Application of character structure knowledge in license plate recognition

xiaoxiao2021-03-17  216

Application of character structure knowledge in license plate recognition

Source: Application of Electronic Technology: China University of Science and Technology Song Jian was

Abstract: A character recognition method based entirely of structural knowledge is proposed. The method is determined using the determination tree to classify the letters and numbers in the automotive license when the character is characterized and the stroke type, data and position as the identification of the table. Key words: license plate number character structure determine tree character identification car license identification is based on image segmentation and image identification theory, analyzing images containing automobile licenses, thereby determining the location of the car license in the image, and further extracting and identifying text character. The character images divided from different license plate images have a large size, and the difficulty of identification is increased. Although the image conversion method can be used to normalize all character images to the same size, the normalization process inevitably lost useful character information, resulting in image distortion, helplessness to improve the identification accuracy, and waste time Reduce the recognition speed. In this paper, by analyzing the characteristics of the car license, a method of identifying a stroke feature and structural knowledge of characters is proposed to identify the letters and numbers in the automobile license. Tests show that the method recognizes the speed, high accuracy, and is not affected by the size of the character image. 1 license plate character structure and identify the characters used in mainland Chinese auto licenses include 59 Chinese characters, 25 English letters (alphabetics are not available) and 10 Arabic numbers in three types, and are printing, structural fixation, strokes specification. Figure 1 is an image of all letters and numbers used in license plate numbers. The structure of these characters has three types in the horizontal direction: the left and right symmetry, left or right, left and right. There are three structures in the vertical direction, namely the upper and lower symmetry, the upper large lower, and the upper is large. If the number "8" used in the license plate number is, it belongs to the left and right symmetry and is symmetrical up and down. From Figure 1, you can see that all letters and numbers have two major categories: straight strokes and arc drawings. Straight strokes can be divided into cross-strokes, vertical strokes, left slope strokes (equivalent to "" in Chinese character strokes) and right inclined strokes (equivalent to "捺" in Chinese characters). The arc pen painting is a curve section, which is divided into two categories: an arc stroke and a backhand stroke. The so-called an arc stroke refers to the arc pen painting does not form a closed loop, such as the letter "C". The closure strokes form a closed loop, such as the number "0". According to this feature of the character image, this paper uses the following method to classify the letters and numbers, forming a recognition determination tree, each character is a leaf: (1) Search in the character image to be identified. Quantity and location. (2) According to the result of the result of the result of the enclosed ring, the classes, double closed loop character classes, and no closed loop character classes. (3) Treatment for each class. (4) Double closed loop characters are only "8" and "B", so they can distinguish these two characters as long as the vertical pen is drawn. The left half of "B" has a long vertical, and "8" is not. (5) A single-sealing closure character has "A", "D", "O", "P", "Q", "R", "0", "4", "6", and "9". These characters are divided into three categories according to the location of the closed loop: the closed loop is in the upper part; the closed loop is in the middle, and then identifies according to the structural characteristics and the extracted stroke feature. . The closure ring is "P", "R" and "9" in the upper character. If there is a closed loop in the upper half of the character image, the vertical pen is drawn from the left half; if there is no vertical stroke of the left side, the character is "9"; if the left half is extracted to the vertical pen draw, continue to extract Right oblique strokes; extract to the right oblique stroke, the character is "R"; otherwise "p".

