JP2013054547A - Object recognition device and object recognition method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an object recognition device which can recognize an object in optional color, can recognize a recognition target even when the recognition target has similar color to a background, and also can perform processing at high speed.SOLUTION: An object recognition device extracts a plurality of pieces of color component information from a color image, generates a plurality of binarized images from the color image on the basis of the plurality of pieces of extracted color component information and a plurality of thresholds corresponding to the color component information, extracts an area included in the generated binarized images from the color image, and on the basis of the extracted area, recognizes an object in the area.

Description

  The present invention relates to a recognition apparatus and a recognition method for recognizing an object of an arbitrary color in a color image at high speed and lightweight using color information.

  Patent Document 1 and Non-Patent Document 1 disclose a method of recognizing a sign and a road sign from a color image. Since these perform recognition based on the color component information, they have a high processing speed. However, the color of the object to be recognized is predetermined as red / blue / yellow / white / black in Patent Document 1 and red / blue / yellow in Non-Patent Document 1. In both cases, the recognition object and the background cannot be recognized if they have similar colors.

  Non-Patent Document 2 discloses a method of recognizing an advertisement signboard from a color image. This can recognize an object of any color, and can recognize even if the recognition object and the background are similar in color. However, since the local invariant feature amount based on the edge is extracted and the advertisement signboard is recognized, the processing speed is slow.

JP 2004-213127 A

Moan, Yoshinao Aoki “Traffic Sign Recognition in Color Images”, IEICE Transactions. D-II2124-2135, 2004-12-01 Naoyuki Ichimura “Recognition of multiple advertising billboards based on local invariant features”, IPSJ Research Report. CVIM, 2005, 123-130, 2005-11-17

As mentioned above, in the prior art,
-Processing speed is fast.
-Recognize objects of any color.
-It can be recognized even if the recognition object and the background are similar colors.
There is no recognition device or recognition method that satisfies all of the above conditions.

  Therefore, an object of the present invention is to provide an object recognition device and an object recognition method that can recognize an object of any color, can recognize even if the recognition target and the background are similar colors, and have a high processing speed. To do.

  In order to achieve the above object, an object recognition apparatus according to the present invention is an object recognition apparatus for recognizing an arbitrary object from a color image. Based on a plurality of extracted color component information and a plurality of threshold values corresponding to the color component information, a generating unit that generates a plurality of binarized images from the color image, and a binary generated by the generating unit A region extracting unit that extracts a region included in the converted image from the color image; and a recognition unit that recognizes an object in the region based on the region extracted by the region extracting unit.

  The extraction means is preferably means for extracting color component information of hue, saturation, brightness, luminance, color difference (red), and color difference (blue) from the color image.

  Further, it is also preferable that the recognition means confirms the accuracy of the recognized object and does not perform any further object recognition when recognizing an object having a certain level of accuracy.

  Moreover, it is also preferable that the recognition means performs an object recognition process of the region based on the shape of the region and a pattern of a certain part in the region.

  In addition, it is preferable that a pattern storage unit that stores various image patterns is further provided, and the recognition unit specifies a type of an object to be recognized according to a combination of patterns acquired from the pattern storage unit.

  Moreover, it is preferable that the recognition means performs a recognition process using a predetermined area in the extracted area.

  Further, it is preferable that the recognition means performs recognition processing using a fixed portion centered on the center of gravity in the extracted region or a fixed region having the highest contrast of color information in the region.

  In order to achieve the above object, an object recognition method according to the present invention is an object recognition method for recognizing an arbitrary object from a color image, and includes an extraction step of extracting a plurality of color component information of the color image, and the extraction step. A generation step of generating a plurality of binarized images from the color image based on the extracted plurality of color component information and a plurality of threshold values corresponding to the color component information, and the binary generated by the generation step A region extracting step for extracting a region included in the converted image from the color image, and a recognition step for recognizing an object in the region based on the region extracted by the region extracting step.

  By using the object recognition apparatus and the object recognition method according to the present invention, it is possible to improve recognition performance when an object of an arbitrary color is recognized from a color image. Further, in the present invention, since the object extraction is performed using the color component narrowing-down technique, an original image having a lower resolution than that in the case of using an algorithm using a local invariant feature based on an existing edge is used. On the other hand, it is possible to perform recognition processing at high speed while maintaining high recognition accuracy.

