JPH1125258A - Device and method for recognizing number plate - Google Patents

Abstract

PROBLEM TO BE SOLVED: To accurately perform character recognition of a number plate that is photographed diagonally by accumulating preliminarily information of a typical histogram of an image which scans a number plate at a prescribed angle and selecting an inclination angle in a scan direction so that information of a histogram that is acquired by a recognizing means may come close to the information of the typical histogram. SOLUTION: A 1st line sensor camera 3 photographs a parked vehicle while moving and accumulates a photographed image signal in memory 2. An image analyzing part 5 extracts an image signal of the part of a number plate of the parked vehicle by detecting a vertical edge. A data creating part 6 preliminarily accumulates information a typical histogram of an image that is scanned at a prescribed angle about a number plate. Then it selects an inclination angle of a scan direction so that information of a histogram which is acquired by counting respectively the number of horizontal and vertical pixels of the number plate can come close to the information of the typical histogram.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an apparatus for recording image information of a parked vehicle parked on a road together with position information and time information. The present invention is an apparatus developed for use in controlling illegally parked vehicles, but can be used for applications other than control. The device of the present invention is mounted on a cracking vehicle, and the cracking vehicle simply passes by the side of the subject parking vehicle, thereby obtaining image information including a license plate of the parking vehicle,
The present invention relates to an information processing apparatus capable of recording position information and time information in association with each other.

[0002]

2. Description of the Related Art In police for illegal parking, when a police officer identifies illegal parking, a license plate and other information identifying a vehicle are recorded together with the parking position and time. For this record, vehicles were equipped with cameras to crack down on parking violations.
Pictures are extremely useful as evidence.

[0003] This conventional example will be described with reference to FIGS. 7 to 9. FIG. 7 is a diagram illustrating a detection device for a parked vehicle. FIG. 8 is a diagram showing an example of an illegal parking sticker. FIG. 9 is a block diagram of a main part of a detection device for a parked vehicle. The applicant of the present application has proposed a parked vehicle detection device based on two prior applications (Japanese Patent Application Nos. 8-248127 and 8-248146, not disclosed at the time of filing the present application). As shown in FIG. 7, an image of the parked vehicle 12 is captured by a detection vehicle including the first line sensor camera 3 and the second line sensor camera 4. The license plate 13 is photographed by the first line sensor camera 3. The second line sensor camera 4 is used to capture an image for grasping the positional relationship between the parked vehicle 12 and the detected vehicle. The image signal photographed in this manner is analyzed and used as data for creating an illegal parking sticker as shown in FIG.

[0004] This detection device is configured as shown in FIG. An image signal captured by the first line sensor camera 3 is written into the memory 2 by the image recording unit 1. The image signal written in the memory 2 is read out and input to the image analyzer 5.

FIG. 10 is a flowchart showing the operation of the image analyzer 5. The image analyzer 5 reads an image signal from the memory 2 (S1), and extracts an area (candidate area) on which the license plate 13 is imprinted from the image signal (S2). A boundary between the upper numeral and character group (for example, “Shinagawa 33”) and the lower numeral and character group (for example, “A12-34”) of the license plate 13 is detected (S3). First, a boundary between the lower character group (hereinafter, “character” is defined as a term including both a numeral and a character) is detected (S4), and the extracted character is recognized (S5). Subsequently, the boundary of the upper character group is detected (S6), and the cut out character is recognized (S7).

The boundary detection described above will be described with reference to FIGS. FIG. 11 is a diagram showing an example of a license plate. FIG. 12 is a diagram for explaining detection of a boundary between an upper character group and a lower character group. FIG. 13 is a diagram for explaining detection of a boundary between characters in a lower character group. As shown in FIG. 12, the image of the license plate shown in FIG. 11 is scanned in the horizontal direction to count the number of pixels where characters are displayed. The method of counting the number of pixels is a known technique and will not be described in detail.However, an image actually projected on the screen may be counted by scanning, or the image actually projected on the screen may still be counted. It is also possible to virtually create the situation displayed on the screen at the stage of the image signal that has not been generated, perform scanning, and count.

[0007] By counting in this manner, the count value of the number of pixels is increased in the area where the character exists, as shown in the left histogram of FIG. This histogram has the number of pixels on the horizontal axis and the position on the vertical axis. Therefore, it can be determined that the portion sandwiched between the two regions having a large number of pixels is the boundary of the detected portion. Since the size and the character arrangement of the license plate are determined according to the standard, the boundary can be accurately obtained by comparing the distribution of the number of pixels on the histogram with a predetermined pattern.

