JP3501907B2 - Color identification device - Google Patents

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a color discriminating apparatus for discriminating a color of a target region existing in a color image by comparing it with a predetermined registered color, and more particularly to a color discriminating device in which the color of the target region contained in the color image is a background region. The present invention relates to a color identification device that accurately identifies the color of the target area while removing the influence of the color of the background area when mixed with the color.

[0002]

2. Description of the Related Art Conventionally, in a fruit and vegetable sorting field, after storing fruits and vegetables in a corrugated cardboard box stamped with grade letters and grade letters for indicating the quality of fruits and vegetables, the letters of equal grade are read. It is sorted and shipped by equal rank.

Here, since the equal rank characters are classified according to not only the type of character but also the character color,
When automatically recognizing such equidistant characters, the character color must also be identified.

For this reason, conventionally, the color information of the original image having RGB values is converted into the Munsell color system HSL space or the La ^* b ^* color system uniform perceptual color space in which the color tone of the composite color is easily controlled. Techniques are used to convert and identify colors in such a space.

For example, color information of an original image having RGB values is HSL-converted to obtain a hue (H: hue) and a saturation (S: satur).
ation) and brightness (L: lightness) components,
The distances between the decomposed components and the registered colors H, S, and L that are registered in advance are obtained, and it is determined whether the color of the original image to be identified matches the registered color based on the obtained distances.

[0006]

However, the above-mentioned conventional technique is effective when the original image to be identified has color information in a good state, but the color information of the original image affects the influence of the background color. When it is received and deteriorated, the identification accuracy is naturally lowered.

[0007] Specifically, when the equal-class characters printed on the cardboard box are mixed due to the influence of the color of the cardboard box itself, the character color of the equal-class characters is different from the original character color. Since it is already different, even if such a character color is converted into the HSL space or the uniform perceptual color space, the character color cannot be properly identified.

In particular, when a stamp-printed character is faint or the ink is thin, the color of the character is closer to the background color due to the influence of the background color. Therefore, if the color similarity is simply calculated by the Euclidean distance or the like. Most colors are similar to the background color,
It becomes indistinguishable.

Therefore, the present invention solves the above problems by
To provide a color identification device capable of accurately identifying the color of a target area while removing the influence of the color of the background area when the color of the target area included in a color image is mixed with the color of the background area. With the goal.

[0010]

To achieve the above object, a first aspect of the present invention is to register a predetermined color in advance as a registered color and compare the color of a target area included in a color image with the registered color. In the color identification device identified by, the detection means for detecting the color of the target area and the color of the background area located on the background of the target area, the color of the target area and the color of the background area detected by the detection means First done in color space
Angle and the registered color and the color of the background area detected by the detection means to calculate a second angle in the color space, the first angle calculated by the angle calculation means and the And a color identification unit for identifying whether the color of the target area corresponds to the registered color based on an angle difference from the second angle.

In a second aspect based on the first aspect, the detecting means separates the target area and the background area from the color image, and the target area and the background area separated by the separating means. A conversion unit that converts a color of each pixel into a unit color that does not include a lightness component, and a unit color that has the highest appearance frequency among the unit colors of each pixel of the target region converted by the conversion unit is the color of the target region. And a color selection unit that selects the unit color having the highest appearance frequency among the unit colors of each pixel of the background region converted by the conversion unit as the color of the background region.

[0012]

BEST MODE FOR CARRYING OUT THE INVENTION Embodiments of the present invention will be described below with reference to the drawings. In the present embodiment,
A case where the character color of the character included in the color image is set as the identification target will be described.

FIG. 1 is a diagram showing the configuration of a color identification apparatus 10 used in this embodiment.

The color identification apparatus 10 shown in FIG. 1 simply compares a color registered in advance (hereinafter referred to as "registered color") with a character color of a character to be identified (hereinafter referred to as "identified character"). Instead, the degree of coincidence between the angle formed by the character color of the identification character and the background color attached to the background of the identification character on the color space and the angle formed by the registered color and the background color on the color space Based on this, it is identified whether the character color of the identification character matches the registered color.

