JP2004077501A - Color classification device - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a color classification device that is simple in device structure, low in cost and durable to mechanical vibration, and does not limit light sources and is satisfactory color-classifiable even if their spectrum is changed. <P>SOLUTION: The color classification device is provided with an imaging means for imaging light from an object, optical means for focusing the light from the object on the imaging means, a plurality of band-pass filters positioned between the object and the imaging means to transmit respective different wavelength bands of light, a classifying means for calculating a classification spectrum for classification using a statistical technique, from spectra through the object imaged by the imaging means, and classifying the object with the classification spectrum, and an exposure condition controlling means for controlling exposure conditions at imaging through each of the plurality of band-pass filters. The color classification device is characterized in that the exposure condition controlling means include an exposure value limiting means for limiting an exposure value in dependence on the plurality of band-pass filters. <P>COPYRIGHT: (C)2004,JPO

Description

{Circle around (1)} The present invention relates to a color classification device that classifies objects mainly using colors.

Conventionally, a color identification device for identifying the color of an object has been used in the management of paint color and dyeing degree at a production site of each industry, or in the color measurement of a product, or in the color measurement of a subject in medical and academic fields. It's being used.

In the conventional technique disclosed in Patent Document 1, two classes are classified by performing statistical processing on the reflection spectrum of an object.

Specifically, the reflection spectrum of an object whose class is known is statistically processed using Foley Sammon transform (FS transform) (see Non-Patent Document 1).

The FS conversion is a method of classifying two classes, specifically, a spectrum di for classification that maximizes a Fisher ratio (Fisher （ratio) R (di) obtained by the following equation (1). That is.

^{R (di) = (di t} S1di) / (di t S2di) ... (1) where, di ... classification spectrum di ^t ... classification spectrum (transposition) S1 ... interclass covariance matrix S2 ... intraclass covariance matrix after The spectrum di for this classification is called a classification spectrum.

分類 Since this classification spectrum di has the same number of dimensions as the spectrum of the object, it should be accurately expressed as di (λ), but is simply expressed as di for simplicity.

{Circle around (2)} Then, two kinds of classification spectra for increasing the Fisher ratio are obtained.

The classification spectrum di that maximizes the Fisher ratio is d1, and the classified spectrum di that maximizes the Fisher ratio is d2 among the spectra orthogonal to d1.

、 ２ Two classes are classified by projecting each data into the space constituted by the classification spectra d1 and d2.

分類 The classification spectra d1 and d2 are obtained from the following equation (2).

d1 = α1 S2 ⁻¹ Δ d2 = α2 S2 ⁻¹ [I− ( ^Δt S2 ⁻² Δ) / ( ^Δt S2 ⁻³ Δ) S2 ⁻¹ ] Δ (2) where α1 and α2 are normal Is a conversion coefficient, Δ is X1−X2 (difference spectrum between class 1 and class 2), and I is a unit matrix.

投影 In order to project each data into the space constituted by the classification spectra d1 and d2 obtained in this way, an inner product of the classification spectrum and the reflection spectrum of the object is obtained. If the reflection spectrum of the object is f (λ) (λ = wavelength), the inner products t1 and t2 can be expressed by the following equations.

t1 = f (λ) · d1
t2 = f (λ) · d2 (3) Here, represents an inner product operation.

In the technique disclosed in Patent Document 1, a classification boundary is determined from the values of t1 and t2 as shown in FIG. 40, and a filter having the characteristics of this classification spectrum is realized using a diffraction grating 501 and a liquid crystal 502 as shown in FIG. are doing.
JP-A-3-267726 Q. Tian, M .; Barbaro et al., "Image Classification by the Foley-Sammon Transform", Optical Engineering, Vol. 25, no. 7, 1986

By the way, the classification spectra d1 and d2 generally have complicated shapes as shown in FIG. 42, and take positive and negative values, so that the mounting accuracy of a diffraction grating, a liquid crystal filter and the like is strictly required.

Further, since the light source is limited in advance in the device disclosed in the above publication, it is not suitable for classification of different light sources, and when the spectrum of the light source changes, good classification cannot be performed. There is also a disadvantage that is high.

The present invention has been made in view of the above points, and has a simple device configuration, is low in cost, can withstand mechanical vibration, and can be used even when its spectrum changes without limiting the light source. It is an object of the present invention to provide a color classification device capable of performing good color classification.

