JPH09231376A - Number plate detecting system - Google Patents

Abstract

PROBLEM TO BE SOLVED: To prevent the influence of noise in a number plate for a vehicle or the like or the color or the like of the number plate at the time of detecting the position of the number plate. SOLUTION: A character part detecting unit U detects a character part in a number plate and an original image segmenting part 7 segments the range of an original image corresponding to the character part. An original image binarizing part 8 binarizes the range and an original image area dividing part 9 divides the binarized image into plural areas based upon the connection of pixels. An original image character extracting part 10 extracts an area including the character part and a character group generating part 11 generates a character group by combining these extracted areas. Then a plate detecting part 12 detects the position of the number plate based upon the generated character group.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a technique for detecting the position of a license plate in an image of a vehicle such as an automobile or a motorcycle.

[0002]

2. Description of the Related Art In recent years, a license plate detection system for detecting a vehicle number printed on a license plate of a vehicle such as a running automobile or a motorcycle is being used for confirmation of passage on a toll road or automatic billing . In such a license plate detection system, first, a TV camera or the like is used to capture an image of the entire vehicle from the front or rear surface of the vehicle. Next, the photographed image is differentiated to obtain the contours of the characters "Naniwa", "Shonan", "33", "45", etc. in the license plate. Then, the differentiated image is binarized in order to clarify the difference in brightness of the image. Further, in order to detect the position of each character, the binarized image is divided into a plurality of regions based on the connection of pixels of the same value. After that, a plurality of character groups are generated by combining those areas. The position of the license plate is detected by comparing the generated layout of the character group with a template of “positional relationship between the character group and the license plate” prepared in advance.

FIG. 5 is a diagram showing a license plate. In the case of detecting this license plate by the conventional technique, by performing the above-described differentiation processing, binarization processing, and division processing, the area including "na", the area including "ni", and the area including "wa" are included. Region, region containing "3", region containing "3", region containing "O", region containing "2", region containing "8", region containing "-", region containing "2",
An area including "1" is created, and a plurality of character groups "Naniwa", "33", and
“O” and “28-21” are generated. Then, by comparing the positional relationship between the character groups and the size and number of each character group with the model of the “positional relationship between the character groups and the license plate” prepared in advance, the position of the license plate can be determined. Will be detected. The license plate position is detected by performing the same processing on the license plate "Shonan 45 Ki 00-15".

[0004]

By the way, in the conventional license plate detection system as described above, when an image of a vehicle is taken, it is originally divided by dirt or a light ray of disturbance attached to the license plate. Two regions to be detected are detected as one connected region,
Alternatively, one area that should be originally connected may be detected as two divided areas,
There was a problem that the position of the license plate could not be detected accurately. When the color scheme of green and white on the license plate is different from the color scheme of a normal license plate as in the license plate of FIG. 5B, the color distribution of the image after binarization is shown in FIG. There is a problem that the position of the license plate cannot be accurately detected because the color distribution of the image after binarizing the license plate of 5 (a) is opposite.

[0005]

The present invention employs the following structure in order to solve the above problems. <Structure> The license plate detection system of the present invention is a character detection unit that detects a character portion from an original image obtained by photographing a vehicle having a license plate, and an original that cuts out a portion corresponding to the character portion from the original image as a partial image. An image cutout unit, an original image binarization unit that binarizes the cutout partial image, and an original image region division unit that divides the binarized partial image for each region where pixels are connected. When,
An original image character extraction unit that extracts a region including a character portion based on a characteristic amount of the divided region; and a character group generation unit that generates a character group by combining the region including the character portion, It is characterized by comprising a plate detection unit for detecting the position of the license plate from the generated character group.

<Contents of Configuration> The contents of each configuration will be described below. The character portion refers to a range including characters and their surroundings. With the character detection unit, from the image of the vehicle,
It is a unit that detects the character portion on the license plate. For example, it is composed of a differentiating unit for differentiating an image, a dividing unit for dividing the differentiated image, a combining unit for combining the divided images, and the like. A character group is a set of related characters.

