Hi,
This is a preliminary idea for a BE project. I would appreciate some feedback on feasibility and pointers to resources.

The idea is to place a digital camera on the entrance of a car parking lot, with stripes on the entrance indicating where the car should aim to enter it (this would aid in a generally uniform picture). From the picture I need to extract the image of the person driving and the license plate of the car.

The idea is an administrator types a car number and sees the date/time this car has entered and who was driving it. Please note I do not need identify the person just crop the face out of the picture.

License plate extraction can be done in 2 stages, stage one I convert the image to grayscale and then to black and white. Crop it and then somehow extract the number from the number plate. I have a fair amount of idea how to do the grayscale, black and white stuff and the cropping, however I am clueless on how to approach the image to number conversion.

My approach would be to develop and implement an algorithm on C# with a custom image. Then tackle the DSP thing. Because if I get unexpected results I would be unsure where the problem lies: programming, my understanding of the programming, my algo, the processor etc. so I would like to test my algo and then tackle the DSP.

My level with DSP is beginner's, however I am pretty good with programming microcontrollers. So my questions are:

- Is my idea feasible? Am I overlooking something?
- Thoughts on my approach to the project (C# and then DSP)
- How should I proceed with the DSP selection? What are the features that I will require?
- Pointers to image to number conversions.

all thoughts are welcome ...

Thanks!

 

Hello,

Extracting the characters of a licenceplate would be more a job for OCR (optical character recognition).
http://en.wikipedia.org/wiki/Optical_character_recognition

Bertus

 

I know this is done by some systems used for traffic control and systems in police cars.

 

I know this is done by some systems used for traffic control and systems in police cars.

Yes, I saw a program starring two police officers.)). Thay had new technology. When they were sitting by the side of the road with their lazer-speed-measure device, they also took a picture of the licence plate. The picture was run thru OCR, and linked to a database. The database contained information on owner, insurance and other autorelated fees.

So when they pulled the car over, they had almost all information at hand.

I think Mythbusters also had something going on with this.

I'm sure you can do it.

 

I'm sure you can find some OCR code around somewhere. Otherwise thats what you need to research, Optical Character Recognition. This takes and image and makes text out of it. There are many different programs that do this and some are free so I would imagine there would be some opens source code about doing it.

 

You will most likely want a sharpening process to find edges before going through the OCR. The boundaries of the LP and face would need to be found.

here's a couple links that should get you started with image processing.
http://www.imatest.com/docs/sharpening.html
http://www.imagemagick.org GET THE SOURCE CODE

 

Thanks everyone .... I am looking into OCR ....

imagemagick has extensive and highly readable tutorials ..... I am going through them right now ...

All these pointers indicate that what I need is not a DSP but a computer ... since the project is for electronics it seems I will have a difficult time convincing the management .....

 

Nah it can and would be done with a dsp or high powered uC. But its just way way way simpler on a computer.