According to the present invention, in order to solve the above problems,
Imaging means for imaging light from the object;
Optical means for forming an image of light from the object on the imaging means,
A plurality of bandpass filters that pass bands of different light wavelengths disposed between the object and the imaging unit,
Classification means for calculating a classification spectrum for classification using a statistical method from the spectral spectrum of the object imaged by the imaging means, and classifying the object using this classification spectrum,
Exposure condition control means for controlling exposure conditions at the time of imaging of each of the plurality of band-pass filters,
With
A color classification device is provided, wherein the exposure condition control means includes an exposure amount limiting means for limiting an exposure amount corresponding to the plurality of bandpass filters.

According to the present invention, in order to solve the above problems,
Imaging means for imaging light from the object;
Optical means for forming an image of light from the object on the imaging means,
A plurality of bandpass filters that pass bands of different light wavelengths disposed between the object and the imaging unit,
Classification means for calculating a classification spectrum for classification using a statistical method from the spectral spectrum of the object imaged by the imaging means, and classifying the object using this classification spectrum,
Exposure condition control means for controlling exposure conditions at the time of imaging of each of the plurality of band-pass filters,
With
A color classification apparatus is provided, wherein the exposure condition control means includes exposure condition storage means for storing exposure conditions at the time of imaging of each of the plurality of bandpass filters.

According to the present invention, in order to solve the above problems,
Imaging means for imaging light from the object;
Optical means for forming an image of light from the object on the imaging means,
A band-variable bandpass filter that is arranged between the object and the imaging unit and that can change a wavelength band of light to be passed,
Classification means for calculating a classification spectrum for classification using a statistical method from the spectral spectrum of the object imaged by the imaging means, and classifying the object using this classification spectrum,
With
An exposure condition control means for separately controlling exposure conditions at the time of imaging of pass band characteristics of different wavelengths set in the band variable band-pass filter is provided.

According to the present invention, in order to solve the above problems,
Imaging means for imaging light from the object;
Optical means for forming an image of light from the object on the imaging means,
A plurality of bandpass filters that pass bands of different light wavelengths disposed between the object and the imaging unit,
Classification means for calculating a classification spectrum for classification using a statistical method from the spectral spectrum of the object imaged by the imaging means, and classifying the object using this classification spectrum,
Normalizing means for normalizing at least one of the spectral characteristic and the luminance of the light source illuminating the object,
With
A color classification apparatus is provided, wherein the normalizing means includes a light source observing means for obtaining data of the light source.

According to the present invention, in order to solve the above problems,
Imaging means for imaging reflected light of the object;
Optical means for imaging the reflected light of the object on the imaging means,
A band-variable bandpass filter that is arranged between the object and the imaging unit and that can change a wavelength band of light to be passed,
Classification means for calculating a classification spectrum for classification using a statistical method from the spectral spectrum of the object imaged by the imaging means, and classifying the object using this classification spectrum,
Normalizing means for normalizing at least one of the spectral characteristic and the luminance of the light source illuminating the object,
And
A color classification apparatus is provided, wherein the normalizing means includes a light source observing means for obtaining data of the light source.

According to the present invention, in order to solve the above problems,
Imaging means for capturing an image composed of light from an object,
A band-variable band-pass filter arranged between the object and the imaging unit, capable of changing a pass band characteristic of a wavelength,
In a plurality of classes defined for the classification of the object, using the spectral spectra of a plurality of types of first objects, each of which is a known object, each of the classes is a classification spectrum for the classification. Classification spectrum calculating means for calculating
Means for defining a space for performing the classification of the object in each of the classes using the classification spectrum,
The characteristic of the band variable bandpass filter is changed a plurality of times, and the corresponding class is obtained from a plurality of images obtained by imaging the second object, which is the unknown object, in a plurality of bands. Means for projecting the spectral data of the object into the space;
A classification unit configured to classify the second object in each of the classes based on the result of the projection.

