Recognition of automobile state. numbers?

Question

Pattern recognition

There is a task to make a diploma on such a topic: we get images from the camera, we recognize the numbers of cars. I understand that it's difficult. To what side to look?

Answer 1

I repeat once again: I want to write a thesis on this topic, I need not finished products, but recognition algorithms

Answer 2

blo , thanks, I'll take a look!
uups , this is a finished product, I need recognition technology

Answer 3

You should separate the tasks: first you need to select the image of the license plate from the image, and then recognize it. For starters, look in the direction of the perceptron, write in a personal or ICQ, I will share detailed information
PS Similar work has already been done by prof. Turlapov from UNN, is there any point in repeating?

Answer 4

article and materials are not available (

Answer 5

OpenCV

Answer 6

www.rossi-potok.ru/

Answer 7

Look for APK "Hurricane", used to be called Cobra. Everything has been working for a long time.

Answer 8

Look at cuneiform algorithms and source codes. It might help.
By the way, the leptonica library is used there - also a pretty good thing to learn.

Answer 9

Now I am writing a program just for these purposes .
The algorithm is as follows:
1. Convert to semitones
2. Apply the Sobel or Canny operator (preferably canny)
3. Threshold binarization and search for a rectangle (usually the autonumber is filled by 23% with approximately black dots)
4. Number segmentation - splitting into characters
5. Recognition itself - here either neural networks or something else
. Analysis of the contrast of the image is quite good for localizing the number, then you don’t have to write the search for rectangles

Answer 10

I think not to bother and take the finished product