JPH09185711A - Shape recognizing method and its device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To recognize the shape of an object by clearly separating a background part and a shadow part. SOLUTION: This device places the object 2 in the background of a uniform reflection rate, arranges a light source 3 to illuminate the object 2 so as not to generate a shadow at the object 2 itself and recognizes the shape of the object 2 from the image of the reflected light. In this case, the reflected light is collor-resolved to not less than two of different wave-length areas to pick up the image of each resolved color and two kinds of color-resolved images in the color-resolved image are specified to recognize the shape of the object based on the ratio of the two kinds of specified color resolving signals.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method and apparatus for accurately recognizing the shape of an object whose surface diffuses light, such as a natural object such as food or wood.

[0002]

2. Description of the Related Art As a method for optically recognizing various objects, various methods as shown in the following (1) to (4) have been developed. A method of illuminating an object with transmitted illumination (backlight illumination), binarizing a grayscale image captured by a monochrome (black and white) camera, and extracting a region of the object (“video information”).
Page 37-43 of the July 1991 issue). …
In this case, the shadow of the object is prevented from being formed by the transmission diffusion illumination. A method of illuminating an object with reflected illumination (brightfield illumination or uniform illumination), binarizing the grayscale image captured by a monochrome camera, and extracting the area of the object (“video information”).
Page 37-43 of the July 1991 issue). …
In this case, the shadow of the object is prevented by illuminating the coaxial epi-illumination or multiple light sources from multiple directions. An image of the shadow area of the recognition object generated by the illumination from each light source is extracted from the images obtained by sequentially illuminating with an illuminator consisting of multiple light sources, and the logical sum of the images of each shadow area is extracted. A method of recognizing the shape of a recognition target by calculation (special opening / closing 6-307833). ... In this case, the difference between the grayscale image level of the shadow boundary with the object is used. In order to detect "red seaweed" that can agree with the inspector's judgment, the reflectance of the wavelengths of 479 nm and 630 nm is measured, the ratio of the reflectance of these two wavelengths is calculated, and this is compared with the evaluation reference value. Then, the method to evaluate the quality of "Akame nori" mentioned above (special opening and closing No. 7-27698) As a method to distinguish the color of glass cullet (scrap glass), two types of specific light with different wavelengths are detected and their intensity is detected. How to identify the color of the cullet from the difference
(025)

[0003]

In the case of the invention, since the transmitted light is detected, it is necessary to provide a light transmitting body between the object and the illumination light source, and the object is continuously conveyed in the manufacturing process. In the case of automatically determining, the above-mentioned light transmissive body is contaminated with time, the transmittance changes, and the amount of received light changes, so that there is a drawback that stable measurement cannot be performed. Further, in the case of automation, since the target object is usually conveyed by a belt conveyor or the like, it is not realistic to image the target object using a transparent body. In the case of 1, the grayscale image level of the target object fluctuates due to fluctuations in the light amount of the illumination light source and uneven lighting, and it is not possible to stably extract the target object region. In this case, it is necessary to repeat the illumination-imaging many times, and it is impossible to accurately recognize the object being conveyed at high speed. In addition, a large number of light sources are required, resulting in high cost. In this case, the quality of the object is only determined by the color, and the shape of the object is not extracted by excluding the shadow or background of the object. in the case of,
Similarly to, it is not possible to extract the shape of the object by excluding the shadow and background of the object. Therefore, the object of the present invention is to provide a high-speed processing for an object that is being automatically conveyed without providing a light-transmitting body, without being affected by fluctuations in the light amount of the illumination light source and uneven illumination. A method and apparatus for recognizing an object, which is capable of accurately extracting only the shape of the object itself by removing the shadow and background of the object.

[0004]

In order to achieve the above object, the present invention in the aspect of the method is to place an object in the background and arrange a light source so that the object itself is not shaded. In the method of illuminating an object and recognizing the shape of the object from the image of the reflected light, it shows high reflectance in the wavelength bands of the two kinds of light to be used, and the ratio is almost 1
An object to be recognized is placed in the background, and the reflected light is color-separated into two or more different wavelength regions, and an image for each separated color is captured. A shape recognition method characterized in that a color separation image is specified, and the shape of the object is recognized based on a ratio of the specified two kinds of color separation image signals. Further, in the present invention in terms of the device surface, the object is placed in the background, the light source is arranged so that the object itself does not have a shadow, and the object is illuminated. In a device for recognizing a shape, a recognition target object is placed in the background where the reflectance is high in the wavelength bands of two kinds of light to be used and the ratio is approximately 1, and the reflected light is different in wavelength range. An image input unit that performs color separation as described above and captures an image for each separated color, and the two types of color separated images in the color separated images obtained by the image input unit are specified and specified. A separation ratio calculating means for calculating a ratio of the two types of color separation image signals,
The shape recognition device is characterized by recognizing the shape of the object by comparing the value of the ratio obtained by the decomposition ratio calculation means with a predetermined value.

