The method here is just the use of <
9CBS BLOG Publish a Review Verification Code Recognition> Thinking and Method for identifying 9CBS new features, only part of the signature is valid. However, by improving the algorithm, it should be possible to get a good effect.
The new feature code is as follows:
Preliminary analysis, this signature uses the following ways to increase the identification difficulty:
Not simply archive randomly appearing interference line randomly appeared interference point Different character colors Different characters The distance between the same characters is random, and it is possible to connect together. Using a bevel word, even if it is not connected, it is not possible to obtain one of the characters in a rectangular manner.
Now start processing against the above problems
Let's take a look at it, use the drawing of Windows to save the saved signature in a monochrome bitmap,
Now, change: turning, such as 1, 2, 4 interference is meaningless.
Below is an effect of zooming in 400%
This step can be called monochrome processing, and the code to implement this function in the program is as follows.
Now handling interference points.
The interference point here is relatively small, and it is basically alone, and it is simpler. It is basically judged that there are several points in the 8 locations in a certain point. The following is the effect of processing:
Code:
Treatment here, 1-4 interference is basically removed.
Let's take a character (corresponding to 5 points), I use an imperfect way here, first get the left and right range of each character, then get the upper and lower range of this character, so the rectangular box where the characters are coming out, here The characters used are too gentle, and the next recognition is not difficult;
The effects after the range are as follows:
The code is limited to the space, this part of the code is not posted.
So far, No. 6-7 has not been mentioned, which is part of the difficulty of this signature.
For example, in 9 samples starting, the third line of the second and third lines is too distant after the default monochrome algorithm is complete.
Second character is a great difficulty to the analysis character position (at least the current processing method), below 2 typical examples:
These are all optimized algorithms.
This article is not over, please pay attention to this address: http://blog.9cbs.net/qqwwee_com/archive/2006/02/09/595607.aspx
TRACKBACK: http://tb.blog.9cbs.net/trackback.aspx?postid=595607