Localization of License Plate Number using Dynamic Image Processing and Genetic Algorithm

Localization of License Plate Number system helps to monitor traffic intelligently during rush hours. Moreover, it finds application in commercial parking systems to access the parking space. In automatic number plate recognition method the primary step is to localize the license plate in the captured image. In order to identify license plate number, the number plate location has to be first pointed out in the image. The aim of localization is to preserve only the number plate area from the input image. Later on from this number plate area, the characters are then segmented and recognized. By making use of dynamic image processing techniques and Genetic Algorithm, license plate number in an image can be recognized.

Overview:

Captured image is converted to gray scale for further processing.  Each pixel in a 24 bit color image contains the Red, Green and Blue color components, with each component consuming 8 bits of information. From these three components, 8-bit gray scale value for each pixel is calculated. Adaptive Binarization is performed on the gray image for highlighting the hard borders and number plate symbols. The key of image binarization is its threshold selection. An appropriate threshold can greatly suppress the noise.

In Mathematical morphology image filtering and geometric analysis is done using structuring elements (SE). Mathematical morphological operations are opening, closing dilation and erosion etc.   Dilation followed by erosion fills the noisy holes in the number plate area and connects the broken symbols in the binary image.Dilation - grows image regions. Dilation is an operation that "thickens" or "grows" or the objects in a binary image. Erosion is used to remove irrelevant details from the binary image. Erosion means it “shrinks” objects in a binary image.

Connected Component Analysis (CCA) is a technique in image processing which scans an image and labels the connected pixels in an image. The connected components are then examined to filter out extended and large components. The extracted objects from the CCA stage are filtered on the basis of their size.  Size filtering is applied because we have a good idea of how large the license plate symbols will be.

Finally a genetic algorithm processes the binary image to detect the license on the license plate. The formulation of the GA phase includes:

1. Random generation of initial population.
2. Fitness evaluation of each chromosome
3. Repeat :​

• Selection
• Cross over
•  mutation  

  4.    Fitness evaluation of chromosome until the stop criterion is satisfied.                                                                                   

Genetic algorithms are used for optimizing functions. It is based on the evolution of a population of solutions which under the action of some precise rules that is to be optimized. GA manipulates a population of fixed size. Population is formed by chromosomes.

Genetic operators: Selection, crossover and mutation are used to create a new population during each iteration. During selection, algorithm chooses the most relevant candidates. Crossover builds 2 new chromosomes from 2 old chromosomes (parents). Mutation realizes the inversion of one or several genes in a chromosome.

Effective combination of genetic algorithm and image processing techniques helps to localize the number plate in the captured image. Localization of the number plate from the captured image is shown here.

Captured image                                  Number plate detected