怪我咯2017-04-17 13:14:29
I have done license plate recognition before, and verification code recognition should be similar. It is generally divided into four steps:
Preprocessing: Grayscale and binarize the verification code image
Character cutting: Character cutting of binary images
Feature extraction: extract the features of each character and generate a feature vector or feature matrix
Character recognition: Compare the feature vector or matrix of the sample library and the cut characters, and use the recognition algorithm to identify
迷茫2017-04-17 13:14:29
"Detailed Explanation of Visual C++ Digital Image Pattern Recognition Technology" Buy this book and read it
ringa_lee2017-04-17 13:14:29
First read in the image, which is a matrix, then binarize the matrix, and then operate on the matrix. I forgot the specific algorithm. Each value corresponds to a different number and letter, just write a switch.
伊谢尔伦2017-04-17 13:14:29
OCR is the way
Another way is to connect to a paid coding platform. The platform will connect to real-person coding. You only need to send the verification code picture to the platform (usually you can report the error if you type it wrong)
There are many such platforms, so I won’t post the detailed address. Just search and you’ll find them
PHP中文网2017-04-17 13:14:29
Binarized road.
Looking at the Ministry of Railways’ mythical beast-level verification code, I feel a little tired.
