JP3291873B2 - License plate recognition device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an apparatus for recognizing a license plate of a passing vehicle, for example, which is effective in automating toll collection on a highway or a parking lot and saving labor.

[0002]

2. Description of the Related Art Stopping and collecting tolls each time at a tollgate on a highway or at a tollgate in a parking lot has many problems that significantly impede the flow of vehicles. In addition, it is necessary to allocate a dedicated staff to collect such a fee, which is not economically preferable.

For this reason, for the purpose of facilitating the flow of vehicles and saving labor in toll collection, for example, on an expressway or a parking lot, the license plate of a passing vehicle is automatically recognized to automate toll collection. However, for example,
Various proposals have been made as disclosed in JP-A-76300 and JP-A-4-160692. Then, in order to automatically recognize the number described on the license plate, for example, JP-A-2-196400 and JP-A-2-196400
A license plate recognition method using image processing or a pattern recognition method as disclosed in Japanese Patent Application Laid-Open No. 226479 has been proposed.

In the proposed means for recognizing and reading the license plate of a vehicle, the license plate is imaged together with the vehicle by a single television camera or the like, and the license is recognized from the captured image. are doing. For this reason, for example, when the license plate is directly irradiated with the sun's light and the reflected light is present, when there is backlight, or when the license plate is concealed by another vehicle, etc. May be mistaken or unrecognizable.

[0005]

SUMMARY OF THE INVENTION The present invention has been made in view of the above-described circumstances, and recognizes a license plate attached to a vehicle under any conditions, and makes it possible to easily identify a number drawn on the license plate. Another object of the present invention is to provide a license plate recognizing device for a vehicle or the like that can read and recognize the vehicle reliably and can identify each passing vehicle without opposing the flow.

[0006]

SUMMARY OF THE INVENTION A license plate recognition device according to the present invention is used for a vehicle in which the vehicle is viewed from another angle in front of the vehicle.
Images of both identical license plates and their respective images
Performs signal feature extraction. And based on the extracted features
And each character recognition together with the recognition reliability
Based on character recognition results and their recognition reliability
To obtain the final result of character recognition of the license plate
Things.

[0007]

In the license plate recognizing device thus constructed, the scene of the vehicle including the license plate is
Images are respectively taken by a plurality of, for example, two TV cameras or the like, so that, for example, two image signals are obtained. These two image signals are separately extracted by different characteristic extracting means, respectively, a license plate portion is cut out from each image signal, and characters such as numbers drawn there are read, Be recognized.
In this case, the quality of the plurality of image signals differs depending on the imaging angle, the degree of light hit, and the like. Therefore, in each of the image signals, the recognition reliability is output together with the recognition character, and the determination unit outputs the recognition reliability. The character recognition of the license plate is performed based on the recognition reliability or the like. Therefore, the license plate recognition processing with very high reliability is performed, and the vehicle can be identified according to the flow of the vehicle.

[0008]

An embodiment of the present invention will be described below with reference to the drawings. FIG. 1 shows the configuration, in which a plurality of TV cameras 111, 112,... Are installed corresponding to gates and the like provided on a road through which a vehicle passes. This TV camera 111
, 112, ... are images of the entire view of the vehicle traveling on the road from different directions, and in particular, ensure that the license plate portion of the passing vehicle is in view.
For example, the vehicle is imaged from the front. In this case, for example, T,
Each of the V cameras 111, 112,...

An image signal representing an image captured by each of the plurality of TV cameras 111, 112,... Is input to a recognition system 12, which recognizes the TV.
Are provided with feature extraction units 131, 132,... To which image signals from the cameras 111, 112,. Each of the feature extraction units 131, 132,... Extracts the presence of a license plate in each of a plurality of captured image signals, and further extracts characters such as numbers drawn on the license plate.

The feature output extracted by each of the feature extraction units 131, 132,...
,... Are input to the recognition units 141, 142,.
In each case, recognition of the input extracted character is executed. In the determination unit 15, the character recognition units 141 and 14
2... Among the recognized characters, the character with the highest reliability is determined as a recognition result, and this is output as a recognized character. In the determination of reliability, for example, when character recognition is performed by template matching, a template with the highest matching degree is determined to have high reliability.

FIG. 2 shows a plurality of TV cameras 111, 112,.
This shows an example of the installation state of the vehicle. For example, a gate 21 is provided for a toll road 20 for which tolls are collected, and a plurality of gates 21 are provided for the gate 21 in the traveling direction of the vehicle 22, for example, a first and a second. Attach TV cameras 111 and 112. In this case, the first TV camera 111
The second TV camera 112 is installed at the lower part of the side of the gate 21, and is set, for example, to capture the image from the lower left front of the traveling vehicle 22. Here, the first and second TV cameras 111 and 11
2 includes a vehicle 22 including at least 23 license plates
The field of view over which the entire view moves is set.

When the license plate 23 of the running vehicle 22 is recognized, in the image taken from one angle, the license plate and the license plate are drawn on the license plate due to the influence of light reflection and concealment by other vehicles. In some cases, the character cannot be recognized, or the character may be erroneously recognized. However, as in this embodiment, a plurality of TV cameras 111, 112,.
Recognition is performed based on each of the plurality of captured images obtained in this manner, and the most reliable recognition result among them is output. Therefore, the situation of erroneous recognition or unknown recognition is effectively reduced.