. The closure ring is "A", "4" and "6" in the lower character. If there is a closed ring in the lower half of the character image to be identified, the vertical pen is drawn from the right half; if there is a vertical stroke of the right side, the character is "4"; if the right half is not extracted to the vertical pen, continue Plug-in; extract the crossword, the character is "a"; otherwise it is "6". . The enclosed loop in the middle character has "D", "O", "Q" and "0". In practical applications, "O" and "0" images are identical, and can be treated as the same character. If there is a closed ring in the middle of the character image to be identified, first determined whether or not "Q" is determined by the upper and lower symmetry; if the upper and lower symmetry is "0" ("O") or "d"; then follow the left and right symmetrical characteristic distinguished characters "0" and "d". (6) The characters with no closed loop are "C", "E", "F", "C", "H", "J", "K", "L", "M", "N", " S "," t "," u "," v "," w "," x "," y "," z "," l "," 2 "," 3 "," 5 "and" 7 " These characters are identified by extracting strokes. The specific steps are as follows: • Takes horizontal pens and vertical pen. • If the character image is to be identified, there is no cross-sharing and vertical pen drawing, then the character is "s", "v" or "x". • If the character image is to be identified, only the stroke of the stroke is not vertical stroke, the character is "2", "3", "7" or "Z". · If the character image is to be identified, only a vertical stroke, there is no horizontal painting, then the character is "1", "c", "j", "k", "m", "n", "u", "w '" Or "y". • The character image to be identified is both a stroke and a vertical stroke of the strokes to "5", "e", "f", "c", "h", "l" or "t". "S", "v" and "x" identification. Extract left slope strokes and right inclined strokes, "S" no two strokes, thus identifying "s". "X" two oblique points are located The intermediate position of the character image, and the two oblique points of "V" intersect the lower portion of the character image to identify "X" and "V". · Identification 2 "," 3 "," 7 ", and" Z ". There are only two cross-strokes in these four characters, so that "Z" can be identified. "3" and "7" cross-strokes are located at the upper, and the "2" cross painting is located in the lower part, so that "2" can be identified. For "3" and "7", identify using left slope strokes.

"7" has left slope strokes, and "3" is not. · Identify "1", "C", "J", "K", "M", "N", "U", "W" and "Y". According to the number of vertical pen drawings into three types, "1", "c", "j", "k" and "y" are a vertical stroke, "M", "n", and "u" are Two vertical pen paintings, and "W" has three vertical pen paintings, which complete the identification of "W". For a vertical stroke, it is determined that the position of the stroke is on the left ("c" and "k"), the middle ("1" and "y") or in the right ("J"), to identify "plant" According to the "C" and "K" are distinguished according to the "C" and "K", "1" and "Y" are distinguished according to the length of the middle vertical stroke. And "U" identifying. For "M" and "U", depending on the structural feature, it is unrecognized. This article uses the ratio of the foreground pixel number and the number of background pixels and the ratio of the number of background pixels. According to these two characters Only the upper part of the character can be calculated. · Identify "5", "e", "f", "g", "h", "i", "t." These characters, only "e" has three Both strokes, "F" has two crosswords, the rest is a horizontal stroke. In the remaining characters, divided into two groups according to the number of vertical pen: "5", "l" and "t" is a vertical stroke, "G" and "h" are painted for two vertical pen. "H" has the same length, and the two vertical pen paintings of "G" are short, which is distinguished "g" and "h" Sign. The vertical pen painted in the middle, "5" and "L" on the left. "L" vertical pen painting long, "5" vertical strikes, this is completed "5", "t "And" L "identification. Search closed loop is actually searching for connectivity fields in character images. In the binary image of characters, it is assumed that the character pixel value" 1 ", the background pixel value is" 0 ", then: (1" ) There are only two connected domains in the character image of the closed loop, namely the character communication domain, and the background communication field, B and F. (2) in Figure 2 (2) There are only three communication domains in the character image of a closed loop. That is, a character communication domain and two background communication domains, B1, B2, and F. (3) in Figure 2 (c) have four connected fields in the character images of two closed loops, namely one character communication domain and three Background communication domain, B1SHE, B2, B3, and F. The algorithm of search closed loops is as follows: (1) Read the binary character image. (2) Find a background of a pixel value "0" Pixel point B. (3) Search B's communication domain, and mark all pixels in the communication domain as a background 1. (4) Pixels that are "0" in the image. (5) If all "0" pixels It has been marked as a background 1, then the number of closed loops in the image is 0, jumps to (11).

转载请注明原文地址:https://www.9cbs.com/read-129584.html

New Post(0)