That is, since the influence of the background color cannot be removed by simply comparing the character color of the identification character with the registered color,
In this embodiment, the color identification is performed in consideration of the background color.

The color identification device 10 has a color registration mode in which a registered color to be compared with the identification characters is registered in advance, and an identification mode in which the identification characters are identified by comparing the registered colors. These modes are switched using an operation unit (not shown).

As shown in FIG. 1, this color identification apparatus 10
Is an image input unit 11, a cutout processing unit 12, a character background separation unit 13, an identification processing unit 14, and a normalization processing unit 15.
And a histogram creation unit 16, an angle calculation unit 17, a color registration processing unit 18, and a registered color storage unit 19.

Here, among these functional units, the image input unit 11, the cutout processing unit 12, the character background separation unit 13, the identification processing unit 14, the normalization processing unit 15, the histogram creation unit 16, the angle calculation unit 17, and the The registered color storage unit 19 functions in the identification mode, and the image input unit 11, the normalization processing unit 15, the histogram creation unit 16, the color registration processing unit 18, and the registered color storage unit 19 perform the color registration. Function in mode.

The angle calculating means and the color identifying means in the first invention correspond to the angle calculating section 17 and the identification processing section 14, respectively, and the separating means, the converting means and the color selecting means in the second invention are: Character background separation part 1
3 corresponds to the normalization processing unit 15 and the histogram creation unit 16.

The image input unit 11 is an input unit that optically takes in color image data having a registered color or an identification character and outputs it to the color registration processing unit 18 or the cutout processing unit 12, respectively.

That is, when the color identification apparatus 10 operates in the color registration mode, the color registration processing unit 1 processes the image data.
8 and outputs the image data to the cutout processing unit 12 when operating in the identification mode.

The cut-out processing unit 12 cuts out a rectangular area (hereinafter referred to as a “character template area”) containing an identification character from the image data received from the image input unit 11 and outputs it to the character background separation unit 13. Is.

When performing such cutout, the character template region of the identification character is cut out after positioning using template matching such as normalized correlation or generalized Hough transform.

FIG. 2 is a diagram showing an example of a character template area cut out from the image data. The character template area 20 shown in FIG. 2 is a character area 20a indicating the letter "L" and other than the character area 20a. Background area 20b.

The character background separating unit 13 is a cutout processing unit 1.
Reference numeral 2 denotes a processing unit that separates the character template region cut out from the image data into a character region and a background region, and outputs them to the identification processing unit 14.

For example, the character background separating unit 13 is shown in FIG.
When the character template area 20 shown in (1) is received, the character area 20a indicating the letter'L 'and the background area 20b are separated.

When the character processing section 14 receives the character area and the background area from the character background separating section 13, the identification processing section 14 correlates the background area color and the character area color with each other and the background area color and the registered color relationship with each other. Is performed, and the identification process is performed using the respective mutual relationships as identification indices.

Specifically, the identification processing unit 14 uses the normalization processing unit 15 to normalize the colors of the background area and the character area in the RGB space, and then uses the histogram creation unit 16 to normalize them. A two-dimensional histogram is created with the two components (r, g) of the color as the vertical axis and the horizontal axis.

The coordinates (rb, gb) having the maximum frequency of the two-dimensional histogram corresponding to the background area are set as the feature amount of the background area, and the coordinates (rc, gc) having the maximum frequency of the two-dimensional histogram corresponding to the character area are set. ) Is the feature amount of the character area. The color registration processing unit 18 calculates in advance the feature amount (rt, gt) of the registered color, and the registered color storage unit 1
It is stored in 9.

Next, the angle calculator 17 calculates the angle θc of the color of the background area and the color of the character area and the angle θt of the color of the background area and the registered color, and these angles are within a predetermined range. If it is, the color of the character area is identified as the same as the registered color.

The reason why the maximum frequency of the two-dimensional histogram is used in this embodiment is that the background color can be detected stably even if the background is soiled by a stamp or the like.