[0005]

BEST MODE FOR CARRYING OUT THE INVENTION Embodiments of the present invention will be described below with reference to the accompanying drawings to provide an understanding of the present invention. The following embodiments are examples of embodying the present invention, and are not of the nature to limit the technical scope of the present invention. Here, FIG. 1 is a block diagram showing an outline of a shape recognition apparatus according to an embodiment of the present invention, FIG. 2 is a graph showing that the reflectance for each frequency differs depending on the color of the object, and FIG. 4 is a graph showing that there is no difference in reflectance for each object, FIG. 4 is an image taken of an object (chikuwa), FIG. 5 is a binarized image obtained by the conventional method, and FIG. 6 is obtained by the method of the present invention. It is a captured image. First, an outline of a shape recognition device according to an embodiment of the present invention will be described with reference to FIG. In the figure, an illumination light source 3 irradiates an object (“chikuwa” in this case) 2 in a background 1 in which the spectral reflectance of white or the like is uniform from 400 nm to 700 nm in the visible light region. The object 2 and the background portion in the vicinity thereof are imaged by the two-dimensional imaging device 4. The image pickup device 4 is a three-plate type color television camera, which is R [red], G [green], and B.
An image is obtained by decomposing the three components of [blue]. The output from the image pickup device 4 is transmitted as a video signal to an analog / digital converter (A / D converter) 5, A / D converted, and then image-processed by an image processing device 6 as described later. .

In the image processing device 6, two types of color separation images (in this example, R [red] and B [color red]) are selected from the captured color separation images.
A binarization process for calculating the ratio of [blue]) as a feature amount (α), comparing it with a predetermined value to identify the shape of the object 2, and distinguishing the area of the object 2 from other areas. Etc. and outputs the result to the operation result output device 7. Also,
The result is also output to the image monitor 8 to display the image of the extracted area of the object 2. In this embodiment, the object 2
Since the shape of the object 2 itself is extracted by removing the background 1 part and the shadow part from the image of the reflected light from the object 2
It is conceivable that the extraction accuracy will decrease if a shadow of itself occurs inside the. Therefore, the illumination light source 3 needs to irradiate light from a direction in which a shadow does not appear inside the object. However, such an internal shadow is unlikely to occur unless the object is extremely uneven, so that it is sufficient to arrange one or two illumination light sources 3 for an object such as ordinary food. In this embodiment, a set of two fluorescent lamps is used. The contents of the image processing in the image processing device 6 will be described in detail. In this embodiment, the object to be recognized is placed in the background where the reflectance is high in the wavelength bands of the two kinds of light to be used and the ratio thereof is approximately 1. A white plate is used as an example of the background as described above. On the other hand, the object "Chikuwa" 2 has a brown surface, and changes from a white-yellow color close to the dough to a light brown, brown or dark brown depending on the state of baking. doing.

2A, 2B, and 2C are three colors of white yellow (a), light brown (b), and dark brown (c), respectively, according to the baking condition of the object. The spectral reflectance distribution is shown. The wavelength range is 400 nm to 700 nm. Speaking of the color components of the image pickup device 4 in FIG. 1, 400 to 500 nm is blue [B], 500 to 600 nm is green [G], and 600 to 7.
00 nm almost corresponds to red [R]. Figure 2 of object 2
Since the spectral reflectance distributions change depending on the baking conditions of (a), (b), and (c) shown in (3), the values of the three color component regions [R, G, B] are within the object 2. You can see that it depends on the place. For example, red [R] corresponding to dark brown
It can be seen that in the region, the absolute value is large and the change is small, and in the blue [B] region, it is small and the change is large. On the other hand, FIG. 3 shows the spectral reflectance distribution of the white plate used as the background 1. The spectral reflectance distribution of the white plate in the background is almost uniform, and the three color component regions [R,
It can be seen that the values of G, B] are almost the same. In addition, although the shadow level around the object changes the light level,
The spectral reflectance distribution of the shadow is determined by the illumination source and the spectral characteristics of the white plate, and its distribution is uniform regardless of frequency, similar to the spectral reflectance distribution of the white plate.

