JP2008225812A - Image discrimination device for specification object - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide an image discrimination device of a specification object, capable of coping with various type of specification objects, and highly precisely discriminating those specification objects. <P>SOLUTION: A first correction image A2 is generated by determining the colors of all pixels of an original image A1 including the periphery of a specification object from the color of the first peripheral region, and an identification candidate region A3 is detected based on the first correction image A2, and a second correction image A4 is generated by determining the color of the pixel corresponding to the identification candidate region A3 of the original image A1 from the color of the second peripheral region which is smaller than the first peripheral region, and the second correction image A4 is compared and collated with the pattern image A5 of the specification object so that the specification object can be identified. <P>COPYRIGHT: (C)2008,JPO&INPIT

Description

  The present invention relates to a standard image identification apparatus.

2. Description of the Related Art Conventionally, a technology of a standard image identification device that identifies a license plate as a standard product outside a vehicle has been known (see Patent Document 1).
JP 2006-309650 A

  However, in the conventional invention, since the standard is a license plate, the captured color image can be converted into a black and white image and can be accurately identified. However, if the standard is a sign or the like, more Since it is necessary to identify the color and shape, noise is likely to occur at the image processing stage, making identification difficult.

  SUMMARY OF THE INVENTION The present invention has been made to solve the above-mentioned problems, and an object of the present invention is to provide a standard image identification device that can handle a wide variety of standard products and can perform high-precision identification. It is.

  According to the first aspect of the present invention, a first correction image is generated in which the colors of all pixels of the original image including the periphery of the standard product are individually determined from the colors of the first peripheral region, and the first correction image is generated. And generating a second corrected image in which the color of the pixel corresponding to the identification candidate area of the original image is individually determined from the color of the second peripheral area smaller than the first peripheral area, A standard product is identified by comparing and collating the second corrected image with a pattern image of a standard product.

  In the first aspect of the present invention, a first corrected image in which the colors of all the pixels of the original image including the periphery of the standard product are individually determined from the colors of the first peripheral region is generated, and the first An identification candidate area is detected based on one correction image, and a second correction image in which the color of the pixel corresponding to the identification candidate area of the original image is individually determined from the color of the second peripheral area smaller than the first peripheral area is obtained. Since the standard product is identified by comparing and comparing the second corrected image and the standard pattern image, the standard product can be used for a wide variety of standard products and can be identified with high accuracy. Can be provided.

  Embodiments of the present invention will be described below with reference to the drawings.

Example 1 will be described below.
FIG. 1 is a system configuration diagram of a standard image recognition apparatus according to a first embodiment of the present invention, FIG. 2 is a diagram for explaining multicolor classification processing by the first corrected image generation unit according to the first embodiment, and FIG. FIG. 4 is a flowchart for explaining the operation of the standard image recognition apparatus according to the first embodiment. FIG. 4 is a diagram for explaining multicolor classification processing by the second corrected image generation unit according to the first embodiment.

  FIG. 5 is a diagram showing an original image of the first embodiment, FIG. 6 is a diagram showing a first corrected image by the first corrected image generation unit, and FIG. 7 is a diagram showing the target pixel in the multi-color classification process of the first corrected image. 2A and 2B are a diagram illustrating a case where a range of surrounding pixels for determining a color is reduced and a partially enlarged view of FIG.

First, the overall configuration will be described.
As shown in FIG. 1, in the standard product image identification apparatus according to the first embodiment, a central processing unit 1, an original image input unit 2, a first correction image generation unit 3, and a second correction image generation unit 4 An enlarged / reduced image generating unit 5, pattern image data 6, and various devices 7 are provided.

  The central processing unit 1 performs input / output control of the original image input unit 2, the first correction image generation unit 3, the second correction image generation unit 4, the enlarged / reduced image generation unit 5 and the pattern image data 6, and performs image processing. The sign is identified and the result is output to various devices 7.

  The original image input unit 2 outputs a 24-bit color original image captured by a CCD camera 8 mounted on the vehicle to the original image input unit 2 toward the outside of the vehicle.

