چکیده:
Automatic detection of vehicle license plate locations, as the first and most important step in most intelligent traffic control systems such as license plate readers, speed cameras, image processing, etc., plays an important and fundamental role in the mechanization of violation detection. The more the error in this section is reduced, the more optimized the overall results of the aforementioned systems will be. In this regard, in this article, we have attempted to compare two completely different methods using a large number of diverse images. The first method is based on finding the plate according to the dimensions of the plate, while the second method is based on finding the plate based on the set of adjacent characters. In the first method, using the Sobel vertical operator and morphological operators, an attempt is made to find the license plate frame in the input images, and in the second method, using the Laplace filter and combining it with morphological methods, an attempt is made to find the characters and finally the license plate frame. In both methods, pixel density sensing is used to find the license plate frame.
خلاصه ماشینی:
There are various solutions for license plate extraction, including edge detection (Ballard 3, 1991) and (Kanayama, Fujikawa, Fujimoto 4, 1991, 798), Hough Transform (Kim, Lee, Lyou, Park, 5, 1997, 511), histogram analysis (Cho 6, 1998, 94), and the use of morphological operators 7 (Shridhar, Miller 8, 1999, 17).
/ Figure 2 - Input sample Grayscale image / Figure 3 – Result of Sobel edge detection operations / Figure 4 – Result of image opening operations and noise removal 1 - Image open 1-3- Unifying close edges: In this stage, in order to detect the license plate location using the morphological image closing 1 method, all edges closer than 5 by 21 pixels are joined together, and with these operations, all vertical edges of the image are converted into unified rectangles, which can be seen in Figure 5.
1-4- Noisy rectangles: Given that the image obtained from the previous stage contains very small rectangles and practically these rectangles have no other use except creating additional processing in subsequent stages, therefore, in this stage, we remove rectangles whose length is less than fourteen pixels using the image opening method, in which we see the result of applying this operator in Figure 6.
/ Figure 11– Image resulting from converting Grayscale to binary 2-3- Noise removal using morphological methods: Despite the results obtained from the previous stage, the relevant image still contains noises that cause processing slowness and also have an undesirable effect on license plate recognition results.
"A new and fast method for detecting the location of vehicle license plates from complex images based on morphological operations".
(
, 
, 
)
