March 8
Open source OCR library and opensourced OCR Library
I suddenly remembered how to create a forum posting machine, but now most of them require image authentication, and an OCR library is required to recognize images. So I found it on the Internet and there are many implementations:
- Ore, OCR easy, http://lem.eui.upm.es/ocre.html
- Clara OCR, http://directory.fsf.org/claraocr.html
- Ocrad, based on a feature extraction, http://directory.fsf.org/ocrad.html
- Gocr, http://sourceforge.net/projects/jocr
- Ocrchie: Modular optical character recognition software,
Http://http.cs.berkeley.edu /~ Fateman/kathey/ocrchie.html
- There are still many ongoing projects on sf.net, but most of them are still in the hypothetical stage.
Now the problem is to test these libraries. However, I guess these OCR attacks are still good for Chinese websites. Images like MSN will definitely be ineffective.
Download these programs and test them separately,
Gocr and ocrad can be compiled and run in Windows with slight modifications. Gocr uses vc6 and ocrad uses GCC.
Ocres, ocrchie, and Clara OCR must be compiled by X Window or GTK.
After testing Sina, Baidu, and Taobao, they all have more or less interference noise, so they are basically unidentifiable.
It is best to use gocr to correctly identify three of the four numbers.
It seems that this recognition method is basically not feasible before image preprocessing.
It's over before it starts.