From the above, three color components [R, G,
Considering a case where two types of color components are specified from B] and the ratio of the two signals is used as the feature amount (α) (in this embodiment, in the case of a combination of red [R] and blue [B]). Determines that the difference between the background and the shadow is the largest, and the ratio R /
B is the feature value α), ・ R and B are almost equal in the background part illuminated, α = 1 ・ R and B are almost equal in the shadow part not illuminated, α = 1 ・ In the target part, the values of R and B change depending on the baking condition, and the relationship α ≠ 1 holds. In fact, depending on the baking condition of the target object 2, it was found that α is 1.2 when the color is white yellow, which is close to that of the dough, and α = 1.2 to 2.8 when the color is light brown to dark brown. From this, the object "Chiwa" of this embodiment
In the case of, by extracting the region in which the range of the feature amount (α) is α = 1.2 to 2.8 as the region of the object,
The object 2 can be identified. In the present embodiment, from the viewpoint of separating the object 2 from the background and the shadow, in order to simplify the image calculation processing, 1.1 is set as the threshold value, and points α ≧ 1.1 are recognized as the object. . The results are shown in FIGS.

FIG. 4 shows the result of imaging the object "chikuwa" with the imaging optical system shown in FIG. From FIG. 4, it can be seen that a shadow is generated around the object. FIG. 5 shows a result of binarizing the image data of FIG. 4 with an appropriate threshold value as an example of a conventional technique. Accurate separation of the object from Fig. 5
You can see that it has not been extracted. FIG. 6 shows the result of calculating the feature amount (α) by the method of the present embodiment and extracting the region of the object using α ≧ 1.1 as the evaluation criterion. From FIG. 6, it can be seen that the object 2 can be accurately separated and extracted by the method of this embodiment.

[0010]

[Embodiment] Although a white plate is used as the background 1 in the above embodiment, any color can be selected as long as it is a background color in which the background and the shadow and the object can be separated by two kinds of color component ratios (= feature values). You can Further, in the present invention, if there is a portion of the spectral reflectance distribution that is the same as the background or the shadow in the object, it is excluded from the object and cannot be selected as the object. In that respect, the mirror surface portion of the mechanical component or the plastic component is totally reflected and thus is imaged as the same portion as the white portion, which is unsuitable as an object. The above problem does not occur because it is intended for items such as bread, paste products (chikuwa, kamaboko, etc.), vegetables, seafood, and other natural foods that have light diffusely reflected on the surface.

[0011]

As described above, the shape recognition method according to the present invention illuminates an object by placing the object in the background and arranging a light source so that the object itself does not have a shadow. , In the method of recognizing the shape of the object from the image of the reflected light, the object to be recognized is shown in the background where the reflectance is high in the wavelength bands of the two kinds of light to be used and the ratio is about 1. Then, the reflected light is color-separated into two or more different wavelength regions, and an image for each separated color is captured.
Specify the two types of color separation images in the color separation image,
The shape recognition method is characterized by recognizing the shape of the object on the basis of the ratio of the specified two types of color-separated image signals. Further, the shape recognition apparatus according to the present invention has the above background An object is placed, a light source is arranged so that a shadow does not appear on the object itself, the object is illuminated, and the shape of the object is recognized from an image of the reflected light. The object to be recognized is placed in the background where the reflectance is high in the wavelength bands of two kinds of light and the ratio is almost 1, and the reflected light is color-separated into two or more different wavelength ranges, and each separated color Image input means for picking up an image for each image, and the two kinds of color separation images in the color separation images obtained by the image input means are specified, and the ratio of the specified two kinds of color separation image signals is calculated. And a decomposition ratio calculation means for
The shape recognition device is configured to recognize the shape of the object by comparing the value of the ratio obtained by the decomposition ratio calculation means with a predetermined value. Therefore, since it is not necessary to interpose a transparent body, there is no problem regarding the stain and it can be applied to automatic inspection such as automation of the process. Further, since the ratio of the two types of color separation image signals is obtained, there is no influence due to fluctuations in the light amount of the light source, uneven lighting, and variations due to the shading level of the shading. It is suitable for application to shape recognition of non-standard foods (sausage, chikuwa, bread, etc.) that easily produce shadows on objects even with reflected illumination (bright field illumination or uniform illumination).

[Brief description of the drawings]

FIG. 1 is a block diagram showing an outline of a shape recognition device according to an embodiment of the present invention.

FIG. 2 is a graph showing that the reflectance for each frequency varies depending on the color of the object.

FIG. 3 is a graph showing that there is no difference in reflectance for each frequency with respect to the background.

FIG. 4 is an image taken of an object (chikuwa).

FIG. 5 is a binarized image obtained by a conventional method.