The first corrected image generation unit 3 uses all the pixels of the original image input from the original image input unit 2 via the central processing unit 1 in colors (white, black, red, blue, yellow) used for signs. After the color classification, the first corrected image is output to the central processing unit 1.
At this time, the first corrected image generation unit 3 determines the colors of all the pixels of the original image individually, as shown in FIG. 2, the pixels of the first peripheral area 10 in which the color of the target pixel 9 is indicated by a bold line. Based on the decision.
Specifically, in the first embodiment, one of the five colors described above is obtained as the color of the target pixel 9 based on the luminance value, the saturation value, and the RGB value of each pixel in the two block ranges around the target pixel 9.

The second corrected image generation unit 4 classifies all the pixels in the label candidate region of the original image, which will be described later, input via the central processing unit 1 into one of the five colors described above, and then performs the second correction image. Is output to the central processing unit 1.
At this time, the second corrected image generation unit 4 determines the color of all pixels in the label candidate region of the original image individually, as shown in FIG. The determination is made based on the pixels in the region 11.
Specifically, in the first embodiment, one of the five colors described above is obtained from the luminance value, the saturation value, and the RGB value of each pixel in the one block range around the target pixel 9, and is set as the color of the target pixel 9. As a result, the color of the target pixel 9 is determined in the second peripheral area 11 smaller than the first peripheral area 10 used by the first corrected image generation unit 3.

  The enlarged / reduced image generation unit 5 enlarges and processes a marker candidate region of an original image (described later) input via the central processing unit 1, or enlarges / processes the second correction image according to the size of the pattern image (described later). It is designed to reduce and process.

The pattern image data 6 stores the number of mark pattern images as standard items, and can be sequentially read from the central processing unit 1 as necessary.
Each pattern image is drawn using the above-described five colors with a size of, for example, 16 × 16 pixels with respect to each corresponding label.

  Next, the operation of the standard image recognition apparatus of the first embodiment will be described based on the flowchart of FIG.

  First, in step S1, an original image A1 as shown in FIG. 5 is input from the original image input unit 2 to the central processing unit 1.

  Next, in step S2, the first corrected image generation unit 3 classifies all the pixels of the original image A1 into the colors (white, black, red, blue, yellow) used for the label, and then in FIG. The first corrected image A2 as shown is output to the central processing unit 1.

Next, in step S3, the central processing unit 1 detects a marker candidate area A3 in which a marker is drawn in the first corrected image A2 (see the corresponding diagram).
In the first embodiment, the marker candidate region A3 is detected based on the outer shape of the marker, but this is not restrictive.

  At this time, since the first peripheral area 10 is used when the original image A1 is classified into multiple colors, the outer color of the sign and the surrounding color are not combined, and sesame salt-like noise is less likely to occur. It becomes possible to detect with high accuracy.

  Next, in step S4, the central processing unit 1 cuts out an image region corresponding to the marker candidate region A3 of the original image A1 into a rectangular shape, and then in step S5, the enlarged / reduced image generation unit 5 performs vertical and horizontal double (or 4). (See the corresponding diagram).

  Next, in step S6, after the second corrected image generation unit 4 classifies all the pixels of the image enlarged in step S4 into the colors (white, black, red, blue, yellow) used for the label, The second corrected image A4 is output to the central processing unit 1 (see the corresponding diagram).

  At this time, since the multi-color classification is performed using the second peripheral area 11 smaller than the first peripheral area 10, the details are not impaired and a clearer image can be obtained.

  Even when the size of the cut-out image is small, it is enlarged in step S4, so that it is possible to prevent image details from being damaged.

  Furthermore, since only the marker candidate area A3 is classified into multiple colors, it is possible to prevent the influence of the outer color of the marker and the surrounding color and the influence of sesame salt-like noise, and the original image A1 is processed in full size. The processing load can be reduced compared to the case.

  Next, in step S7, the enlarged / reduced image generating unit 5 normalizes the second corrected image A4 according to 32 × 32 pixels having the same size as the pattern (see the corresponding diagram).

Next, in step S8, the central processing unit 1 collates the normalized second corrected image A4 with a plurality of pattern images A5 to identify a sign (see the corresponding diagram).
A general template matching or the like is used as a matching method.

Next, in step S <b> 9, the central processing unit 1 outputs the marker identification result to the various devices 7.
Even when the second corrected image A4 does not match the pattern image A5, the fact is output to the various devices 7.

  At this time, a box may be overlapped on the sign of the original image A1 so as to be visible to the driver using a head-up display device or the like (see the corresponding diagram).

