【文献阅读】Hybrid model for Chinese character recognition based on Tesseract-OCR

总结：openCV(image preprocessing)+KNN(phrase processing)+Tesseract-OCR engine
个人感觉此篇论文质量不高，实验细节未论述，实验结果没有统计分析，言辞重复，存在低级错误

Introduction

Chinese OCR is more difficult

• The number of English letters is only 26. But the number of Chinese characters that used commonly are about 2,500.
• the strokes of Chinese characters are complex and similar.
• The differences between the different fonts of Chinese are large.

OCR engines

Tesseract-OCR engine

• the first OCR engine, supports more than 100 languages (tesseract-
  ocr/tessdata, https://github.com/tesseract-ocr/tessdata).
• The OCR engine of Tesseract- version 4.0 uses Long Short-Term Memory (LSTM).
• In the Tesseract-OCR Simplified Chinese language library,the character recognition of separate words is based on the feature of standard Chinese characters.

OCRopus

• also a OCR engine based on LSTM.

Ocular OCR engine

• mostly uses the recognition of historical artefact.

Swift OCR

• is a simple and fast OCR, Written in Swift.

Simple-ocr-openCV

is a simple python OCR engine based on OpenCV and NumPy

Background

Process of OCR

The main work of this study includes image preprocessing and phrase processing.

2.4.1 Image preprocessing

The methods of image preprocessing include binarisation, noise reduction, image tilt correction, and the like

3 OCR hybrid recognition model

3.1 Image correction

3.2 KNN phrase detection and correction