JP3053723B2 - License plate character recognition method and device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a license plate character recognition method and apparatus for capturing a running vehicle with a television camera from the front and processing a still image obtained therefrom to recognize a license plate character of the vehicle. Things.

[0002]

2. Description of the Related Art Conventionally, this type of license plate character recognition apparatus has already been put to practical use and applied in various fields. In these, an image pickup device having a built-in television camera and a strobe device is installed on a dedicated gate over the road, and the license plate is imaged from the front of the image pickup device.

[0003] Since the mounting angle of the license plate varies depending on the vehicle, the length of the captured image changes in the lengthwise direction of the license plate. Therefore, in the character recognition process, the size normalization process is performed in consideration of the change in size, but since the character shape itself does not change, it is not necessary to perform the correction process on the character shape. Was.

FIG. 4 shows a process for recognizing upper characters in a conventional license plate character recognition device. In FIG. 4, reference numeral 41 denotes license plate image data;
Is a vertical addition process for detecting the character position from the upper character area, 43 is a character normalization process for normalizing the size of the detected character data, and 44 is a character from the normalized data using the erect template dictionary pattern. This is a pattern matching process to be recognized.

FIG. 5 shows a process of recognizing a use code character in a conventional license plate character recognition device. In FIG. 5, reference numeral 51 denotes license plate image data; 52, vertical addition processing for detecting a character position from a use code character area; 53, character normalization processing for normalizing the size of the detected character data; This is a pattern matching process for recognizing characters from the converted data using the erect template dictionary pattern.

[0006]

However, in such a conventional license plate character recognition device,
In order to install the imaging device, installation work of a gate structure is required, and a large cost is required for equipment installation. Therefore, a system that can be downsized and mounted on a vehicle to improve mobility can be considered, but in this case, it is difficult to install the imaging device on a road lane. In order to address these problems, it is conceivable to install the image pickup device on the road shoulder side or median strip side and take an image obliquely from the front. This causes a problem of causing distortion. When recognizing distorted license plate characters, the characters cannot be detected and recognized by conventional methods. That is, in the conventional method, the character string data is added in the vertical direction of the image, and the space between the characters is detected from the added data value. Therefore, it can be used when a character can be partitioned by a circumscribed rectangle, but a character distorted by oblique imaging is partitioned by a circumscribed parallelogram.
Characters cannot be detected by adding in the vertical direction. Also, the shape of the character is distorted, and the character recognition capability of the distorted shape is required. Since this distortion angle depends on the license plate mounting angle, it varies from vehicle to vehicle.

SUMMARY OF THE INVENTION The present invention solves such a conventional problem, and a method of correctly recognizing a character of a license plate which is geometrically distorted due to a mounting angle when an image of a vehicle is taken obliquely from the front. It is intended to provide a device.

[0008]

In order to achieve the above object, the present invention provides a license plate character recognition device which includes
Geometrically distorted number play by imaging from the front
The character position of the character is corrected so that the space between characters can be detected.
Character position detection section and the position corrected characters
Template dictionary pattern and distortion template dictionary pattern
A character recognition unit that recognizes using the
Wants the imaging device to image the vehicle from the left or right side
From the distortion template registered in advance
Select a dictionary and select the selected distortion template dictionary
Number using turn and erect template dictionary pattern
This is to recognize the characters on the plate .

[0009]

Therefore, according to the present invention, even when an image of a traveling vehicle is taken obliquely from the front, the characters on the license plate can be correctly recognized without being affected by the mounting angle of the license plate.

[0010]

【Example】

Embodiment 1 Hereinafter, an embodiment of the present invention will be described with reference to the drawings. FIG. 1 is a block diagram showing a configuration of an image processing apparatus according to a first embodiment of the present invention, and FIG. 2 is a schematic diagram showing a character position detection / recognition algorithm in the apparatus.

In FIG. 1, reference numeral 11 denotes an image pickup device including a television camera and a strobe lighting device, 12 denotes a vehicle detection sensor for detecting a running vehicle, 13 denotes a device main body for recognizing characters on a license plate, and 14 to 19. It consists of modules. 14 is an overall control unit that controls the entire recognition processing operation, 15 is an AD conversion unit that converts the video signal of the imaging device 11 into digital data, 16 is an image memory that stores the image data digitally converted by the AD conversion unit 15, Reference numeral 17 denotes a plate detection unit that processes the image memory 16 to detect the position of the license plate. Reference numeral 18 denotes a character position detection unit that detects the position of a character on the plate from the license plate position information of the plate detection unit 17. Reference numeral 19 denotes a character position detection. A character recognition unit that recognizes a character from the character position information of the unit 18.