The flow of processing in the feature extracting unit and the character recognizing unit in the apparatus configured as described above will be described with reference to the flowchart shown in FIG. This processing flow shows a portion corresponding to one image signal. First, in step 201, an image of a road on which a vehicle travels is input from an installed TV camera. Based on this input image, it is determined in step 202 whether or not a vehicle is included in the image. As the determination means, a method of examining the magnitude of a horizontal edge component which is a feature of the vehicle, a method of observing a temporal change of an image, and the like can be considered.

If the presence of the vehicle is confirmed in the image in step 202, the flow advances to step 203 to perform line copying. In this line copy processing, if the TV camera is of an interlace (jump operation) type, for example, a position shift occurs in an image of two fields, so the effect is reduced by copying every line. .

In step 204, a vertical edge is extracted from the input image signal. There are many vertical edges in the letters and numbers usually drawn on the license plate. In step 205, by detecting a region that matches the features of the license plate, the position of the license plate is detected and cut out. For the license plate to be detected, the density of the vertical edge is large, and 2
Because the characters are arranged in columns, the vertical edge is also 2
The license plate portion is cut out based on these characteristics.

In the next step 206, step 20
In step 5, a horizontal edge is extracted from the detected license plate area, and in step 207, the position of the character is detected and cut out from the horizontal edge projection histogram. Then, in step 208, the characters extracted in step 207 are binarized, and then recognition is performed by template matching, feature amount matching, or the like.

In step 209, it is determined from the matching level in step 208 whether or not the character recognition has succeeded. If not successful, it is determined that there is no character, that is, the plate cutout has failed. Return to 205 and cut out the plate again.

If it is determined in step 209 that the character recognition has been successful, the flow advances to step 210 to determine whether all the characters to be recognized have been read, and it is determined that not all the characters have been read. If so, the process proceeds to step 211, where the position of the unread character is estimated from the position of the read character, and the unread character is cut out. Then, after binarization is performed in step 212, recognition of unread characters is performed by template matching or feature amount matching.

At the time of character recognition at steps 208 and 211, the reliability of the recognition result is calculated. At step 213, the recognition result is output together with the reliability.

In the method of calculating the reliability, basically, the higher the degree of matching in the matching process, the higher the reliability.
In the case of template matching of a binary image, assuming that the input image is I _jk and the template is T _jk , the reliability R is calculated by the following equation, for example. Here, assuming that the size of the image is X pixels in the horizontal (j) direction and Y pixels in the vertical (k) direction, the reliability becomes maximum when “R = XY”.

[0021]

(Equation 1)

In the case of template matching, weighting may be performed for each pixel, and in the case of feature matching, weighting may be performed according to the type of feature. Fig. 1
The recognition processing is executed in each system of the plurality of TV cameras 111, 112,... Indicated by. The determination unit 15 determines the highest reliability among the results sent from each system in the final Output as recognition result. Table 1 below shows one example of the contents of the judgment.

[0023]

[Table 1]

In the embodiment, an example is shown in which recognition by a plurality of cameras is applied to detection of a license plate of a vehicle. However, recognition in a natural environment where environmental conditions cannot be controlled, for example, roads The same algorithm can be applied to character recognition of a plate constituting a sign.

[0025]

As described above, according to the license plate recognizing device of the present invention, a plurality of cameras capture license plate images from different directions, and recognize images of the license plate. The final judgment is made based on the recognition result of. If an image captured from one angle is not recognized due to the effects of natural phenomena,
Alternatively, even in a situation where erroneous recognition is performed, a captured image from another angle can be used, so that accurate recognition can be performed. In addition, since the reliability is determined for each character,
Even when one camera can read only the right half of the plate and the other can read only the left half of the plate, it is possible to recognize the entire plate surface, which is better than using a single camera. As a result, unknown or incorrect recognition results can be significantly reduced.

[Brief description of the drawings]

FIG. 1 is a configuration diagram illustrating a license plate recognition device according to an embodiment of the present invention.

FIG. 2 is a view for explaining the arrangement of cameras in the embodiment.

FIG. 3 is a flowchart illustrating a flow of a recognition process of the recognition device.

[Explanation of symbols]

111, 112, ... TV camera, 12 ... recognition system, 131,
132,... Feature extraction unit, 141, 142,.
Judgment unit, 20 ... road, 21 ... gate, 22 ... vehicle, 23 ... license plate.

Continuation of the front page (56) References JP-A-5-189598 (JP, A) JP-A-5-35912 (JP, A) JP-A-62-1121586 (JP, A) JP-A-3-25959 (JP) , U) Toshihiro Ueda Yoshihisa Yamanaka Hitoshi Yoneda, Image Processing Type Vehicle Number Reader, Nissin Electric Technical Report, Japan, Nissin Electric Co., Ltd., November 27, 1992, Vol. 37, No. 3, p. 24-27 (58) Field surveyed (Int. Cl. ⁷ , DB name) G06K 9/00-9/72 G06T 1/00 280 G08G 1/017 JICST file (JOIS)

Claims (1)

(57) [Claims] 1. A vehicle according to claim 1, further comprising:
A plurality of image pickup means for picking up an image of the same license plate and converting them into image signals, respectively; a plurality of feature extraction means to which image signals are supplied from each of the plurality of image pickup means and which perform characteristic extraction of each of the image signals; A plurality of character recognition means for performing each character recognition together with the recognition reliability based on the features extracted by each of the plurality of feature extraction means; and a character recognition result of the plurality of character recognition means and its recognition reliability. A determination unit for obtaining a final result of character recognition of the license plate based on the identification plate, and the vehicle is identified by an output from the determination unit.