JPH09147049A - Number plate position detector - Google Patents

Abstract

PROBLEM TO BE SOLVED: To reduce erroneous extraction of a position other than the number plate position and to reduce the burden for processing. SOLUTION: A vehicle picture is subjected to smoothing and horizontal- direction second-order differential processing by a smoothing and horizontal- direction second-order differential processing part 12 to obtain a differential picture having a large value in the part of a picture pattern like a number plate characters part. This differential picture is binarized by a part 13, and this binary picture is divided into areas by a part 14 to extract the areas like number plate characters by a part 15, and a combination of areas like a number plate is extracted from positional relations between areas or the like by parts 16 and 17, thus determining the number plate position.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention is provided in a vehicle number recognition device or the like for recognizing a vehicle number or the like written on a license plate in a vehicle image, and is a number for accurately detecting the position of the license plate in the vehicle image. The present invention relates to a plate position detecting device.

[0002]

2. Description of the Related Art Conventionally, this type of device is disclosed in Japanese Patent Publication No. 4-143.
No. 94 discloses that a vehicle image is subjected to spatial quadratic differential processing to obtain a differential image, the differential image is binarized by a predetermined threshold value to obtain a binary image, and area division is performed on the binary image. Through the processing, the area of the partial image of the pattern indicated by the connection of the characteristic data and its vertical and horizontal dimensions are obtained to detect the partial image that is a candidate number for the vehicle number in the license plate, and between the detected multiple partial images. The group of partial images having a predetermined positional relationship is examined as one block, and a block having a high probability as a vehicle number in the license plate is obtained from the number of partial images included in each block and the block size. Then, a predetermined area centered on the position of the block is detected as the license plate position in the vehicle image.

[0003]

However, the above-mentioned conventional license plate position detecting device method has the following problems and it is difficult to solve them. That is, the spatially second-order derivative filter used conventionally extracts a characteristic pattern such as a license plate character in the vehicle image by spatially secondarily differentiating the input vehicle image in the horizontal and vertical directions. There is. However, there are many horizontal structures such as bumpers and grills near the license plate of the vehicle, and a large number of these horizontal structures are also extracted by the spatial second derivative filter, and the area / dimension check unit and the positional relationship check are performed. Since the number of candidate areas in a copy becomes very large, it has been a cause of erroneous recognition and processing load.

Further, the spatial second derivative filter is easily affected by noise, and if the input vehicle image contains noise, the license plate position is erroneously detected.

[0005]

In order to solve the above-mentioned problems, the present invention relates to a license plate in a vehicle image used for a vehicle number recognition device or the like for recognizing a vehicle number etc. written on a license plate in a vehicle image. In a license plate position detecting device for detecting the position of a vehicle, an image of a vehicle is taken by using an image input means, and the image is input as a multi-value digitized input image, and a smoothing of the input image. Smoothing / horizontal quadratic differential processing unit for performing differential processing and horizontal quadratic differential processing to generate a differential image having a large value in a license plate character portion, and a license plate character portion by binarizing the differential image And the like, a binarization processing unit that obtains a binary image mainly containing, etc., a region division processing unit that divides the binary image for each connected region, and an area and a large area for each connected region. A region feature processing unit that calculates a region feature amount such as a number and extracts a plurality of license plate character-like regions, and a positional relationship between the plurality of license plate character-like regions extracted by the region feature processing unit and between the region feature amounts. The relationship between the areas for extracting the combination of areas likely to be license plates is determined based on the relationship between the areas, and the position of the license plate is determined based on the position of the combination of areas like the license plates extracted by the area relationship processing unit. A plate extracting unit that outputs this as a detection result is provided.

[0006]

1 is a block diagram showing an example of an embodiment of a license plate position detecting device of the present invention. In the figure, a vehicle image input unit 11 captures an image of a vehicle using an image input means such as a TV camera and inputs this as a multi-value digitized input image.

The smoothing / horizontal quadratic differential processing unit 12 performs smoothing processing on the input image and horizontal quadratic differential processing to generate a differential image having a large value in the license plate character portion or the like. The binarization processing unit 13 binarizes the differential image to obtain a binary image mainly containing license plate character portions and the like.

The area division processing section 14 divides the binary image into connected areas. The area feature processing unit 15 calculates the feature amount such as the area and size of the area for each of the connection areas, and extracts the area that looks like a license plate character. The inter-region relationship processing unit 16 extracts a combination of license plate-like regions by judging from the positional relationship between the plurality of license plate character-like regions extracted by the region feature processing unit 15 and the relationship between the region feature amounts. To do. The plate extraction unit 17 determines the license plate position based on the position of the combination of the regions that are likely to be license plates extracted by the inter-region relation processing unit 16 and outputs this as the detection result.

The operation of the embodiment of the present invention will be described below with reference to FIG. First, the vehicle image input unit 11 captures an image of a vehicle using an image input means such as a TV camera, and A / D-converts the image to input it as a multi-value digitized input image. Alternatively, a vehicle image previously stored in the storage device as a multivalued digital image is input as an input image.

FIG. 2 is a block diagram of the smoothing filter, and FIG. 3 is a block diagram of the horizontal quadratic differential filter. In the smoothing / horizontal quadratic differential processing unit 12, the input image is shown in FIG. Smoothing processing is performed using a smoothing filter as shown in the figure or a well-known median filter to reduce noise components in the input image, and a horizontal second derivative filter as shown in FIG. By performing the second differential processing, a differential image having a large value in the image pattern portion such as the license plate character portion is generated.

