3.4.1 Localization
3.4.1 Localization
The binary and noise free image is scanned and using AForge.NET library, rectangular shapes of white intensity with width greater than 80 pixels and height greater than 30 pixels (corresponding to the aspect ratio of Indian LP’s) are detected. Once found they are marked in red as shown in figure 6 below. The Blobcounter class is used that counts and extracts stand alone objects in images using connected components labeling algorithm. The algorithm treats all pixels with values less or equal to Background Threshold as background, but pixels with higher values are treated as objects' pixels. For blobs' searching the class supports 8 bpp indexed grayscale images and 24/32 bpp color images that are at least two pixels wide. …show more content…
The two types of vehicle plates must have two types of mask. If the location of the dash line “-” is varied, a new mask must be produced. One vehicle plate type needs one mask that is impracticable. Therefore, an adaptive method to segment each character is proposed.
The characters on the plate need to be isolated in order to be recognized. Characters are isolated based on the width of the largest character. The image is inverted in color, scanned and checked for the column of black pixels denoting end of the character. Taking into consideration width of the largest / widest character the plate image is then cropped also checking the former stated condition for end of the character. Each character is segmented and stored separately. A counter is set to count the number of characters in the license plate.
Each of the segmented characters is then stored separately to be passed to the next character recognition module for identification. This is illustrated with the results as shown in figure …show more content…
The algorithm operates on all black pixels P1 that can have eight neighbors. The neighbors are, in order, arranged as:
P9 P2 P3
P8 P1 P4
P7 P6 P5
Figure 13 Neighborhood arrangement used by thinning algorithm
There are two iterations the algorithm passes through which consist of successive passess applied to contour points of the given region where a contour point is any pixel with value 1 and having at least one 8 neighbor valued 0. The neighborhood arrangement as shown above in Figure 13 is the arrangement used.
The first step of thinning algorithm flags a point p for deletion if following conditions are satisfied:
a) 2<=6 where N(p1) is the number of non zero neighbors of p1; that is , N(p1)= p2+p3+p4+…..+p8+p9;
The pixel p1 should have more than 2 neighbor pixels and less than 6 neighbor pixels.
This condition is violated when contour point p1 has only one or seven of the 8 neighbors valued 1. Having only one such neighbor implies that p1 is the end point of the skeleton stroke and obviously should not be deleted. Deleting p1 if it has seven such neighbors would cause erosion into the
The binary and noise free image is scanned and using AForge.NET library, rectangular shapes of white intensity with width greater than 80 pixels and height greater than 30 pixels (corresponding to the aspect ratio of Indian LP’s) are detected. Once found they are marked in red as shown in figure 6 below. The Blobcounter class is used that counts and extracts stand alone objects in images using connected components labeling algorithm. The algorithm treats all pixels with values less or equal to Background Threshold as background, but pixels with higher values are treated as objects' pixels. For blobs' searching the class supports 8 bpp indexed grayscale images and 24/32 bpp color images that are at least two pixels wide. …show more content…
The two types of vehicle plates must have two types of mask. If the location of the dash line “-” is varied, a new mask must be produced. One vehicle plate type needs one mask that is impracticable. Therefore, an adaptive method to segment each character is proposed.
The characters on the plate need to be isolated in order to be recognized. Characters are isolated based on the width of the largest character. The image is inverted in color, scanned and checked for the column of black pixels denoting end of the character. Taking into consideration width of the largest / widest character the plate image is then cropped also checking the former stated condition for end of the character. Each character is segmented and stored separately. A counter is set to count the number of characters in the license plate.
Each of the segmented characters is then stored separately to be passed to the next character recognition module for identification. This is illustrated with the results as shown in figure …show more content…
The algorithm operates on all black pixels P1 that can have eight neighbors. The neighbors are, in order, arranged as:
P9 P2 P3
P8 P1 P4
P7 P6 P5
Figure 13 Neighborhood arrangement used by thinning algorithm
There are two iterations the algorithm passes through which consist of successive passess applied to contour points of the given region where a contour point is any pixel with value 1 and having at least one 8 neighbor valued 0. The neighborhood arrangement as shown above in Figure 13 is the arrangement used.
The first step of thinning algorithm flags a point p for deletion if following conditions are satisfied:
a) 2<=6 where N(p1) is the number of non zero neighbors of p1; that is , N(p1)= p2+p3+p4+…..+p8+p9;
The pixel p1 should have more than 2 neighbor pixels and less than 6 neighbor pixels.
This condition is violated when contour point p1 has only one or seven of the 8 neighbors valued 1. Having only one such neighbor implies that p1 is the end point of the skeleton stroke and obviously should not be deleted. Deleting p1 if it has seven such neighbors would cause erosion into the