According to the present invention, in order to solve the above problems,
Imaging means for capturing an image composed of light from an object,
A plurality of bandpass filters, each having a different passband characteristic of wavelength, disposed between the object and the imaging unit,
In at least three classes defined for the classification of the object, each spectral spectrum of two first objects, which are the objects that are known to correspond to any two classes of the classes, respectively. Using, a classification spectrum calculation means for calculating a classification spectrum for the classification,
Means for defining a space for performing the classification of the object in each of the classes using the classification spectrum,
Using each of the bandpass filters, the spectral spectrum of the second object obtained from a plurality of images obtained by imaging the second object whose class is the unknown object. Means for projecting the data of
A classification unit configured to classify the second object in each of the classes based on the result of the projection.

According to the present invention, a plurality of bandpass filters each having a different band are prepared, and each of the plurality of bandpass filters is arranged between the object and the imaging unit. By calculating a classification spectrum for classification using a statistical method from the spectral spectrum of the object imaged by the means, and performing the classification of the object using this classification spectrum, the apparatus configuration is simple and low. It is possible to provide a color classification device that is cost-effective, can withstand mechanical vibrations, and can satisfactorily perform color classification even when its spectrum changes without limiting the light source.

前 Before describing the embodiments of the present invention, the basic principle of the present invention will be described.

According to the present invention, a color classification device having a simple and inexpensive configuration is realized by using a plurality of band-pass filters that transmit only a specific wavelength as a filter for classification. In addition, in order to perform color classification even under different light sources, the reflection spectrum of an appropriate reference plate is measured under the same conditions as when the object is photographed, and the reflection spectrum of the object is measured. In this case, the influence of the light source (illumination light) is removed.

That is, with λ as the wavelength, the reflection spectrum of the object is f (λ), the reflection spectrum of the reference plate is s (λ), the reflection spectrum of the illumination light is L (λ), and the sensitivity spectrum of the photographing system (photographing If the transmission spectrum of the lens, the sensitivity spectrum of the image sensor, etc.) are M (λ), the imaging spectrum gi (λ) of the object and the imaging spectrum gs (λ) of the reference plate are respectively gi (λ) = f (λ). ) × L (λ) × M (λ) {gs (λ) = s (λ) × L (λ) × M (λ)}, and the spectrum gi ′ (λ) of the object is expressed as {gi ′ (λ) = gi (λ) / gs (λ) = f (λ) / s (λ) (4)

Thus, the influence of the reflection spectrum L (λ) of the illumination light can be removed, and if gi ′ (λ) is used, classification can be performed under different light sources.

If the luminance of the illumination light is further different, the power of the signal gi ′ (λ) after division may be normalized.

Hereinafter, embodiments of the present invention will be described with reference to the drawings.

(First Embodiment) {Depending on the reflection spectral characteristics of an object to be measured, the intensity of reflected light is low in a filter of a certain band, and the SNR of observed data is deteriorated, so that the classification accuracy is lowered. is there.

A first embodiment for solving this problem will be described with reference to FIG.

The color classification device according to the present embodiment includes an optical system 10 including a lens and the like provided in front of a housing 1 as shown in FIG. 1, an aperture 101, an aperture control circuit 126, and a plurality of sheets as shown in FIG. .., 12E having different band characteristics, a filter position sensor 123, a motor 24, a motor drive circuit 124, and an image sensor for capturing images of an object and a reference plate. 14, an amplifier 15, an image sensor driving circuit 22, an A / D converter 16, a frame memory 18, a classification operation circuit 28, an exposure value memory 29 for storing an aperture value for obtaining an appropriate exposure for each bandpass filter, and a control circuit 26. Consists of

As shown in FIG. 2, the rotary color filter 12 is provided with filter position detection holes 125A, 125B,..., 125E and a filter initial position detection hole 126.

The position of each bandpass filter 12A, 12B,..., 12E is detected by the initial position detecting hole 126 and the filter position detecting holes 125A, 125B,. Then, the motor drive circuit 124 is controlled so that the rotation of the filter is synchronized with the imaging of the image sensor 14 by the signal from the filter position sensor 123.

The image that has passed through the rotary color filter 12 is formed on an image sensor 14, converted into a digital signal by an A / D converter 16 via an amplifier 15, stored in a frame memory 18, and transferred from the frame memory 18 to a monitor. A signal and data of a predetermined area in the image are sent to the classification operation circuit 28.