FIG. 4 is a block diagram of a smoothing / horizontal quadratic differential filter, which corresponds to the input image shown in FIG.
It is also possible to perform smoothing processing and horizontal quadratic differential processing by using a filter having both the smoothing function and the horizontal quadratic differential function as shown in FIG. The binarization processing unit 13 binarizes the differential image to obtain a binary image mainly containing license plate character portions and the like. For the binarization process, a method of setting a threshold value in advance, a method of automatically determining the threshold value, which is a known technique, a method of adaptively setting the threshold value for each pixel, or the like can be used.

The area division processing section 14 divides the binary image into connected areas in which pixels of the binary image are connected by a labeling process which is a known technique. Also, 2
The binary image obtained by the binarization processing unit 13 may be subjected to the noise removal of the binary image by a well-known process such as contraction and expansion, and then the above-described area division process may be performed.

The area feature processing section 15 calculates the area, size, and other area feature amounts of each area for each connection area,
A plurality of regions like license plate characters are extracted from the plurality of connected regions divided by the region division processing unit 14.
In the inter-region relation processing unit 16, it seems to be a license plate by judging from the positional relation between each of the plurality of regions which are likely to be license plate characters extracted by the region characteristic processing unit 15, the relation between the region characteristic amounts of each region, and the like. Extract a combination of regions.

The plate portion extraction unit 17 determines the license plate position in the input image based on the combination position of the license plate-like regions extracted by the inter-region relation processing unit 16 and uses this as the license plate position detection result. Output. In the smoothing / horizontal quadratic differential processing unit 12 described above, smoothing processing using a smoothing filter as shown in FIG. 2 or a median filter which is a known technique, and horizontal as shown in FIG. Horizontal quadratic differential processing using a directional quadratic differential filter and smoothing / horizontal directional using a filter that has both the smoothing function and the horizontal quadratic differential function as shown in FIG. Although the example of performing the secondary differential processing has been described, the present invention is not limited to the above-described embodiment, and smoothing / horizontal direction secondary differential processing can be performed by processing other than the filter processing.

[0015]

As described above, according to the present invention,
In the smoothing / horizontal quadratic differential processing unit, the input vehicle image is processed to have a quadratic differential effect in the horizontal direction, resulting in a large value in the image pattern part such as the license plate character part. It is possible to obtain a differential image, and it is possible to solve the problem that a large number of horizontal structures other than the license plate in the vicinity of the license plate of the vehicle are extracted as in the conventional case. As a result, the portions other than the license plate are erroneously extracted, and the number of candidate areas to be processed is reduced, which reduces the processing load and increases the processing speed.

Further, since the smoothing / horizontal quadratic differential processing unit processes the input image so as to have a smoothing effect, noise contained in the input image can be reduced, and noise other than the license plate can be reduced. It is much less likely to accidentally extract parts.

[Brief description of the drawings]

FIG. 1 is a block diagram showing an example of an embodiment of a license plate position detection device of the present invention.

FIG. 2 is a block diagram of a smoothing filter

FIG. 3 is a configuration diagram of a horizontal second-order differential filter.

FIG. 4 is a block diagram of a smoothing / horizontal quadratic differential filter.

[Explanation of symbols]

 11 vehicle image input unit 12 smoothing / horizontal quadratic differential processing unit 13 binarization processing unit 14 region division processing unit 15 region feature processing unit 16 region relation processing unit 17 plate unit extraction unit

Claims (4)

[Claims] 1. A license plate position detecting device for detecting a position of a license plate in a vehicle image, which is used for a vehicle number recognizing device or the like for recognizing a vehicle number or the like written on a license plate in a vehicle image, and an image input means is provided. A vehicle image input unit that captures an image of a vehicle using it and inputs this as a multi-value digitized input image, and a license plate character unit that performs smoothing processing and horizontal second derivative processing of the input image A smoothing / horizontal quadratic differential processing unit for generating a differential image having a large value, and a binarization processing unit for binarizing the differential image to obtain a binary image mainly including license plate character portions and the like. A region division processing unit that divides the binary image into connected regions, and calculates region feature amounts such as the area and size of the regions for each connected region to obtain license plate characters. A combination of the region feature processing unit that extracts a plurality of regions, and the number plate region that is determined by the positional relationship between the plurality of license plate character regions extracted by the region feature processing unit and the relationship between the region feature amounts. An inter-region relationship processing unit that extracts the license plate position, and a plate unit extraction unit that determines the license plate position based on the position of the combination of the license plate-like regions extracted by the inter-region relationship processing unit and outputs this as the detection result. A license plate position detecting device characterized by being provided with. 2. The license plate position detecting device according to claim 1, wherein the smoothing / horizontal quadratic differential processing unit uses a smoothing process using a smoothing filter and a horizontal quadratic differential filter. By performing horizontal second derivative processing on the multi-value digitized input image, smoothing
A license plate position detection device characterized by performing a second-order differential process in the horizontal direction. 3. The license plate position detection device according to claim 1, wherein the smoothing / horizontal quadratic differential processing unit uses a filter having both a smoothing effect and a horizontal quadratic differential effect. A license plate position detecting device characterized by performing smoothing / horizontal quadratic differential processing on a multi-value digitized input image. 4. The license plate position detecting device according to claim 1, wherein noise removal processing is performed by performing contraction and expansion processing on the binary image obtained by the binarization processing unit. License plate position detection device.