<Operations and Effects> In the license plate detection system of the present invention, after the rough position of the character portion is detected by executing the differentiating process, the dividing process, the combining process, etc. on the original image of the vehicle, the next time, Does not further utilize the image generated in the process up to the detection, but uses the rough position of the character portion to detect the detailed position of the character portion from the original image again. Therefore, noise is included in the original image, and when the rough position of the character part is detected, it is affected by noise, and the regions to be connected are divided or the regions to be divided are connected. Even at this time, since the detailed position of the character portion is detected again from the original image, the influence of noise can be reduced. In particular, when the process of detecting the approximate position of the character portion involves the differential process / binarization process, when the image is binarized, the differential process has already been easily affected by noise. Therefore, it is possible to reduce the influence of noise by detecting the detailed position of the character portion from the original image again.

Further, in the license plate detection system of the present invention, when the rough position of the character portion is detected through the differential processing, the pixels in the image obtained by the differential processing are converted into absolute values. Then, when detecting the detailed position of the character part from the original image, the original image is binarized with reference to this absolute value. This makes it possible to recognize whether it is the character part of the license plate based on white or the character part of the license plate based on green. Without, the position of the license plate can be detected.

[0009]

BEST MODE FOR CARRYING OUT THE INVENTION The license plate detection system of the present invention will be described below with reference to embodiments.
The embodiment is a specific example 1 and a specific example 2.
The main feature of is to detect the detailed position of the character part again after detecting the rough position of the character part,
The main characteristic point of the specific example 2 is that the pixels in the image subjected to the differential processing are converted into absolute values.

<Structure of Concrete Example 1> FIG. 1 is a block diagram of a license plate detection system of Concrete Example 1. This license plate detection system includes an original image input unit 1, a character detection unit U, an original image cutout unit 7, an original image binarization unit 8, an original image area division unit 9, an original image character extraction unit 10, and a character group generation. It is composed of a unit 11 and a plate detection unit 12. Further, the character detection unit U includes an image smoothing unit 2, an image differentiating unit 3, a differential image binarizing unit 4, a differential image area dividing unit 5, and a differential image character extracting unit 6.

The original image input unit 1 captures an image of a vehicle. The image smoothing unit 2 removes noise included in the image by smoothing the captured image of the vehicle. The image differentiating unit 3 differentiates the smoothed image. When there are characters whose contours are clear in the captured image, for example, characters on the license plate, the contours of those characters can be detected by this differentiating process.

The differential image binarization unit 4 binarizes the differentiated image. The differential image area dividing unit 5 divides the binarized differential image into a plurality of areas. The differential image character extraction unit 6 extracts, from each area of the differential image, an area estimated to include a character based on the connection state of the pixels in each area.

The original image cutout unit 7 is a differential image character extraction unit 6
The area on the original image corresponding to the position of the area extracted in step S4 is cut out. The original image binarization unit 8 binarizes the area on the clipped original image. The original image area dividing unit 9 divides the binarized original image into a plurality of areas. The original image character extraction unit 10 extracts an area estimated to include a character from the divided areas based on the connection state of pixels in each area. The character group generation unit 11 generates a character group by combining the areas with each other based on the size and the feature point of the extracted area. Plate detector 12
Detects the position of the license plate from the positional relationship between the character groups while referring to the model of the license plate prepared in advance, that is, the relationship between each character group in the license plate and the license plate.

<Operation> FIG. 2 is an operation flowchart of the license plate detection system of the first specific example. The operation of the license plate detection system of Specific Example 1 will be described below with reference to the flowchart thereof. Step S10: The original image input unit 1 captures an image of a vehicle or the like and A / D-converts the captured image to form a multi-valued image. It is also possible to previously record the vehicle image as a multi-valued image in a recording device such as an HD or MO not shown. Step S20: The image smoothing unit 2 smoothes the multivalued image. As a result, the noise component existing in the image can be removed.

Step S30: The image differentiating section 3 differentiates the smoothed image. As a result, a large differentiation result can be obtained at a place where the contrast is clear, such as the contrast between the back (white portion) of the license plate and the character (green portion). Then, from these large differentiation results, it is possible to know where the outline of the character is. Note that smoothing and differentiation can be simultaneously performed by a filter having both the function of the image smoothing unit 2 in step S20 and the function of the image differentiating unit 3 in step S30. Step S40: The differential image binarization unit 4 binarizes the differentiated image using a preset threshold as a reference. This makes it possible to obtain an image in which the character portion in the license plate is clear.

