By a car license plate recognition, we mean a software system processing images and providing an alphanumeric transcription of car plates included in an image. We divide the task into four sub-tasks: license plate localization, license plate extraction, characters segmentation and characters recognition. All four sub-tasks are discussed in the context of standard approaches and own solution based on a chain of standard and soft computing image processing algorithms is presented. In this chain, the F-transform approximate pattern matching algorithm plays the crucial role.
The general car plate localization and reading in ideal conditions are well-solved applications. On the other side, we have dealt with a car plate reading in difficult conditions, where - according to our experience - the existing systems might fail. By difficult conditions, we mean night images, images with bad weather conditions - ice, snow, rain, images with debris, images containing multiple texts, car plates in non-defined format, or over-exposured images.
We have split the task into four sub-tasks: car plate localization, car plate extraction, character segmentation and character recognition. For all the sub-tasks we proposed a solution based on a connection of fast standard and soft computing image processing methods. The standard methods, such as connect component analysis and percentile thresholding, allow us to realize simple and fast processing, while the soft-computing ones bring robustness.
© 2019 | We are part of the Institute For Research And Applications Of Fuzzy Modeling