Similarly, for the image of the license plate shown in FIG. 11, the number of pixels where characters are displayed is counted by scanning in the vertical direction as shown in FIG.

By counting in this manner, as shown in the lower histogram of FIG. 13, the count value of the number of pixels increases in the area where the character exists. This histogram has a position on the horizontal axis and the number of pixels on the vertical axis. Therefore, it can be determined that the portion sandwiched between the two regions having a large number of pixels is the boundary of the detected portion. Since the size and the character arrangement of the license plate are determined according to the standard, the boundary can be accurately obtained by comparing the distribution of the number of pixels on the histogram with a predetermined pattern.
A character is cut out at each boundary thus obtained and the character is recognized.

As shown in FIG. 12, when a boundary between a character group at the upper part of the license plate and a character group at the lower part of the license plate is detected, when scanning in the vertical direction is performed, the scanning start position is set at this boundary. By performing scanning from to the lower character group, it is possible to prevent the upper character group of the license plate from affecting the boundary detection of the lower character group. Similarly, by performing scanning from this boundary to the upper character group, it is possible to prevent the lower character group of the license plate from affecting the boundary detection of the upper character group.

The recognition of the character group at the top of the license plate ("Shinagawa 33") can be described in the same manner as the recognition of the character group at the bottom.

Since the license plate often bends at the boundary between the upper character group and the lower character group for structural reasons, it is handled individually, so that accurate character recognition can be performed even when the license plate is bent. It can be performed.

[0013]

In such a detection system for a parked vehicle, the angle at which the first line sensor camera 3 photographs the license plate 13 is not obliquely from the front but obliquely. Further, although it cannot be read from FIG. 7, if the height of the mounting position of the first line sensor camera 3 from the road surface is higher (or lower) than the height of the license plate 13 from the road surface, the vertical direction Is also oblique.

FIGS. 14 and 15 show histograms of license plates photographed obliquely. As shown in FIG. 14, a license plate that is photographed diagonally is scanned in the horizontal direction and the number of pixels in which characters are displayed is counted. As shown in the histogram on the left side of FIG. 14, the number of pixels corresponding to the presence or absence of characters is counted. Are not formed, and the boundary between them is unknown. Similarly, FIG.
As shown in FIG. 15, when the license plate that is photographed obliquely is scanned in the vertical direction and the number of pixels in which characters are displayed is counted, the number of pixels corresponding to the presence or absence of characters is large as shown in the histogram at the bottom of FIG. Six regions are not formed, and their boundaries are unknown. In such a case, there is a problem that it is difficult to perform character recognition normally.

It is an object of the present invention to provide a license plate recognizing device which has been made in such a background and can perform accurate character recognition even on a license plate photographed obliquely. . SUMMARY OF THE INVENTION It is an object of the present invention to provide a license plate recognizing device capable of setting photographing conditions gently.

[0016]

SUMMARY OF THE INVENTION A first aspect of the present invention is a license plate recognizing apparatus, which is a line sensor camera for taking an image of a parked vehicle while moving beside the vehicle, and an image captured by the camera. A memory for storing the image signal, and a means for reading the image signal from the memory and analyzing the image signal.
Means for extracting an image signal of the license plate portion of the parked vehicle by detecting a vertical edge, and a numeral written on the license plate from the image signal of the license plate portion extracted by the extraction means; Means for recognizing a character, wherein the recognizing means scans an image based on the image signal of the license plate in a horizontal direction and a vertical direction of the license plate, respectively, and determines the number of pixels on which the number and the character are displayed. A license plate recognizing apparatus includes means for counting information in the horizontal direction and vertical direction to obtain histogram information, and means for recognizing the arrangement of numbers and characters on the license plate in accordance with the information of the histogram.

Here, a feature of the present invention is that typical histogram information of an image obtained by scanning a license plate at a predetermined angle is stored in advance, and the information of the histogram obtained by the recognizing means is stored. There is a means for selecting an inclination angle in the scanning direction so as to approach information of the typical histogram.

As described above, by performing scanning from a plurality of angles within a predetermined range, not only in the horizontal direction and the vertical direction, the inclination angle of the license plate photographed obliquely at any one of the angles coincides with the inclination angle. There is an angle,
The character arrangement state can be accurately grasped at the coincident angle.

