Inkjet Printing Character Recognition Base on Deep Learning

目录

声明

Abstract

摘要

Acknowledgments

Contents

List of Figures

List of Tables

Chapter 1INTRODUCTION

1．1Background

1．2Research Significance

1．3Research Development and Current Situation

1．4Research Objectives

1．5Paper Structure

Chapter 2REVIEW OF RELATED THEORIES AND TECHNIQUES

2．1Development of Deep Learning

2．1．1Origin Stage

2．1．2Development Stage

2．1．3Outbreak Stage

2．2 Deep Learning Framework

2．2．1Caffe

2．2．2 Tensorflow

2．2．3PvTorch

2．3Convolutional Neural Network

2．3．1Structure

2．3．2 Classical Network Model

2．4Recurrent Neural Network

2．4．1Overview

2．4．2LSTM

2．5Character Dectection Algorithm

2．5．1Faster RCNN

2．5．2 YOLO

2．6Character Recognition Algorithm

2．7Summary

Chapter 3TEXT RECOGNITION ALGORITHM BASED ON DEEP LEARNING

3．1 Image Preprocessing

3．1．1 Normalization

3．1．2 Gray stretch

3．1．3 Binarization

3．1．4 Hough Transform

3．1．5 Tilt Correction

3．2 Recognition Algorithm Based on CTPN and CRNN

3．2．1 CTPN

3．2．2 CRNN

3．2．3 CTC

3．2．4 Advantages

3．3 Recognition Algorithm Based on Template Matching

3．4 Recognition Algorithm Based on Whole Word Recog-nition

3．5 Summary

Chapter 4EXPERIMENTAL RESUIⅡ1S AND ANALYSIS

4．1 Data Set

4．2 System Design

4．3 Experimental Result

4．3．1 Image Preprocessing

4．3．2 CTPN and CRNN

4．3．3 Template Matching

4．4 Experimental Analysis

4．5 Summary

Chapter 5CONCLUSION AND FUTURE WORK

5．2Future Work

References