Research On Text Recognition Algorithm Of Educational Scene Based On Deep Learning

In recent years,with the rapid development of various information technologies,the acquisition of information through mobile phones or computers has gradually become the mainstream trend in all walks of life.Driven by this trend,the education technology in the education industry makes information technology widely used.At present,the basic form of educational information is mainly text,paper form is the traditional way of text preservation.This kind of traditional way has certain difficulty in storage,searching and sorting.With the integration of information technology,the traditional text is gradually digitized,and the position of computers and handheld terminals in the education industry is gradually rising.The basic form of educational information enables text information to be continuously output and input in the background,resulting in a large amount of information.However,current education and teaching still rely on handwriting and keyboard input as the main way,which limits the flow speed of educational information to some extent.With the continuous development of computer application technology,optical character recognition technology is gradually mature,through the optical character recognition technology can be printed and handwritten text into the text can be processed,greatly improve the speed of text input,reduce the intensity of work and learning burden in education,so as to improve the teaching efficiency.However,at present,there is limited research on the application of optical character recognition technology in the educational scene,especially the large number of text recognition algorithms in the educational scene is not mature,which needs to be further supplemented and explored.Based on the method of deep learning,this paper studies the text recognition algorithm of educational scene.On the basis of the in-depth analysis of the research status of this part,it USES the optical character deep learning algorithm to design the optical character recognition system with the educational scene as the background.Cloud server is used to build a high-speed concurrent and parallel platform,which basically meets the needs of users for text recognition,input and editing.In this paper,the text box is first identified based on YOLOv3 target detection algorithm.YOLOv3 algorithm adopts the framework of convolutional neural network to accurately detect and extract text by behavior unit.This algorithm has high speed and recognition accuracy,and can reach the level of nondestructive testing in practical application,which ensures that the detection and extraction of the Chinese group in the education scene are in a state of high efficiency and high speed,and is the basic guarantee for the construction of thetext recognition algorithm in the education scene.Secondly,the text in the extracted text box is recognized based on the convolutional loop neural network.The cyclic neural network takes into account the context of the text and,combined with the convolutional neural network,can process the text input of variable length.The algorithm can recognize the text sequence flexibly,and has a high accuracy,for the image without serious noise,can reach the completely correct level.Finally,based on the CPU and GPU cloud server,the identification system is built,and the multi-gpu distributed framework is adopted,independent CPU management and distribution deployment is adopted,which has achieved the effect of high concurrent real-time response.The system can be used in the real teaching scene to achieve text acquisition and editing anytime and anywhere,and to achieve the purpose of promoting educational informatization.