  Therefore, it can be mounted on a conventional feature phone having a lower processing capability than a smartphone or the like. As a result, it becomes possible to enjoy an augmented reality application using image recognition even with a conventional feature phone.

The functional block diagram of this invention is shown. The processing flowchart of this invention is shown. An example of region extraction from a binarized image is shown.

  The best mode for carrying out the present invention will be described in detail below with reference to the drawings. First, the functional configuration of the object recognition apparatus 1 in the present embodiment will be described with reference to FIG. As shown in FIG. 1, the object recognition apparatus 1 includes a color space conversion unit 11, a binarized image generation unit 12, a region extraction unit 13, an object recognition unit 14, a color image storage unit 20, a threshold data storage unit 21, and a pattern storage. Part 22. When the object recognition apparatus 1 recognizes an object such as an arbitrary signboard or logo from a color image, the object recognition apparatus 1 presents related information related to the object to the user.

  The color space conversion unit 11 inputs a color image stored in the color image storage unit 20 and converts the color space. For example, an RGB color image is converted into a color image having hue, saturation, and brightness color component information using HSV (Hue Saturation Value) conversion and YCbCr (Luminance Chrominance-red Chrominance-blue) conversion, Conversion into a color image having each color component information of the color difference.

  The binarized image generation unit 12 generates a binarized image from the color image using the threshold corresponding to the color component information acquired from the threshold data storage unit 21 for the color image converted by the color space conversion unit 11. To do.

  The area extraction unit 13 performs a labeling process on the binarized image generated by the binarized image generation unit 12 and extracts an area in the binarized image. Further, an object recognition area is extracted from the same position of the color image.

  The object recognition unit 14 performs object recognition by comparing the region for object recognition acquired by the region extraction unit 13 with the pattern acquired from the pattern storage unit 22. The method for comparing the region and the pattern may be, for example, template matching using a cross-correlation coefficient, a sum of squares of differences, or a sum of absolute values of differences.

  The color image storage unit 20 stores a color image taken by a digital camera, a camera function of a mobile phone, a CMOS camera, or the like.

  The threshold data storage unit 21 stores a plurality of threshold values corresponding to the color components converted by the color space conversion unit 11. The threshold is set for a single color component or a plurality of color components.

  The pattern storage unit 22 stores various image patterns. For example, image patterns of signs and signs are stored.

  Next, the operation of the object recognition apparatus 1 in this embodiment will be described with reference to FIG.

  Step S <b> 1: The color space conversion unit 11 of the object recognition apparatus 1 acquires a color image from the color image storage unit 20.

  Step S2: Next, the color space conversion unit 11 converts the color space of the acquired color image. For example, a color image by RGB is converted into a color image having hue, saturation, brightness, and luminance, color difference (red), and color difference (blue) color component information.

  Step S <b> 3: Next, the binarized image generation unit 12 acquires a threshold corresponding to a single or a plurality of color components from the threshold data storage unit 21.

  Step S4: The binarized image generation unit 12 determines a threshold (for example, S (saturation) = 90) based on the color component information included in the color image color-converted by the color space conversion unit 11, Are compared with each other, and a binary image is generated with the result as a pixel value of 0 or 1. FIG. 3A shows an example of the binarized image generated here. In the case of thresholds for a plurality of color components (for example, S (saturation) = 90, V (lightness) = 50), the saturation and lightness of the pixel value are compared with each other, and binarization is performed based on the result. Generate an image.

  Note that. The threshold used in this determination is a value obtained by setting an arbitrary pixel value from 0 to 255 for each color component information of hue, saturation, brightness, luminance, color difference (red), and color difference (blue). It is set in combination with one or more color component information.

  Step S5: The region extraction unit 13 performs a labeling process on the binarized image generated in step S4 and extracts a region. FIG. 3B shows an example of a labeling process and extracted regions.

  Step S6: The region extracting unit 13 extracts a region at the same position as the region extracted in Step S5 from the color image acquired in Step S1.

  Step S7: Next, the object recognition unit 14 performs object recognition by comparing the region extracted in step S6 with the pattern acquired from the pattern storage unit 22. In step S7, object recognition is performed based on the shape of the region and the design of a predetermined portion in the region. The object recognition method may be, for example, template matching using a cross-correlation coefficient, a sum of squares of differences, or a sum of absolute values of differences.