The normalization processing section 15 receives the pixel values (R, G, and G) of the image data from the identification processing section 14 or the color registration processing section 18.
When B) is received, r = R / (R + G + B) g = G / (R + G + B) is calculated, and the calculated r and g are output to the identification processing unit 14 or the color registration processing unit 18.

FIG. 3 is a diagram showing the concept of the processing performed by the normalization processing unit 15. As shown in FIG. 3, the normalization processing unit 15 has R indicating the color of the background area or the character area.
The color of the background area or the character area is normalized by obtaining the intersection (r, g, b) of the color vector C on the GB space and the unit plane on the RGB space.

That is, in the RGB space, the difference in color can be grasped as the difference in the direction of the color vector.
By normalizing using the intersection with the unit plane in the RGB space, the color of each pixel in the background region and the character region is converted into a color (unit color) not including the lightness component.

Here, the reason why only two components of the r component and the g component are calculated is that the b component is uniquely determined from the relational expression of r + g + b = 1 on the unit surface and the color can be specified.

The histogram creating unit 16 prepares a two-dimensional plane having the R component as the horizontal axis and the G component as the vertical axis, and performs a process of incrementing the frequency of the coordinates (r, g) sequentially received from the identification processing unit 14. It is a processing unit that repeatedly calculates a two-dimensional histogram and outputs the coordinates having the maximum frequency of the created two-dimensional histogram to the identification processing unit 14.

The angle calculation unit 17 receives the coordinates (rb, gb) corresponding to the background region, the coordinates (rc, gc) corresponding to the character region, and the coordinates (rt, gt) corresponding to the registered color from the identification processing unit 14. ) Is received, the angle θc formed by the coordinates (rb, gb) and the coordinates (rc, gc) and the coordinates (rb, g
b) and the angle θt formed by the coordinates (rt, gt) are calculated and output to the identification processing unit 14.

Specifically, two calculation formulas of θc = tan-1 ((gc-gb) / (rc-rb)) θt = tan-1 ((gt-gb) / (rt-rb)) It will be used to calculate such an angle.

FIG. 4 is a diagram showing the concept of the processing of the angle calculating unit 17, and as shown in FIG.
In FIG. 7, the coordinates (rc, gc) corresponding to the character area and the coordinates (rt, gt) corresponding to the registered color are not obtained as the angles with the origin, but the coordinates (rb) corresponding to the background area are obtained. , Gb).

That is, the angle formed with the origin 0 is simply obtained,
If no consideration is given to the color of the background area, when the character color of the identification character is mixed with the background color, it is virtually impossible to identify the original character color of the identification character.

For this reason, the angle calculating unit 17 obtains the angle when the coordinates (rb, gb) corresponding to the color of the background area are used as a reference.

The color registration processing section 18 is a processing section that functions in a color registration mode for registering a registered color that is a target of comparison of identification characters, and specifically, a predetermined color including a color registered from the image input section 11. After cutting out the region, the normalization processing unit 15 and the histogram creation unit 16 are used to calculate the coordinates (rt, gt) having the maximum frequency of the two-dimensional histogram and the registered color storage unit 19 as in the identification processing unit 14. To store.

The registered color storage unit 19 is a storage unit for storing the coordinates on the two-dimensional histogram corresponding to the registered color to be compared with the identification characters.

By using the color identification device 10 having the above-mentioned configuration, the mutual relationship between the color of the background area and the color of the character area and the mutual relationship between the color of the background area and the registered color are obtained, and the mutual relationships are identified. It is possible to perform the identification process by using the index and remove the influence of the color mixture of the character color of the identification character and the background color to perform the identification.

Next, the processing procedure in the color registration mode of the color identification apparatus 10 shown in FIG. 1 will be described.

FIG. 5 is a flowchart showing a processing procedure in the color registration mode of the color identification apparatus 10 shown in FIG.

As shown in FIG. 5, this color identification device 10
Is a color image (RGB) including the registered color,
After importing from 1 (step 501), the area including the color to be registered is designated (step 502).

