Image Processing Reference

In-Depth Information

6.3 Plate Numbers Recognition

This is done by first determining the plate position, then taking the numbers from the plate

and from a look-up table to compare. The numbers are detected by taking the ones that have

the maximum correlations. The look-up table is a template that contains the numbers from 0

to 9.

The action of recognition includes many functions to obtain the exact result. The image will

be transformed to grayscale. Then the extended-maxima transform is computed. Edge detec-

tion and Strel functions are also used. Finally, plate shaping and filter processing are used to

obtain the final result.

Plate position determination is made by the following steps:

- Transform the image to grayscale.

- Compute the extended-maxima transform.

- Apply edge-detection-type Sobel.

- Determine the line by Strel function.

- Dilate the image.

- Verify the rectangular plate shape.

Plate number recognition (PNR) consists of three parts:

- PNR: in this M-file, the image is being read and then opens a text document where the res-

ult is being displayed at. In this M-file, the plate numbers are being divided each separately

in order to the compare. As shown in
Figure 10
.

FIGURE 10
Templates for the plate numbers.

- Creat: in this M-ile, the templates have been made.

- Recognize_numb: in this M-file, it computes the correlation between template and input

image and takes the maximum correlation. Its output is a string containing the leter.

The correlation is made by this part of the code:

(4)

where all
numb
is from the PNR M-file which is the size templates

(5)

This comutes the correlation between the templates and the input image

Search WWH ::

Custom Search