, and combines random forest classifiers to classify characters. The PhotoOCR system uses the HOG feature classifier to score the candidate results obtained by segmentation, and then combines the BeamSearch algorithm of the N-gram language model to obtain the candidate character set, and finally uses the language model and shape m...

AI Technical Summary

Problems solved by technology

The above methods still have the following problems: 1) The recognized features rely on manual definition, and the artificially defined features are difficult to capture the deep semantics of the picture, and the production is time-consuming and not very versatile

2) Recognition based on a single character will be out of context, which can easily lead t

Images