In FIG. 2, reference numeral 21 denotes geometrically distorted license plate image data obtained by oblique forward imaging, reference numeral 22 denotes an upper character region extracted from the license plate image data 21, and reference numeral 23 denotes a character position from the upper character region 22. 24 is a character normalization process for normalizing the size of position-detected character data, and 25 is a character using the upright template dictionary pattern and the distortion template dictionary pattern from the normalized data. Is a pattern matching process for recognizing.

Next, the operation of the above embodiment will be described.
The vehicle detection sensor 12 detects the traveling vehicle and the imaging device 11
When the vehicle captures an image of the vehicle diagonally from the front, the A / D converter 15 performs A / D conversion of the captured data and stores the data in the image memory 16, and then the plate detector 17 detects the position of the license plate. The detected license plate is imaged like the geometrically-distorted license plate image data 21.
The degree of the distortion varies depending on the oblique imaging angle and the mounting angle of the license plate.

Next, the character position detecting section 18 recognizes the upper character "Yokohama 70" from the image data. The character position detecting unit 18 first extracts the upper character region 22 from the geometrically-distorted license plate image data 21. This is different from the upper character data extracted from the license plate image data 41 obtained by the conventional frontal imaging shown in FIG.
The character position cannot be detected by detecting the space between the characters by the vertical addition processing 42 as it is. Therefore, first, distortion correction processing is performed, and then vertical direction addition processing 23 is executed.
Since a character is defined by a circumscribed parallelogram, the conversion process is performed so that the character becomes a rectangle. Since the distortion direction is determined by information indicating which side of the traffic lane the imaging device 11 has taken from the left or right, it is necessary to input that information to the device body 13. The illustrated image data 21 is an image taken from the right side of the driver, that is, from the median strip. From the information that the image is taken from the right side, the character position detection unit 18 shifts the upper character area 22 to the right by one pixel every N scanning lines. In the case of imaging from the left side, the image is shifted to the left. As a result of performing this for all the upper character areas 22, the area becomes a parallelogram. A vertical addition process is performed from this data to detect a character position. When the geometric distortion correction is successfully performed, it is possible to detect a space between characters as shown in FIG. If the space between characters cannot be detected, the correction angle is changed and the process is performed again. That is, the value of the number N of scanning lines at the time of the shift processing is changed. This is repeated until a space between characters is detected.

If the character position is detected by the distortion correction and vertical direction addition processing 23, the character is then recognized by the character recognition unit 19. First, the data is converted into pattern data of a designated size by a character normalization process 24, and a pattern matching process 25 is executed to recognize a character. Here, the data for which the character position can be detected is not always completely corrected for distortion. In order to absorb a slight distortion of the character data to be recognized, two slight distortion patterns other than the erect pattern are registered in the template dictionary pattern of the pattern matching process. In the processing, in order to speed up the pattern matching processing time and prevent erroneous recognition due to similar categories, one of the two types of distortion patterns is used.
Select only one. The selection method is determined based on the information indicating whether the image was taken from the left or right side of the lane.

As described above, according to the first embodiment,
Even if the image of the running vehicle is taken obliquely from the front, there is an advantage that the characters on the license plate can be correctly recognized without being affected by the geometric distortion caused thereby.

(Embodiment 2) Next, a second embodiment of the present invention will be described. In the present embodiment, the configuration of the image processing apparatus is the same as that of the first embodiment shown in FIG. 1, and the description of the configuration will be omitted. FIG. 3 is a schematic diagram showing a character position detection / recognition algorithm according to the second embodiment of the present invention. In FIG. 3, reference numeral 31 denotes geometrically distorted license plate image data obtained by oblique forward imaging,
Reference numeral 32 denotes a use code character area and an upper character area taken out from the license plate image data 31, reference numeral 33 denotes distortion correction and vertical addition processing for detecting a character position from the use code character area and the upper character area 32, and reference numeral 24 denotes a position detected. Character normalization processing for normalizing the size of character data, and reference numeral 35 denotes pattern matching processing for recognizing characters from the normalized data.