FIG. 6 is a perspective view showing an image pickup image obtained by the method of the present invention, showing a fitting jig according to an embodiment of the present invention.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 ... Background 2 ... Object 3 ... Illumination light source 4 ... Imaging device 5 ... A / D conversion device 6 ... Image processing device 7 ... Calculation result output device 8 ... Image monitor

─────────────────────────────────────────────────── ───

[Procedure amendment]

[Submission date] March 28, 1996

[Procedure amendment 1]

[Document name to be amended] Statement

[Correction target item name] 0005

[Correction method] Change

[Correction contents]

[0005]

BEST MODE FOR CARRYING OUT THE INVENTION Embodiments of the present invention will be described below with reference to the accompanying drawings to provide an understanding of the present invention. The following embodiments are examples of embodying the present invention, and are not of the nature to limit the technical scope of the present invention. Here, FIG. 1 is a block diagram showing an outline of a shape recognition apparatus according to an embodiment of the present invention, FIG. 2 is a graph showing that the reflectance for each frequency differs depending on the color of the object, and FIG. Graph showing that there is no difference in reflectance, Fig. 4 is a halftone image displayed on the display of the object (chikuwa), and Fig. 5 is displayed on the display of "chikuwa" obtained by the conventional method. 6 is a halftone image displayed on the display of "CHIKUWA" obtained by the method of the present invention. First, an outline of a shape recognition device according to an embodiment of the present invention will be described with reference to FIG. In the figure, an illumination light source 3 is provided on an object (in this case, "chikuwa") 2 in a background 1 in which the spectral reflectance of white or the like is uniform from 400 nm to 700 nm in the visible light region.
Light is emitted from. The object 2 and the background portion in the vicinity thereof are imaged by the two-dimensional imaging device 4. The image pickup device 4 is a three-plate type color television camera, which is R [red], G [green], and B.
An image is obtained by decomposing the three components of [blue]. The output from the image pickup device 4 is transmitted as a video signal to an analog / digital converter (A / D converter) 5, A / D converted, and then image-processed by an image processing device 6 as described later. .

[Procedure amendment 2]

[Document name to be amended] Statement

[Correction target item name] Brief description of drawings

[Correction method] Change

[Correction contents]

[Brief description of the drawings]

FIG. 1 is a block diagram showing an outline of a shape recognition device according to an embodiment of the present invention.

FIG. 2 is a graph showing that the reflectance for each frequency varies depending on the color of the object.

FIG. 3 is a graph showing that there is no difference in reflectance for each frequency with respect to the background.

FIG. 4 is a halftone image displayed on a display of an object (chikuwa).

FIG. 5 is a halftone image displayed on a display of “CHIKUWA” obtained by a conventional method.

FIG. 6 is a halftone image displayed on the display of “CHIKUWA” obtained by the method of the present invention.

[Explanation of Codes] 1 ... Background 2 ... Object 3 ... Illumination light source 4 ... Imaging device 5 ... A / D conversion device 6 ... Image processing device 7 ... Calculation result output device 8 ... Image monitor

[Procedure 3]

[Document name to be amended] Drawing

[Correction target item name] Fig. 4

[Correction method] Change

[Correction contents]

FIG. 4

[Procedure amendment 4]

[Document name to be amended] Drawing

[Correction target item name] Fig. 5

[Correction method] Change

[Correction contents]

[Figure 5]

[Procedure Amendment 5]

[Document name to be amended] Drawing

[Correction target item name] Fig. 6

[Correction method] Change

[Correction contents]

FIG. 6

Claims (2)

[Claims] 1. An object is placed in the background, a light source is arranged so that the object itself does not have a shadow, the object is illuminated, and the shape of the object is recognized from an image of the reflected light. Method, the object to be recognized is placed in the background where the reflectance is high in the wavelength bands of the two kinds of light to be used and the ratio is approximately 1, and the reflected light is colored into two or more different wavelength bands. An image for each separated color is captured, the two types of color separated images in the color separated image are specified, and the above-described two types of color separated image signals are identified based on the ratio of the specified two types of color separated image signals. A shape recognition method characterized by recognizing the shape of an object. 2. An object is placed in the background, a light source is arranged so that the object itself does not have a shadow, the object is illuminated, and the shape of the object is recognized from the image of the reflected light. In this device, the object to be recognized has a high reflectance in the wavelength bands of the two kinds of light to be used, and the object to be recognized is placed in the background where the ratio is approximately 1, and the reflected light is colored into two or more different wavelength bands. An image input unit that separates and captures an image for each separated color, and the two types of color separated images in the color separated images obtained by the image input unit are specified, and the two specified types are specified. A separation ratio calculating means for calculating a ratio of the color separation image signals, and recognizing the shape of the object by comparing the value of the ratio obtained by the separation ratio calculating means with a predetermined value. Shape recognition device.