We can see that we have identified extraneous objects, such as the bolts of the and the automaker’s name. Now like previously, we can label the connected components within the license plate. We again gray scale the picture so that we can evaluate which components are connected, so that we can run the connected components algorithm (measure.label from the skimage package). The next step is similar to the previous step in that we are searching for connected components on the license plate. Using this rubric we can filter out the vast majority of objects to get the license plate from the photo. From a picture standpoint, a license plate width wise is very unlikely to take up more than 25% of the width of the entire picture and a license plate height wise is unlikely to take up more than 15% of the entire picture. License plates are usually rectangular in shape that is significantly wider than it is tall.

To obtain the license plate from the objects we need to consider the properties of a license plate. Doing so, will return a matrix with labels for all distinct objects. This will label all the distinct objects within a matrix of the picture’s pixels, using a connected component labeling algorithm. We will do this by grey scaling the original photo, and then try to label distinct objects within the photo using the skimage package’s measure.label. We will try to get all connected components within the picture first, from there we can try to extrapolate the license plate object out of the distinct objects found in the picture.

Classify the character, using Optical Character Recognition (OCR).

Segment and order the characters from the license plate.

Segment the license plate from the photo.

The license plate detection system has three steps And afterwards, I’ll discuss some popular open source license plate detection systems. Hopefully this gives you a rubric so that you can try to create your own implementation of a license plate reader. I will go over my own custom implementation of license plate detection. The goal of the plate detection system is to obtain a plate number from a photo of a car or other motor vehicle. In this blog I will go over how to engineer a license plate detection system.

License plate recognition thai python