Next, the operation of the above embodiment will be described.
As described above, when the vehicle is imaged from obliquely forward, the license plate is imaged like the geometrically-distorted license plate image data 31. The degree of the distortion varies depending on the oblique imaging angle and the mounting angle of the license plate. When recognizing the application code “sa” from the image data 31, first, the geometric distortion license plate image data 3
Then, the application code character area 32a is extracted from “1”. this is,
Unlike the conventional use code character data extracted from the front license plate image data 51 shown in FIG. 5 in the related art, there is a geometric distortion. Therefore, after performing the distortion correction processing, the vertical direction addition processing 33 is executed. At the time of distortion correction, upper character position detection processing is performed, and the detected information is used. Since the upper-stage character in which the geometric distortion has occurred is defined by the circumscribed parallelogram, the conversion process is performed so that this is a rectangle. Since the direction of the distortion is determined based on the information on which side of the lane the image pickup device 11 picks up the image, it is necessary to input that information to the device body 13. The illustrated image data 31 is an image taken from the right side of the driver, that is, from the median strip side. From the information that the image is taken from the right side, the character position detection unit 18 shifts the upper character area 32b right by one pixel for every N scanning lines. In the case of imaging from the left side, the image is shifted to the left. As a result of performing this for the entire upper character area 32b, the area becomes a parallelogram. A vertical addition process is performed from this data to detect a character position. When the geometric distortion correction is successfully performed, it is possible to detect a space between characters as shown in FIG. If the space between characters cannot be detected, the correction angle is changed and the process is performed again. That is, the value of the number N of scanning lines at the time of the shift processing is changed. This is repeated until a space between characters is detected.

If the upper character position is detected by the distortion correction and vertical direction addition processing 33, the same shift processing is performed on the application code character area 32a to detect the application code character. Further, the character data is changed to pattern data of the designated size by the character normalization processing 34, and the pattern matching processing 35 is executed to recognize the character. Here, the data for which the character position has been detected may not always have the distortion completely corrected. In order to absorb a slight distortion of the character data to be recognized, two slight distortion patterns other than the erect pattern are registered in the template dictionary pattern of the pattern matching process. In the processing, in order to speed up the pattern matching processing time and prevent erroneous recognition due to similar categories, one of the two types of distortion patterns is used.
Select only one. The selection method is determined based on the information indicating whether the image was taken from the left or right side of the lane.

As described above, according to the second embodiment,
Even if the traveling vehicle is imaged obliquely from the front, there is an advantage that the use code of the license plate can be correctly recognized without being affected by the geometric distortion caused thereby.

[0021]

According to the present invention, as is apparent from the above-described embodiment, the characters on the license plate imaged geometrically distorted by the image from obliquely forward of the vehicle can be correctly recognized. In this case, it is possible to perform the recognition process by installing the imaging device on the roadside or on the median strip without installing a gate structure for overhead imaging on the road.

[Brief description of the drawings]

FIG. 1 is a block diagram illustrating a configuration of an image processing apparatus according to each embodiment of the present invention.

FIG. 2 is a schematic diagram illustrating an upper-character position detection / recognition algorithm according to the first embodiment of the present invention;

FIG. 3 is a schematic diagram showing a use code character position detection / recognition algorithm in a second embodiment of the present invention.

FIG. 4 is a schematic diagram showing a conventional upper character position detection / recognition algorithm.

FIG. 5 is a schematic diagram showing a conventional application code character position detection / recognition algorithm.

[Explanation of symbols]

 REFERENCE SIGNS LIST 11 imaging device 12 vehicle detection sensor 13 device main body 14 overall control unit 15 AD conversion unit 16 image memory 17 plate detection unit 18 character position detection unit 19 character recognition unit 21 geometric distortion license plate image data 22 upper character area 23 distortion correction And vertical addition processing 24 Character normalization processing 25 Pattern matching processing

Claims (2)

(57) [Claims] A vehicle detection sensor for detecting a traveling vehicle;
An imaging device for imaging a traveling vehicle obliquely forward in the vehicle traveling direction
And the number play of the vehicle from the video signal of the imaging device.
Device that detects the character and recognizes the character.
The main body of the device is geometrically distorted by oblique imaging.
The character position on the license plate can be detected with the space between the characters.
Character position detection unit that detects and corrects
Characters into the erect template dictionary pattern and distortion template
Character recognition unit that recognizes using the plate dictionary pattern
The character recognition unit has an image pickup device on the left side of the vehicle or
From the information that the image was taken from the right side, before registering in advance
Select the skewed template dictionary and select this selected skew
Template dictionary pattern and erect template dictionary pattern
Recognition of license plate characters using
License plate character recognition device. 2. A number play in which a character position detector is distorted.
Detects the application code of the default and detects each character in the upper character string
Using the distortion correction angle information obtained by
Code distortion, and use the corrected usage code
The number plate character recognition device according to claim 1, wherein the recognition unit performs recognition .