How to auto adjust ROI for License Plate recognition system with vision assistant

[Min1992]

Hi...
I am trying to write the license plate recognition system for the car park.. I use vision assistant for it....I am a beginner in vision assistant and also in labview...

I just learn about the OCR Training and OCR First Example.vi. (Barcode example)... From that example, I try to write for License plate..
I can able to write for one License plate... but it has problem... the cars are parked in different directions and different structures.. i know my program will not work if the images of license plate from camera are not capture the same distance and the same direction. I try to draw the bigger ROI but if ROI is very big, it cannot recognize the character. For this problem, i have attached the image below.. how can i do for the roi in that image to be able to detect the characters?... Do u have any ideas?

[Vision_Armenia]

Hi there.

It is not a simple task. Moreover, it is the key point in ALPR.

So LabVIEW gives you ability of learning and recognition of license plate, but not detection. You should programmatically find license plate you are looking for, find ROI, then give this ROI to OCR. I could recomend you to use edge detection techniques to find the ROI.

Regarding Vision Assistant - you could start with it, but for the final program, you should go to the lower layer (LabVIEW), and add some mathematics there. Try to divide your program into two parts, one let say will be the ROI finding part, another OCR part (which you already did).

[Min1992]

Hi, Thanks for ur reply....Yes. That task is really tough for me....I will try to learn more and more on labview and vision assistant. I don't have any knowledge on labview and start doing it now...I don't know which part i have to learn and where i can learn about it...as this is my final year project, i have to finish it on time...I hope i can get a way to detect the license plate programmatically.... Any one can help me the detail steps for the ROI to programmtically look for the car license plate number?

[Vision_Armenia · Accepted solution]

OK, I will help you.

I've attached Vision Assistant script you need, OCR file and some sample images. Open your Vision Assistant, load images, then open script file and go through the steps I did. This will be good start for you.

If I answered to your question, don't forget to mark this as a solution 🙂

Wish you good luck.

[Min1992]

hi.. sorry for the late reply.. I have tried ur program... It's really work.. Really thanks for ur help..:)

[Min1992]

Hi Vision Armenia,

I still have one more porblem. I am working on the images of the cars so i tried to extract only the license plate part like your example image. but i still cannot get the right solution for it. I cannot extract only the license plate part of the car. If u know how to extract the part with vision assistant, please help me.. Really thanks..

I have attached some images that i am working now. Please check them for me.. Thanks

[dfsfdsfv]

i can't download test.abc file.can you please mail me

[Vision_Armenia]

Just right click on the test.abc and choose Save Link As ...

[dfsfdsfv]

thank you..its realy working ..great work

[dfsfdsfv]

sir

 the image what you have attached is number plate in a black back ground.i have rgb image and converted it into gray scale image.so how can i locate the plate from it?