That is, since the size and character arrangement of the license plate are standardized according to the standard, the pattern of the histogram is checked in advance, and the angle in the scanning direction and the number are compared by comparing with the information of the histogram obtained on site. It can be determined whether or not the inclination angle of the image of the plate matches.

It is preferable that a means for correcting a reading angle of numbers and characters on the license plate according to the inclination angle is provided.

Thus, even if the license plate is photographed obliquely, characters engraved on the license plate can be accurately recognized.

According to the recognition result of the recognizing means, there is provided means for detecting a boundary between the upper character group (Shinagawa 33) and the lower character group (A12-34) on the license plate, and the selecting means comprises: It is desirable to include means for individually selecting an inclination angle in the scanning direction for each of the two character groups.

That is, when a boundary between a character group at the upper part of the license plate and a character group at the lower part of the license plate is detected, when performing vertical scanning, the scanning start position is set at this boundary.
By performing scanning from this boundary to the lower character group, it is possible to prevent the upper character group of the license plate from affecting the detection of the lower character group boundary. Similarly, by performing scanning from this boundary to the upper character group, it is possible to prevent the lower character group of the license plate from affecting the boundary detection of the upper character group.

Further, the license plate is easily bent at the boundary between the upper character group and the lower character group due to its structural reason. Therefore, by treating these character groups individually, accurate character recognition can be performed even on a bent license plate.

A second aspect of the present invention is a license plate recognition method, in which a vehicle is parked by a line sensor camera while moving beside a parked vehicle, and an image of a license plate portion is extracted from the captured image. This is a method of recognizing numbers and characters on the license plate using an information processing device that performs character recognition. The feature of the present invention is that information of a typical histogram of black pixels appearing in an image scanned at a predetermined angle with respect to a license plate is stored in the information processing device in advance.
The tilt angle in the scanning direction is selected so that the histogram of the black pixels appearing when the image of the license plate portion is scanned approaches the information of the stored histogram.

It is desirable to correct the reading angle of numbers and characters on the license plate according to the inclination angle.

It is desirable to detect a boundary between a character group at the upper part of the license plate and a character group at the lower part of the license plate in accordance with the recognition result, and select an inclination angle in the scanning direction individually for each of these two numbers and character groups. .

[0028]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present invention will be described with reference to FIGS.

The present invention relates to a license plate recognizing device, which moves aside a parked vehicle 12 while moving the vehicle.
, A memory 2 for storing image signals captured by the line sensor camera 3, and an image analysis unit for reading the image signals from the memory 2 and analyzing the image signals. The image analysis unit 5 extracts the image signal of the license plate 13 of the parked vehicle 12 by detecting the vertical edge, and further extracts the license plate 13 extracted by the image analysis unit 5. A data creation unit 6 as means for recognizing numbers and characters written on the license plate 13 from the partial image signal;
Scans an image based on the image signal of the license plate 13 in the horizontal direction and the vertical direction of the license plate 13 respectively, counts the number of pixels in which numbers and characters are displayed in the horizontal direction and the vertical direction of the license plate 13, and calculates the histogram. Is a license plate recognizing device that obtains the information of the number plate and recognizes the arrangement status of numbers and characters on the license plate 13 according to the information of the histogram.

Here, a feature of the present invention is that the data creating section 6 previously stores typical histogram information of an image of the license plate 13 scanned at a predetermined angle. 6 is to select the inclination angle in the scanning direction so that the obtained information of the histogram approaches information of the typical histogram. Further, the data creation unit 6 corrects the reading angle of numbers and characters on the license plate 13 according to the inclination angle.

The data generator 6 detects the boundary between the upper character group (Shinagawa 33) and the lower character group (A12-34) on the license plate 13 according to the recognition result, and determines the two characters. An inclination angle in the scanning direction is individually selected for each of the groups.

[0032]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present invention will be described with reference to FIGS. FIG. 1 is a flowchart showing the operation of the embodiment of the present invention. FIGS. 2 and 3 are diagrams for explaining boundary detection according to the embodiment of the present invention. FIG. 4 is a diagram for explaining the correction of the inclination of the character according to the embodiment of the present invention. FIG. 5 is a flowchart showing a procedure for selecting an inclination angle. FIG. 6 is a flowchart showing a procedure for calculating the width of a non-character portion.
The device configuration is common to FIG.

In the license plate recognition device of the present invention,
It is characterized in that accurate character recognition can be performed even on a license plate imaged obliquely by the data creation unit 6. In order to realize this, the data creation unit 6 shown in FIG. 9 performs processing of image data according to the flowchart shown in FIG.

