The Research On Character Recognition Technology Based On Deep Learning

Deep learning has been widely used in target recognition,speech content recognition and so on.Up to now,the topic of deep learning has gone deep into the research of various industries,and has gradually been cited in the actual life production,and its application is becoming more and more extensive.With the demand of production and the rapid improvement of computer data processing speed,character recognition is becoming more and more popular in actual production and daily life,but,The traditional pattern recognition and feature extraction algorithms can not meet the production needs of real-time and accuracy;At the same time,because of the development of deep learning,the character recognition technology based on deep learning has the advantage that the traditional recognition algorithm can not compare.Therefore,more and more scientists study the character recognition technology based on deep learning.The development of this paper mainly starts with the character recognition technology based on deep learning,aiming at the demand of print character recognition in current production,relying on the security door character recognition project.This paper studies the character recognition technology in complex background,which can be divided into three aspects: image preprocessing,character target location and character content recognition based on deep learning network.In the aspect of character target location,we make full use of the detection and tracking to get the character target information,combining with the water-filled algorithm and the traditional image preprocessing technology,we propose an algorithm which can accurately locate the character position and determine the character region.Remove the area outside the character,eliminate the influence of the environment background,improve the robustness of the existing algorithms.In the aspect of character content recognition,it is mainly divided into two aspects: the reproduction of the basic network framework and the promotion and optimization of the main network LeNet5.First,we deeply understand the currentmainstream recognition algorithm for character recognition.At the same time,it focuses on exploring the thought principle of the application of deep learning in character recognition,and carries on the training of character picture based on the digital recognition frame LeNet5 and mnist database,and the caffe deep learning platform.In order to improve the accuracy and real time of character recognition,the LeNet5 network is optimized and improved on the original framework.