Various devices 7 that output the identification result of the identified sign and the method of using the identification result of the sign are arbitrary.
For example, the identification result of the sign is output to the navigation device to update / add the map information, or the identification result of the sign is output to the head-up display device to indicate to the driver at night or when the visibility is poor in the tunnel. It is conceivable to display.

  Therefore, when the standard image identification device according to the first embodiment is used, the surrounding pixels for determining the color of the target pixel 9 are determined in the multicolor classification of the first corrected image A2 used for the detection of the sign area. While increasing the range, when performing multi-color classification of the second corrected image A4 used for sign collation, the sign detection and collation can be performed by reducing the surrounding pixel range that determines the color of the target pixel 9. It can be performed with high accuracy.

  Further, since the second corrected image A4 is obtained by image processing of the original candidate image A1 extracted from the identification candidate area A3, the image is clearer and larger than when the first corrected image A2 is cut out. The details can be prevented from being damaged.

Here, FIG. 7 shows an image when the range of surrounding pixels for determining the color of the target pixel 9 is reduced in the multi-color classification processing of the first correction image A2 used for detection of the sign.
As is clear from the corresponding diagrams in FIG. 6 and step S6, in this case, the outline color of the sign and the surrounding color are combined, and many sesame salt-like noises are generated. Detection and verification was practically difficult.

Next, the effect will be described.
As described above, in the standard product image identification apparatus according to the first embodiment, the colors of all the pixels of the original image A1 including the periphery of the standard product are individually determined from the colors of the first peripheral region 10. The first corrected image A2 is generated, the identification candidate area A3 is detected based on the first correction image A2, and the color of the pixel corresponding to the identification candidate area A3 of the original image A1 is individually set as compared with the first peripheral area 10. Since the second corrected image A4 determined from the color of the small second peripheral region 11 is generated, and the second corrected image A4 is compared with the pattern image of the standard product and compared, the standard product is identified. It is possible to provide a standard image identification apparatus that can handle various types of standard products and can perform high-precision identification.

  Further, since the first corrected image A2 is generated by determining the colors of all the pixels of the original image A1 including the periphery of the standard product individually from the colors of the first peripheral area 10 to the standard product. Colors unrelated to objects can be deleted in advance, and signs can be detected and collated with high accuracy.

Although the present embodiment has been described above, the present invention is not limited to the above-described embodiment, and design changes and the like within a scope not departing from the gist of the present invention are included in the present invention.
For example, the sizes of the first peripheral region and the second peripheral region can be appropriately set within a range satisfying the first peripheral region> the second peripheral region.

  In addition, the standard is not limited to a sign, but can be any object.

1 is a system configuration diagram of a standard image identification device according to a first embodiment of the present invention. It is a figure explaining the process of the multicolor classification | category by the 1st correction image generation part of the present Example 1. FIG. It is a figure explaining the process of the multicolor classification by the 2nd correction image generation part of the present Example 1. FIG. 3 is a flowchart for explaining the operation of the standard image identification device according to the first embodiment. It is a figure which shows the original image of the present Example 1. FIG. It is a figure which shows the 1st correction image by a 1st correction image generation part. In the process of multi-color classification of the first corrected image, FIG. 7A is a diagram illustrating a case where the range of surrounding pixels for determining the color of the target pixel is reduced, and FIG.

Explanation of symbols

A1 Original image A2 First correction image A3 Marker candidate area A4 Second correction image A5 Pattern image 1 Central processing unit 2 Original image A1 input unit 3 First correction image A2 generation unit 4 Second correction image A4 generation unit 5 Enlarged / reduced image Generation unit 6 Pattern image data 7 Various devices 8 CCD camera 9 Target pixel 10 First peripheral region 11 Second peripheral region

Claims (2)

Generating a first corrected image in which the colors of all pixels of the original image including the periphery of the standard product are individually determined from the colors of the first peripheral region;
Detecting an identification candidate region based on the first corrected image;
Generating a second corrected image in which the color of the pixel corresponding to the identification candidate area of the original image is individually determined from the color of the second peripheral area smaller than the first peripheral area;
A standard product image identification apparatus characterized in that the standard product is identified by comparing and collating the second corrected image with a standard pattern image. The standard product image identification device according to claim 1,
A standard in which the first corrected image is generated by individually determining the colors of all pixels of the original image including the periphery of the standard product from the colors of the first peripheral region to the colors adopted for the standard product. Image identification device.