Solving the license plate detection problem as a genetic algorithm optimization problem

What is it about?

In this paper, the license plate number detection problem is formulated as a genetic algorithm optimization problem. Each bounding box of a license symbol is considered as a gene and the whole license number is represented as a chromosome. A prototype of a certain license plate is represented using a geometric relationship matrix (GRM) in which relative positions and dimensions of the neighboring symbols are recorded in its cells. To localize a license plate in an image, the image is binarized using a simple adaptive binarization technique to overcome illumination problems. Then all the foreground objects in the binarized image are determined using the connected component analysis technique. the job of the genetic algorithm now is to search for the best sequence of CCA objects that minimizes the objective distance measured based on the specific geometric relationship matrix that represents the concerned license plate. To speed up the genetic algorithm, two crossover operators are designed to exchange genes between best fit chromosomes based on their spatial distribution in the vertical and horizontal directions. To overcome CCA problems, partial matching is allowed during the calculation of the objective distance between the current sequence and the GRM of the concerned LP. This flexibility allows detection of license numbers which have some connected symbols or missed ones. The system is implemented using MATLAB and various image samples are experimented with to verify the distinction of the proposed system. Encouraging results with 98.4% overall accuracy are reported for two different datasets having variability in orientation, scaling, plate location, illumination, and complex background. Examples of distorted plate images are successfully detected due to the independency on the shape, color, or location of the plate.

Featured Image

Photo by Semyon Borisov on Unsplash

Photo by Semyon Borisov on Unsplash

Why is it important?

The introduced solution is important because it solves the first step in the vehicle recognition problem which is the basic component in all intelligent transportation systems. In addition to solving the global object detection problem when the concerned object is composed of many subobjects.

Perspectives