Now, suppose that the value of the data of the bandpass filter 12A is about 1/10 of the value of the data of the bandpass filter 12B as the data of the predetermined area in the image. Considering the signal-to-noise ratio (SNR) of the measurement data due to dark current noise, the data of the bandpass filter 12A has an SNR about 10 times worse than the data of the bandpass filter 12B.

Therefore, the aperture is controlled in synchronization with each of the filters 12A to 12E so that the SNR of each filter exceeds a predetermined level.

(4) The aperture values of the respective filters 12A to 12E at that time are stored in the exposure value memory 29, and the aperture is corrected by preprocessing of the classification operation circuit 28.

{That is, when taking in the data of the band-pass filter 12A, the aperture area of the stop is set to ten times that of the band-pass filter 12B.

Then, at the time of the classification operation, the data of the band-pass filter 12A is reduced to 1/10 before the operation.

Next, an example of a series of operations in the color classification device according to the first embodiment configured as described above will be described with reference to the flowchart shown in FIG.

First, a measurement area of an object is set (step S1).

Next, the first band-pass filter is set (band-pass filter 12A) to perform pre-exposure, and the aperture value is set so that the measurement data obtained from the pre-exposure falls within a predetermined value range. The aperture value is stored together with the corresponding bandpass filter number (steps S2 to S7).

を This is sequentially performed for all bandpass filters (bandpass filters 12B,..., 12E) (steps S8 to S10).

After such preliminary exposure is completed, at the time of measurement, the classification operation circuit 28 corrects the measurement data of the corresponding band-pass filter based on the aperture value data stored for each band-pass filter, and then performs the classification operation. I do.

As described above, in the first embodiment, the exposure value memory 29 is provided, and imaging is performed with optimal exposure for each bandpass filter. Therefore, the measurement data SNR is improved, and the classification accuracy is increased.

In addition, in the present embodiment, since the pre-exposure is performed and the measurement can be performed with the optimum exposure for each bandpass filter, classification can be performed with high accuracy even when the measurement target changes.

{(Second Embodiment)} Next, a second embodiment for preventing the SNR from being reduced due to dark current noise of the image sensor when the number of objects is limited to some extent will be described with reference to FIG.

The color classification apparatus according to the present embodiment includes an optical system 10 including a lens and the like as shown in FIG. 4, and a plurality of bandpass filters 12A, 12B,..., 12E as shown in FIG. , 103E, and has fan-shaped openings 103A, 103B,..., 103E having different widths so as to correspond to each band-pass filter of the rotating color filter as shown in FIG. , A rotary diaphragm plate 102 fixed to a motor shaft so as to overlap a corresponding bandpass filter, a filter position sensor 123, a motor 24, a motor drive circuit 124, a solid-state imaging device 14 for capturing images of an object and a reference plate, Each band of the amplifier 15, the image sensor driving circuit 22, the A / D converter 16, the frame memory 18, the classification operation circuit 28, and the rotating color filter 12 Pass filter 12A, 12B, ..., the exposure value memory 29 for storing the opening value of the rotation stop plate corresponding to 12E, made from the control circuit 26 which controls timed units.

As shown in FIG. 2, the rotary color filter 12 is provided with filter position detection holes 125A, 125B,..., 125E, and a filter initial position detection hole 126, as shown in FIG.

The position of each bandpass filter 12A, 12B,..., 12E is detected by the initial position detecting hole 126 and the filter position detecting holes 125A, 125B,. Then, the motor drive circuit is controlled so that the rotation of the filter is synchronized with the imaging of the image sensor by the signal from the filter position sensor 123.

The rotary diaphragm plate 102 rotates together with the rotary color filter 12.

.., 103E of the rotary color filter 12 correspond to the openings 103A, 103B,..., 103E of the rotary diaphragm plate 102, thereby limiting the exposure amount.

The image that has passed through each filter is formed on an image sensor 14, converted into a digital signal by an A / D converter 16 via an amplifier, and stored in a frame memory 18. The image signal and image are transmitted from the frame memory 18 to a monitor. Data in a predetermined area in the middle is sent to the classification operation circuit 28.

Here, the rotary diaphragm plate 102 will be described.

The approximate reflection spectral distribution of the object is checked in advance, and the optimum aperture diameter for each bandpass filter (bandpass filters 12B,..., 12E) is determined, and the rotary aperture plate is determined based on the aperture diameter. And the aperture diameter at this time is stored in the exposure value memory 29 in correspondence with the band-pass filter.