Next, a normalization process is performed to convert each pixel in the designated area into (r, g) using the calculation formula of r = R / (R + G + B) g = G / (R + G + B) (step 503). A two-dimensional histogram H (r, g) of (r, g) of all pixels in the designated area is obtained (step 504).

The two-dimensional histogram H (r,
The maximum frequency (r, g) of g) is registered in the registered color storage unit 19 as (rt, gt) (step 505).

By repeating this registration process, it is possible to register a plurality of colors to be compared with the character colors of the identification characters.

Next, the processing procedure in the identification mode of the color identification apparatus 10 shown in FIG. 1 will be described.

FIG. 6 is a flow chart showing a processing procedure in the identification mode of the color identification apparatus 10 shown in FIG.

As shown in FIG. 6, this color identification device 10
When a color image (RGB) including identification characters is captured from the image input unit 11 (step 601), the cutout processing unit 12 cuts out a character template region by template matching (step 602), and the character background separation unit 13 This character template area is converted from an RGB image into a density image, binarized, separated into a character area and a background area, and output to the identification processing unit 14 (step 60).
3).

Then, the identification processing section 14 which has received the character area and the background area uses the normalization processing section 15 to calculate each of the background areas by the calculation formula of r = R / (R + G + B) g = G / (R + G + B). Normalize the pixel (r,
g) is created (step 604), and the histogram creating unit 16 is used to create (r,
After obtaining the two-dimensional histogram H (r, g) of (g) (step 605), the maximum frequency (r, g) of the histogram H (r, g) is adopted as (rb, gb) (step 606). ).

Next, the identification processing unit 14 similarly normalizes each pixel of the character area to create (r, g) (step 607), and (r, g) of all the pixels in the character area. Of 2
After obtaining the dimension histogram H (r, g) (step 608), the maximum frequency (r, g) of the histogram H (r, g) is adopted as (rc, gc) (step 609).

After that, the identification processing unit 14 centers the background position (rb, gb) on the color (rc, g) of the character area.
c) angle θc and the registered color (rt,
gt) and the angle θt are calculated using the angle calculation unit 17 (step 610), and if the absolute value of the difference between θc and θt is within a certain value, the color of the character area is identified as the registered color. (Step 611).

By performing the series of processes described above, it is possible to identify the character color of the identification character in consideration of the background color.

Here, for convenience of explanation, the case of identifying whether or not one registered color and the character color of the identification character match is shown, but in reality, a plurality of registered colors are prepared in advance. In addition, it is also possible to use which of the registered colors the color of the identification character is close to as an identification index.

For example, as shown in FIG. 7, three registered colors t
1 (rt1, gt1), t2 (rt2, gt2) and t
3 (rt3, gt3) and background color b exist,
The angle formed by the background color b and the registered colors t1, t2, and t3 is θt1 = tan-1 ((gt1-gb) / (rt1-r), respectively.
b)) θt2 = tan-1 ((gt2-gb) / (rt2-r
b)) θt3 = tan-1 ((gt3-gb) / (rt3-r
b)). Here, as is apparent from the figure, since the angle formed by the character color c and the background color b is the closest to the registered color t2, such character color is regarded as the registered color t2.

As described above, in this embodiment,
The identification processing unit 14 uses the normalization processing unit 15 to normalize the colors of the background area and the character area in the RGB space, respectively, and then uses the histogram creation unit 16 to normalize the two colors.
A two-dimensional histogram corresponding to a coordinate (rb, gb) and a character area having a maximum frequency of the two-dimensional histogram corresponding to the background area is created by creating a two-dimensional histogram with the components (r, g) as the vertical axis and the horizontal axis, respectively. The coordinates (rc, gc) having the maximum frequency of are the color of the background area and the color of the character area, respectively, and the angle calculating unit 17 is used to determine the angle θc between the color of the background area and the color of the character area, The angles θt of registered colors are calculated, and if these angles are within a predetermined range, it is configured to identify that the color of the character area and the registered color are the same. Therefore, the following effects can be obtained.