Step S50: Differential image area dividing unit 5
Divides the binary image into a region in which the pixels in the binary image are connected. In addition, after removing noise in the binary image by contraction processing or expansion processing from the image binarized by the differential image binarization unit 4,
The differential image area dividing section 5 can also divide the noise-free image. Step S60: The differential image character extraction unit 6 calculates the feature amount such as the connection of pixels and the size and shape of the region for each area, determines whether the area includes the character portion, and determines the character portion. Extract the containing area.

Step S70: The original image cutout unit 7 determines the area on the original image corresponding to the area on the differential image estimated at Step S60, that is, the area on the image photographed by the original image input unit 1. cut. Step S80: The original image binarization unit 8 binarizes, for each of the cut out regions, a threshold value set in advance as a reference. It is also possible to set the threshold value in the middle of the process without setting it in advance. In this case, the feature amount calculated by the differential image character extraction unit 6 can be taken into consideration, or the feature such as the density distribution of the cut-out region on the original image can be used.

Step S90: Original image area dividing section 9
Divides the binary image into a region in which pixels are connected to each other. As in the case of the differential image region dividing unit 5, the original image region dividing unit 9 divides the binary image into each region after removing noise of the binary image by contraction processing or expansion processing. Is also possible. Step S100: The original image character extraction unit 10 calculates, for each of the divided areas, a feature amount such as a size / shape of the area or a connection of pixels, and determines whether the area includes a character portion.
Extract the area containing the character part.

Step S110: Character group generator 11
Generates a character group by combining the areas with each other based on the positional relationship between the areas and the feature amount. Step S120: The plate detection unit 12 detects the position of the license plate by comparing a prepared template showing the positional relationship between the character group and the license plate with the generated character group. When recognizing what the characters are on the license plate, the image data, the feature amount, etc. are detected for each character of the character group that is estimated to be present on the license plate.

<Description of Specific Example 2> Next, a license plate detection system of Specific Example 2 will be described. FIG.
It is a block diagram of the license plate detection system of the specific example 2. Further, FIG. 5 is an operation flowchart of the license plate detection system of the second specific example. Since the difference from the specific example 1 is the absolute value conversion unit 20, only the operation of the absolute value conversion unit 20 itself and the operation related to the absolute value conversion unit 20 will be described below.

Step S35: The absolute value conversion section 20 generates an absolute value image by calculating the absolute value of each pixel in the differential image calculated by the image differentiating section 3 of step S30. Step S80: The original image binarization unit 8 uses the absolute value binarization unit 2.
Regarding the absolute value image generated with 0, whether the license plate is a white-based type (passenger vehicle type) or a green-based type (commercial vehicle type) based on the absolute value distribution To judge.
Based on this determination, the subsequent original image area dividing unit 9
Determines whether to divide into a plurality of areas with reference to the connection of green pixels or to divide into a plurality of areas with reference to the connection of white pixels.

[Brief description of drawings]

FIG. 1 is a block diagram of a license plate detection system according to a specific example 1.

FIG. 2 is an operation flowchart of a specific example 1.

FIG. 3 is a block diagram of a license plate detection system of specific example 2.

FIG. 4 is an operation flowchart of a specific example 2.

FIG. 5 is a diagram showing a license plate.

[Explanation of symbols]

 U Character part detection unit 7 Original image cutout unit 8 Original image binarization unit 9 Original image area division unit 10 Original image character extraction unit 11 Character group generation unit 12 Plate detection unit

Claims (2)

[Claims] 1. A character detection unit for detecting a character portion from an original image obtained by photographing a vehicle having a license plate, an original image cutout portion for cutting out a portion corresponding to the character portion from the original image as a partial image, and the cutout portion. An original image binarizing unit that binarizes the divided partial image, an original image region dividing unit that divides the binarized partial image for each region where pixels are connected, and the divided region Based on the feature amount of the original image character extraction unit for extracting a region including a character portion, a character group generation unit for generating a character group by combining the region including the character portion, the generated character group A license plate detection system, comprising: a plate detection unit that detects the position of the license plate. 2. The number according to claim 1, wherein the character detecting unit comprises a differentiating part for differentiating the original image and an absolute value converting part for calculating absolute values of pixels of the partial image. Plate detection system.