An image signal of the parked vehicle taken by the first line sensor camera 3 is written into the memory 2 via the image recording unit 1. The image signal written in the memory 2 is read by the image analyzer 5 (S10).

The image analysis unit 5 extracts an area (candidate area) where the license plate is captured from the image signal of the parked vehicle (S11). This extraction utilizes the fact that a car has many horizontal line edges and a portion with many vertical line edges is a license plate portion.

Next, the horizontal inclination angle of the license plate image and the boundary between the upper character group and the lower character group of the license plate, which are features of the present invention, are detected (S12).
In the example of FIG. 2, the image of the license plate is tilted downward and left in the vertical direction, and is tilted so that the left end is closer to the front than the right end in the horizontal direction. When the inclination angle was set, a predetermined pattern could be obtained. Until the predetermined pattern is obtained, the inclination angle in the scanning direction is changed variously in the horizontal direction and the vertical direction, so that the inclination angle in the scanning direction can be appropriately adjusted with respect to the image of the license plate whose inclination angle is unknown in advance. The angle can be detected.

In the case of FIG. 2, the upper character group (Shinagawa 3
The boundary between 3) and the lower character group (A12-34) can be detected.

Similarly, the vertical inclination angle of the image of the license plate and the boundary of the characters of the character group below the license plate are detected (S13).
In the example of FIG. 3, the image of the license plate is inclined downward and leftward in the vertical direction, and is inclined so that the left end is closer to the left than the right end in the horizontal direction, as in the case of FIG. 2.
When the inclination angle in the scanning direction is set at the histogram position shown in FIG. Until the predetermined pattern is obtained, the inclination angle in the scanning direction is changed variously in the horizontal direction and the vertical direction, so that the inclination angle in the scanning direction can be appropriately adjusted with respect to the image of the license plate whose inclination angle is unknown in advance. The angle can be detected.

In the case of FIG. 3, the lower character group (A12-
34), the boundary between the characters can be detected. In this way, as shown in FIG. 2, when a boundary between the character group at the upper part of the license plate and the character group at the lower part of the license plate is detected, when scanning in the vertical direction is performed, the scanning start position is set at this boundary. By performing the scanning from to the lower part, it is possible to avoid that the character group at the upper part of the license plate affects the boundary detection of the character group at the lower part. Similarly, by performing scanning from this boundary toward the upper side, it is possible to prevent the character group below the license plate from affecting the boundary detection of the upper character group.

Since the license plate often bends at the boundary between the upper character group and the lower character group for structural reasons, by treating each of these two character groups individually, accurate character recognition can be achieved. It can be carried out.

Further, the characters on the license plate are corrected according to the detected inclination angle (S14). FIG.
FIG. 9 is a diagram for describing an example of correcting characters. The correction of the character is performed by converting the coordinate position according to the detected inclination angle, as shown in FIG.

The characters cut out in this manner are represented by numerals (S
15) Recognize kana characters (S16). In addition, the upper group of characters on the license plate ("Shinagawa 33")
Are detected (S17), and the cut-out characters are recognized (S18).

The procedure for selecting the tilt angle in the scanning direction will be described in more detail with reference to FIGS. This selection procedure can be applied in common to the horizontal and vertical directions of the license plate, the upper character group, and the lower character group.

The smallest possible angle of inclination mi that can be considered as the angle of inclination θ
It is set to nθ (S20). A histogram is created according to the inclination angle θ (S23). The total D of the change amounts of the histogram is calculated as D = _{i = 0 to npΣ} | H (i) −H (i + p) | (S24). Here, H (i), H (i
+ P) is a histogram, from H (i) to H (i +
The histogram of the range up to p) is shown. Therefore, p
Is a constant determined by the size of the range. Subsequently, the estimated amount A of the width of the non-character portion is calculated (S25). The evaluation amount R (θ) is calculated as R (θ) = a * D + b * A using the total amount D of change in the histogram and the estimated amount A of the width of the non-character portion (S26). Here, a and b are weighting constants, respectively. A value obtained by adding the inclination angle Δθ added to the inclination angle θ is set as a new inclination angle θ (S27).
In this procedure, the inclination angle θ is the maximum possible inclination angle max.
This is repeated until θ is reached (S21). Evaluation amount R
The maximum value is obtained from (θ), and the tilt angle θ at that time is the desired tilt angle (S22).