  Further, in step S7, as a predetermined portion in the region, a fixed portion centered on the center of gravity in the region or a portion having the highest contrast of color information in the region can be used.

  Furthermore, in step S7, according to the combination of patterns acquired from the pattern storage unit 22, the type of object to be recognized (eg, bank logo, mobile carrier logo, etc.) can be specified.

  Step S8: It is confirmed whether or not the object recognition with all threshold values has been completed. If completed, the object recognition process ends. If not finished, the process returns to step S3, another threshold value is acquired, and object recognition is performed again. Thus, in the object recognition apparatus 1 of the present embodiment, the processing from step S3 to step S7 is executed a plurality of times using different threshold values. Thereby, an object of a plurality of colors can be extracted from the color image. Here, as an example of the termination condition, there are a case where the threshold number (= n) is executed, or the object recognition is performed with a certain accuracy or more in step S7.

  As described above, the object recognition apparatus 1 according to the present invention uses a plurality of threshold values for a plurality of pieces of color component information such as brightness and luminance, so that the object can be stabilized even when the illumination state differs for an arbitrary object. Object recognition. In particular, since color components of luminance and color difference are used, it is not easily affected by illumination. For this reason, even if the conventional technique cannot recognize the object due to the influence of illumination and the angle of light, the present invention can stably recognize the object.

  In addition, since recognition processing is performed using color components without extracting feature points from the color image, it can be performed at high speed.

  In step S3, when the boundary value between the outer frame portion and the background portion of the region to be recognized is set as a threshold value, the region can be extracted from the background of similar colors. Further, since a plurality of pieces of color component information and a plurality of threshold values are used, the recognition accuracy can be improved as compared with the prior art.

  Note that step S7 is executed with a plurality of threshold values for single or a plurality of pieces of color component information, so that a plurality of objects may be recognized. In this case, for example, an object whose template matching effectiveness is equal to or higher than a certain level can be recognized as an object in the color image.

  Moreover, all the embodiments described above are illustrative of the present invention and are not intended to limit the present invention, and the present invention can be implemented in other various modifications and changes. Therefore, the scope of the present invention is defined only by the claims and their equivalents.

DESCRIPTION OF SYMBOLS 1 Object recognition apparatus 11 Color space conversion part 12 Binary image generation part 13 Area extraction part 14 Object recognition part 20 Color image memory | storage part 21 Threshold data memory | storage part 22 Pattern memory | storage part

Claims (8)

An object recognition device for recognizing an arbitrary object from a color image,
Extracting means for extracting a plurality of color component information of the color image;
Generating means for generating a plurality of binarized images from the color image based on a plurality of color component information extracted by the extracting means and a plurality of threshold values corresponding to the color component information;
A region extracting unit that extracts a region included in the binarized image generated by the generating unit from the color image;
Recognition means for recognizing an object in the area based on the area extracted by the area extraction means;
An object recognition apparatus comprising:   2. The object recognition according to claim 1, wherein the extraction means is means for extracting color component information of hue, saturation, brightness, luminance, color difference (red), and color difference (blue) from the color image. apparatus.   3. The object recognition apparatus according to claim 1, wherein the recognition unit confirms the accuracy of the recognized object and does not perform object recognition any more when an object having a certain level of accuracy is recognized. .   The object recognition apparatus according to any one of claims 1 to 3, wherein the recognition unit performs an object recognition process of a region based on a shape of the region and a pattern of a certain portion in the region. . A pattern storage unit for storing various image patterns;
5. The object recognition apparatus according to claim 1, wherein the recognition unit specifies a type of an object to be recognized according to a combination of patterns acquired from the pattern storage unit.   The object recognition apparatus according to claim 1, wherein the recognition unit performs a recognition process using a predetermined region in the extracted region.   7. The recognition unit according to claim 1, wherein the recognition unit performs recognition processing using a fixed portion centered on the center of gravity in the extracted region, or a fixed region having the highest contrast of color information in the region. The object recognition apparatus of any one of Claims. An object recognition method for recognizing an arbitrary object from a color image,
Extracting a plurality of color component information of the color image; and
Generating a plurality of binarized images from the color image based on a plurality of color component information extracted in the extraction step and a plurality of threshold values corresponding to the color component information;
A region extraction step of extracting a region included in the binarized image generated by the generation step from the color image;
A recognition step for recognizing an object in the region based on the region extracted by the region extraction step;
An object recognition method comprising:

Images