Hungry for OCR text recognition technology support

zhoushaoli

I got the title of the graduation project "OCR Character Recognition System Design" some time ago, and my mind was at a loss~~~
It is required to realize the conversion of graphics and text, which mainly refers to the extraction of text information in common images (.bmp|.jpg|.gif|.pdf) into common documents (.txt|.doc|), and the recognition efficiency is required to be 80%. Optional development tools.
  I've never touched this stuff before, so I don't know how to start.
  Hope that all the experts will not hesitate to impart relevant knowledge and techniques.

zhoushaoli

English and Chinese character coding, please let me know.

rulazio

You will be awesome if you can achieve it, but you won’t be able to achieve it.

ouyuyu

I read all the information in the font library myself. The font, size, and format of the words in the picture should be exactly the same, and then click one by one to find out what percentage matches, and then give a threshold. Those exceeding this threshold are considered to be matched.
The above is the most ideal state. Write a few words on the drawing board by yourself, and the position and color should also be fixed. The undesirable state is that you have to do text positioning, binarization, and fuzzy recognition. . .
If it allows you to recognize the scanned image, it can clearly tell you that the recognition rate is impossible to achieve.

ayasakura

OCR recognition itself is a technical difficulty
Your teacher really embarrassed you. Tell him you can't do it
He will definitely lower your requirements

terminator

Easiest to identify numbers

Wavelet algorithm, let’s take a look, but the accuracy depends on the input

zhoushaoli

But she told me that she would not. Let me figure it out by myself, ~~`It’s one step at a time

zfy010317

I think recognition should consider algorithms on the one hand, and some constraints on the other hand.

maggic

Let you write analysis and design, that is, you don't need to implement it first, and you have not yet reached the stage of writing code. .

binlee259

The current school is very nb