The procedure for calculating the estimated amount A of the width of the non-character portion shown in step 25 is started with the initial values A = 0 and i = 0 as shown in FIG. 6 (S30). If the histogram H (i) is smaller than the threshold value TH1 (S32), the value ΔA is added to the value of A (S3).
3) Further, "1" is added to the value of i (S34). This procedure is repeated until the value of i exceeds the width W of the license plate area (S31). Histogram H
When (i) becomes larger than the threshold value TH1, the value ΔA is not added to the value of the estimated amount A, and the value of the estimated amount A is calculated here.

[0046]

As described above, according to the present invention,
Accurate character recognition can be performed even for license plates that are photographed diagonally. Therefore, the photographing conditions can be set gently.

[Brief description of the drawings]

FIG. 1 is a flowchart showing the operation of an embodiment of the present invention.

FIG. 2 is a diagram for explaining boundary detection according to the embodiment of the present invention.

FIG. 3 is a diagram for explaining boundary detection according to the embodiment of the present invention.

FIG. 4 is a view for explaining correction of inclination of a character according to the embodiment of the present invention.

FIG. 5 is a flowchart showing a procedure for selecting a tilt angle.

FIG. 6 is a flowchart illustrating a procedure for calculating a width of a non-character portion.

FIG. 7 is a diagram showing a detection device for a parked vehicle.

FIG. 8 is a diagram showing an example of an illegal parking sticker.

FIG. 9 is a block diagram of a main part of a detection device for a parked vehicle.

FIG. 10 is a flowchart showing the operation of the image analysis unit.

FIG. 11 is a diagram showing an example of a license plate.

FIG. 12 is a view for explaining detection of a boundary between an upper character group and a lower character group;

FIG. 13 is a view for explaining detection of a boundary between characters in a lower character group.

FIG. 14 is a view showing a histogram of a license plate photographed obliquely.

FIG. 15 is a view showing a histogram of a license plate photographed obliquely.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Image recording part 2 Memory 3 First line sensor camera 4 Second line sensor camera 5 Image analysis part 6 Data creation part 7 Operation console 8 Display part 12 Parked vehicle 13 License plate

Claims (6)

[Claims] 1. A line sensor camera for photographing a parked vehicle while moving beside the vehicle, a memory for storing image signals photographed by the camera, and reading the image signals from the memory to read the image signals. Means for extracting the image signal of the license plate portion of the parked vehicle by detecting a vertical edge, and the license plate extracted by the extraction means. Means for recognizing numbers and characters written on the license plate from the image signal of the portion, wherein the recognition means scans an image based on the image signal of the license plate in a horizontal direction and a vertical direction of the license plate, respectively. The number of pixels for displaying numbers and letters is set in the horizontal and vertical directions of this license plate. A license plate recognizing device comprising means for counting and obtaining histogram information, and means for recognizing the arrangement of numbers and characters on the license plate according to the information of the histogram, wherein the license plate has a predetermined angle. Typical histogram information of the image scanned in is stored in advance,
A license plate recognizing device, comprising: means for selecting an inclination angle in a scanning direction such that the information on the histogram obtained by the means for recognizing approaches the information on the typical histogram. 2. The license plate recognition device according to claim 1, further comprising means for correcting a reading angle of numbers and characters on the license plate according to the inclination angle. 3. A number and character group (Shinagawa 3) on the upper part of the license plate according to the recognition result of the recognition means.
3) means for detecting a boundary between the lower numeral and the character group (A12-34), wherein the selecting means individually selects an inclination angle in the scanning direction for each of the two numbers and the character group. The license plate recognition device according to claim 1, further comprising means for performing. 4. A vehicle that is parked is photographed by a line sensor camera while moving beside the vehicle, an image of a license plate portion is extracted from the photographed image, and numerals and characters on the license plate are subjected to character recognition. A method of performing recognition using an information processing device, wherein information on a typical histogram of black pixels appearing in an image scanned at a predetermined angle with respect to a license plate is stored in the information processing device in advance, and A license plate recognition method, wherein an inclination angle in a scanning direction is selected such that a histogram of black pixels appearing when an image of a plate portion is scanned approaches a stored information of the histogram. 5. The license plate recognition method according to claim 4, wherein a reading angle of numbers and characters on the license plate is corrected according to the inclination angle. 6. A boundary between a number and a character group (Shinagawa 33) on an upper part of a license plate and a number and a character group (A12-34) on a lower part of the license plate is detected in accordance with the recognition result, and the two numbers and the character group are detected. 5. The license plate recognition method according to claim 4, wherein an inclination angle in the scanning direction is individually selected for each of the following.