Number plate extraction

[mini09]

Hi friends,

            I`m working on a number plate extraction. Initially I just designed for an single image with a fixed ROI. Later with the idea of my friend used pattern matching in the Vision Assistant to detect the number plate and used Number of Search Array and Unbundle by name function to get the points of the rectangle and given it to an Rectangle to ROI and trained OCR. But this is not working exactly as it should be. Please share your valuable experience here. 

[MoviJOHN]

If the camera position is fixed, and you have a triger sensor, you can limit your search to looking in a smaller region.  Use morphology functions to find the plate, and fill holes, so that test is not seen.  Then calculate the approximate area f the rectangle, and compare to the actual area of the blob.  If theyare close, then you probably have found the plate.  Extract the plate using the blob as a mask, and then proceed with OCR.