1) When the color of the character area included in the color image is mixed with the color of the background area, the color of the character area can be accurately identified while removing the influence of the color of the background area.

2) The angle θc of the color of the background area and the color of the character area and the angle θt of the color of the background area and the registered color can be efficiently calculated.

In the present embodiment, the case has been described in which the character color of the character included in the color image is used as the identification target, but the present invention is not limited to this and can be applied to various identification targets. You can

[0064]

As described above in detail, in the first invention, the color of the target area and the first angle formed by the color of the background area located in the background of the target area on the color space are registered in advance. Since the second angle formed by the registered color and the color of the background area in the color space is calculated, the color of the target area is identified based on the calculated difference between the first angle and the second angle. ,
When the color of the target area included in the color image is mixed with the color of the background area, the color of the target area can be accurately identified while removing the influence of the color of the background area.

In the second aspect of the invention, the target area and the background area are separated from the color image, and the color of each pixel of the separated target area and background area is converted into a unit color that does not include a lightness component. The unit color having the highest appearance frequency among the unit colors of each pixel of the converted target area is selected as the color of the target area, and the unit color having the highest appearance frequency of the unit colors of each pixel of the background area is the background. Select as the color of the area. Then, the first angle formed by the color of the selected target area and the color of the background area in the color space and the second angle formed by the color of the selected background area and the registered color in the color space are calculated and calculated. Since the color of the target area is identified based on the difference between the first angle and the second angle, it is possible to efficiently calculate the first and second angles.

[Brief description of drawings]

FIG. 1 is a diagram showing a configuration of a color identification device used in the present embodiment.

FIG. 2 is a diagram showing an example of a character template region cut out from image data by a cutout processing unit shown in FIG.

FIG. 3 is a diagram showing a concept of processing performed by a normalization processing unit shown in FIG. 1.

FIG. 4 is a diagram showing a concept of processing of an angle calculation unit shown in FIG. 1.

5 is a flowchart showing a processing procedure in a color registration mode of the color identification device shown in FIG.

6 is a flowchart showing a processing procedure in an identification mode of the color identification apparatus shown in FIG.

FIG. 7 is an explanatory diagram for selecting a character color of an identification character from a plurality of registered colors.

[Explanation of symbols]

10 ... Color identification device, 11 ... Image input unit, 12 ... Clipping processing unit, 13 ... Character background separation unit, 14 ... Identification processing unit, 15 ... Normalization processing unit, 16 ... Histogram creation unit, 17 ... Angle calculation unit, 18 ... Color registration processing unit, 19
... Registered color storage section, 20 ... Character template area, 20
a ... Character area, 20b ... Background area

Continuation of the front page (56) Reference JP-A-8-84962 (JP, A) JP-A-5-72040 (JP, A) JP-A-5-133809 (JP, A) JP-A-5-61974 (JP , A) (58) Fields investigated (Int.Cl. ⁷ , DB name) G01J 3/00-3/52 JISST file (JOIS)

Claims (2)

(57) [Claims] 1. A color identification device for identifying a color by registering a predetermined color as a registered color in advance and comparing the color of a target area existing in a color image with the registered color, the color of the target area and the target area. Detecting means for detecting the color of the background area located on the background of the background, the first angle formed by the color of the target area detected by the detecting means and the color of the background area on the color space, the registered color and the detecting means. The color of the background area detected in step 2 in the color space
Angle calculating means for calculating the angle of, and the first angle and the second angle calculated by the angle calculating means.
And a color identification unit that identifies whether the color of the target area corresponds to the registered color based on the angle difference from the angle. 2. The detecting means includes a separating means for separating a target area and a background area from the color image, and a unit which does not include a lightness component in a color of each pixel of the target area and the background area separated by the separating means. A conversion unit that converts each color, and a unit color that has the highest appearance frequency among the unit colors of each pixel of the target region converted by the conversion unit is selected as the color of the target region, and the background converted by the conversion unit The color identification device according to claim 1, further comprising a color selection unit that selects a unit color having the highest appearance frequency among the unit colors of each pixel of the area as a color of the background area.