Sei sulla pagina 1di 3

A summary of Localization of License Plate Number Using Dynamic Image Processing Techniques And Genetic Algorithms"

Eric Laska Department of Computer Science University of Minnesota Morris 1. INTRODUCTION


higher-tness members and selected, mutated, and crossedover to produces subsequent generations. These ospring are then expected to do better than their parents when fed through the tness evaluation, and the selection/crossover/mutation process is repeated for many iterations.

In their paper Localization of License Plate Number Using Dynamic Image Processing Techniques And Genetic Algorithms, the authors G. Abo Samra and F. Khalefah provide a new evolutionary computational algorithm (genetic algorithm) to take raw (color or grayscale) images taken of license plates and outputs the locations of license plate symbols. This is accomplished using a genetic algorithm the authors developed. From here a simple optical character recognition algorithm has an easy task of converting these license plate symbols to letters, symbols, and numbers. Presumably this algorithm would be used for law enforcement purposes.

3.

OUTLINE

2.

BACKGROUND

A genetic algorithm is an evolutionary computational algorithm which uses cellular genetics as a model. Given that evolution has been successful enough to create intelligent species (us) using genetics, as well as practically all life as we know it, genetics is a natural model to base evolutionary computation on. Chromosomes are used to encode information into a genetic algorithm in a manner akin to using the letters ACGT (seen the movie GATACA?) which we use to represent the genetic code of DNA. When we realize that this genetic code is just like a base-4 number, we can easily extend genetic encoding to binary, base-2, numbers. Evolution then takes place using mutation and crossover operators which alter genetic letters or switch regions of genetic code between two individuals, respectively. We encode a goal of what we wish to evolve as logic which we may use to evaluate the tness of genetic code, called a tness function. Once evolution has generated genes which meet or exceed some predened tness requirement, evolution may be halted and a good solution is returned (although evolution may still continue, tness will level o eventually and is expected to increase slower and slower). A population of individuals is created initially, from which

To process a color image of a license plate into a list of license plate symbols requires many steps. The authors begin by converting the image to grayscale, then converting it a black/white-only (binary) image through a process known as adaptive binarization which highlights the license plate symbols and hard boundaries. Next, a morphological operation is applied to deal with noise in the picture. A connected component analysis determines which objects in the binary image are the license plate characters. And nally size ltering is used to conclude which components are indeed the license plate characters when we feed all this data into a genetic algorithm. This procedure is seen in Figure 1.

This work is licensed under the Creative Commons AttributionNoncommercial-Share Alike 3.0 United States License. To view a copy of this license, visit http://creativecommons.org/licenses/by-nc-sa/3.0/us/ or send a letter to Creative Commons, 171 Second Street, Suite 300, San Francisco, California, 94105, USA.

Figure 1: Flowchart of localizing license plate symbols from an image

4.

IMAGE PROCESSING OF PLATE


We begin with a color image of a license plate.

Figure 5: Morphological closing using 3-pixel diskelements to reduce noise A connected component analysis then extracts objects from the processed binary image. Figure 2: Raw color image of license plate Next it is converted to grayscale, followed by an application of adaptive binarization, which uses neighborhood gray levels to classify pixels as foreground or background. This handles local variations in brightness very well.

Figure 3: Grayscale image of license plate

Figure 6: Connected component analysis of processed binary image (N=2287) Figure 4: Binary image of license plate Next, a morphological operation is applied to remove noise from the binary license plate image. Size ltering is applied because we have a good idea of how large license plate symbols are.

5.

RESULTS

This algorithm works quite well, with a 98.4% accuracy. On a 2.6Ghz PC with 2GB of RAM a 640x480 image takes 0.12 seconds to locate the license plate characters, whereas on a 2048x1536 image it takes 0.34 seconds. Their algorithm has a sublinear algorithmic complexity, which is awesome. The authors then show around a hundred instances of applying their algorithm. Even with highly-variable lighting and positional conditions, as well as highly-variable license plate designs, their algorithm works very consistently, and in a reasonable amount of computational time.

Figure 7: Binary image after applying size ltering (M=64) Finally a genetic algorithm developed by the authors processes the processed binary image to detect the license on the license plate. This algorithm solves the 2D compound object detection problem. Upper left corner coordinates, width, and height of bounding boxes are encoded into chromosomes. Relative positions and sizes between bounding boxes, dealing with variations in skew, perspective, orientation, and scaling of the image. Stochastic universal sampling is used in selection of ospring. Individuals are placed along a line corresponding to their tness evaluation and an equally-spaced distribution are selected along the line. 90% of individuals are exposed to evolutionary operators. The crossover operator used was created by the authors, with weights based on co-alignment of chromosomes. This rapidly increases convergence time of evolution (nding a solution) by logically keeping relevant pieces of the license plate image grouped together. Objective distance between license plate characters is converged to and is used in determining the correct bounding boxes of license plate characters. The end result is a license plate with license plate symbol locations determined.

Figure 8: Binary image after applying the genetic algorithm

Potrebbero piacerti anche