The classification operation circuit 28 performs a classification operation after correcting the measurement data of the corresponding bandpass filter based on the aperture value data stored in the exposure value memory 29.

As described above, in the second embodiment, the rotary diaphragm plate 102 and the exposure value memory 29 are provided, and the imaging is performed with the optimum exposure for each bandpass filter. Therefore, the SNR of the measurement data is improved, and the classification accuracy is improved. .

In the second embodiment, since there is no mechanical operation such as changing the aperture diameter, high-speed and stable measurement can be performed.

In addition, in the second embodiment, a plurality of different rotary diaphragm plates designed according to the intended use are prepared, and can be used for various purposes by replacing them.

{Third Embodiment} Next, a description will be given, with reference to FIGS. 6, 7, and 8, of a third embodiment for preventing the SNR from being reduced due to dark current noise of the image sensor when the number of objects is limited to some extent.

The color classification device according to the present embodiment includes an optical system 10 including a housing 1, a lens, and the like as shown in FIG. 6, and a plurality of bandpass filters 12A, 12B,..., 12E as shown in FIG. , And ND filters 105A, 105B,..., 105E having different transmittances corresponding to the respective band-pass filters of the rotating color filter as shown in FIG. For capturing images of the rotating ND filter 104, the filter position sensor 123, the motor 24, the motor driving circuit 124, the object and the reference plate, which are directly fixed to the rotating color filter so that the openings overlap the corresponding bandpass filters. Solid-state image sensor 14, amplifier 15, image sensor drive circuit 22, A / D converter 16, frame memory 18, classification operation circuit 28, Each bandpass filter 12A of the color filters 12, 12B, ..., the exposure value memory 29 for storing the transmission of the rotary ND filter 104 corresponding to 12E, made from the control circuit 26 which controls timed units.

Like the first and second embodiments, the rotary color filter 12 is provided with filter position detection holes 125A, 125B,..., 125E and a filter initial position detection hole 126 as shown in FIG. .

The position of each bandpass filter 12A, 12B,..., 12E is detected by the initial position detecting hole 126 and the filter position detecting holes 125A, 125B,. Then, the motor drive circuit is controlled so that the rotation of the filter is synchronized with the imaging of the image sensor by the signal from the filter position sensor 123.

The rotating ND filter 104 rotates together with the rotating color filter 12.

.., 105E of the rotating ND filter 104 correspond to the bandpass filters 12A, 12B,..., 12E of the rotating color filter 12, and limit the exposure amount.

The image that has passed through each bandpass filter is formed on an image sensor 14, converted into a digital signal by an A / D converter 16 via an amplifier, and stored in a frame memory 18. The image signal is transmitted from the frame memory 18 to a monitor. , Data of a predetermined area in the image is sent to the classification operation circuit 28.

Here, the rotating ND filter 104 will be described.

The reflection spectral distribution of the main body of the object is checked in advance, the optimum transmittance is determined for each band-pass filter (band-pass filters 12B,..., 12E), and rotation is performed using an ND filter having this transmittance. The ND filter 104 is created, and the transmittance at this time is stored in the exposure value memory 29 in correspondence with the band-pass filter.

The classification operation circuit 28 performs classification operation after correcting the measurement data of the corresponding bandpass filter based on the transmittance data stored in the exposure value memory 29.

As described above, in the third embodiment, the rotation ND filter 104 and the exposure value memory 29 are provided, and the imaging is performed with the optimal exposure for each bandpass filter. Therefore, the SNR of the measurement data is improved, and the classification accuracy is improved. .

In addition, in the third embodiment, since there is no mechanical operation such as changing the aperture diameter, high-speed and stable measurement can be performed.

In addition, in the third embodiment, a plurality of different rotating ND filters designed according to the application are prepared and can be used for various purposes by replacing them.

In the third embodiment, a rotary slit plate 104A as shown in FIG. 8 may be prepared instead of the rotary ND filter 104.

In the rotating slit plate, the circumferential length of the openings 106A, 106B,..., 106E corresponding to the respective band-pass filters is a length corresponding to the transmittance of each ND filter in FIG.

This limits the exposure time of each bandpass filter, and as a result, the amount of exposure can be controlled, and the same effect can be obtained.

