Vehicle number plate recognition

The technology works best with digital images. Infra-red and visible spectrum cameras can be used, though infra-red provides suitable images under a wider range of conditions. The text can be compared with databases containing vehicle number plate values used by law enforcement agencies to generate alerts for unregistered or stolen vehicles as examples.
_______________________
#-Link-Snipped-#

well you can use a simple webcam to recognise them,for that you need to go deep into image processing,the idea is to break the whole image into smaller cubes and compare those cubes with some reference cubes and recognise but accuracy is low in this type of systems

There are two approaches possible. The first step in both the methods is obviously extracting the number plate from the image and extracting all the characters from the number plate. It'll be better if you frame each character into a separate image.

Then, the second part is identifying each character. For this, you can use sample images, compare your images with the test images and decide which one is the closest match.
The other method is based on neural networks, wherein you train the program to identify how each character is, and then the program will identify the characters based on the initial training. The accuracy will improve as more and more images are used.

but letters and numbers are written in many font sizes,so it will require constant leraning system,dont ya think thats bit complex???

Sorry if this is too basic, but to the best of my knowledge we could have something like pattern recognition system in place, right?

definately,but what should be the quality of image,since we need some reference value!!

ignitedindiani am working on a video processing project.

a video cam captures all the vehicles in road.

i have to identify speeding vehicles and get their number from number plate.

which platform is good? how do i start??

I have successfully connected my webcam in matlab. is that the way to go?? what next???

MATLAB is my development tool of choice for technical development (and it would be ideal for this project), but if you are more comfortable with something else, then I suggest using that instead.

Your basic next steps are:

1. Isolate the license plate within the image
2. Prepare the isolated image (improve contrast, clean noise, etc.)
3. Isolate individual glyphs ("characters")
4. Extract relevant features for each glyph
5. Recognize each glyph in turn

The first 3 items are image processing functions, while the last 2 are machine learning tasks, and will require some sort of classifier (linear discriminant, neural network, etc.).


-Will Dwinnell
<a href="https://matlabdatamining.blogspot.com/" target="_blank" rel="nofollow noopener noreferrer">Data Mining in MATLAB</a>

sauravgoswamibut letters and numbers are written in many font sizes,so it will require constant leraning system,dont ya think thats bit complex???

Unless the license plates being recognized are unusual, I do not think this will be much of a problem, since license plates within most countries are fairly consistent about typeface. I suspect that the larger challenges will be:

1. Accurately isolating the license plate (not as easy as it sounds, even if the vehicle is standing still).
2. Being able to ignore extraneous items, whether they belong on a license plate (background images, non-character symbols) or not (dirt, bugs, etc.).
3. Dealing with poor or inconsistent lighting and interfering weather (rain, snow).

I may be wrong, of course, but my expectation is that, if the image processing part of such a project is performed successfully, then the recognition part should not be too difficult.


-Will Dwinnell
<a href="https://matlabdatamining.blogspot.com/" target="_blank" rel="nofollow noopener noreferrer">Data Mining in MATLAB</a>

silverscorpionThen, the second part is identifying each character. For this, you can use sample images, compare your images with the test images and decide which one is the closest match.

This is conventionally called 1-nearest neighbor modeling.

silverscorpionThe other method is based on neural networks, wherein you train the program to identify how each character is, and then the program will identify the characters based on the initial training. The accuracy will improve as more and more images are used.

While these are two possible candidates, there are many others, such as tree induction, logistic regression, discriminant analysis, etc. Selecting among these means choosing from among various trade-offs, such as recall accuracy vs. recall time.


-Will Dwinnell
<a href="https://matlabdatamining.blogspot.com/" target="_blank" rel="nofollow noopener noreferrer">Data Mining in MATLAB</a>

Dual camera systems are becoming popular here, this gives twice the odds of successfully getting the plate (in states that require a front and rear plate) plus you also get a shot of the driver, which many areas require as legal proof it was you driving your car, not a friend.

Lol, now I'm curious.. is there a law regarding the prohibition of wearing masks while driving?

No, only while making withdrawls at a bank... =)

Today, there is a news in mirror newspaper (Pune, INDIA) that a guy was driving a vehicle having Dubai passing, in India. So I was wondering if this project can help detecting non-listed vehicles by detecting number on number plate.

Hi friend,
Do you have code of license plate recognition from video in matlab. Can you please send us and I will definitely support for that

Guys,
i m doing the very same project in final year...

I have segmented the characters from the number plate and looking forward to complete the matching the extracted characters with database characters... i am unable to design any algo which provides comparison of the test characters(of number plates) with the database characters.... anybody eith any damn solutions???? eagerly waiting for any help to appreciate

Thanks in advance......
😀
Chirag