{(Fourth Embodiment)} A further versatile fourth embodiment will be described with reference to FIGS.

The color classification device of the present embodiment includes a housing 1, an optical system 10 including a lens and the like as shown in FIG. 9, and a plurality of bandpass filters 12A, 12B,..., 12E as shown in FIG. Rotating color filter 12, filter position sensor 123, motor 24, motor drive circuit 124, transmittance variable filter 122 capable of changing transmittance, transmittance control circuit 121 controlling transmittance of transmittance variable filter 122, object Image sensor 14, amplifier 15, image sensor drive circuit 22, A / D converter 16, frame memory 18, classification operation circuit 28, and transmittance for obtaining appropriate exposure of each band-pass filter for capturing images of an object and a reference plate , And an exposure value memory 29 for storing the

As shown in FIG. 2, similarly to the first to third embodiments, the rotary color filter 12 is provided with filter position detection holes 125A, 125B,..., 125E and a filter initial position detection hole 126. .

The position of each bandpass filter 12A, 12B,..., 12E is detected by the initial position detecting hole 126 and the filter position detecting holes 125A, 125B,. Then, the motor drive circuit is controlled so that the rotation of the filter is synchronized with the imaging of the image sensor by the signal from the filter position sensor 123.

The image that has passed through each filter is formed on an image sensor 14, converted into a digital signal by an A / D converter 16 via an amplifier 15, stored in a frame memory 18, and an image signal is sent from the frame memory 18 to a monitor. The data of a predetermined area in the image is sent to the classification operation circuit 28.

Next, the operation of the fourth embodiment will be described with reference to the flowchart shown in FIG.

First, a measurement area of the object is set (step S11).

Next, the first band-pass filter is set (band-pass filter 12A) to perform preliminary exposure, and the transmittance of the transmittance variable filter 122 is set so that the measurement data obtained by the preliminary exposure falls within a predetermined value range. Is stored in the exposure value memory 29 together with the corresponding bandpass filter number at this time (steps S12 to S17).

を This is sequentially performed for all bandpass filters (bandpass filters 12B,..., 12E) (steps S18 to S20).

After the completion of such preliminary exposure, at the time of measurement, the transmittance of the transmittance variable filter 122 is changed in synchronization with the bandpass filters 12A, 12B,..., 12E. After correcting the measurement data of the corresponding band-pass filter based on the transmittance data stored for each, a classification operation is performed.

As described above, in the fourth embodiment, the transmittance variable filter 122 and the exposure value memory 29 are provided, and the configuration is such that imaging is performed with the optimum exposure for each bandpass filter. Therefore, the SNR of the measurement data is improved, and the classification accuracy is improved. Get higher.

Further, in the present embodiment, since it is possible to perform pre-exposure and perform measurement with the optimum exposure for each bandpass filter, classification can be performed accurately even when the measurement object changes, so that a versatile color classification device can be realized. it can.

In the fourth embodiment, as shown in FIG. 11, the transmittance variable filter 122 and its transmittance control circuit 121 shown in FIG. 9 may be removed to control the exposure time of the image sensor. This further simplifies the circuit configuration.

<< (5th Embodiment) >> Next, 5th Embodiment which simplified measurement of the reference plate mainly at the time of outdoor use is demonstrated with reference to FIG. 12, FIG.

The color classification apparatus of the present embodiment includes a housing 1, an optical system 10 including a lens and the like, as shown in FIG. 12, and a plurality of bandpass filters 12A, 12B,..., 12E as shown in FIG. Rotating color filter 12, filter position sensor 123, motor 24, motor driving circuit 124, image sensor 14, amplifier 15, image sensor driving circuit 22, A / D converter 16 for capturing images of an object and a reference plate , A frame memory 18, an illumination condition detection circuit 137 for detecting an illumination condition from data in the frame memory 18, a classification operation circuit 28, a control circuit 26, and a transmission-type reference plate 131 provided in an upper front part of the housing 1. The optical system includes an optical system 133 for forming an image of the reference plate on the image sensor, and a mirror 134 whose angle is variable about one end.

At the time of measurement with respect to the reference plate 131, the mirror 134 is set to a position indicated by a broken line 134 'in FIG. The transmissive reference plate 131 provided at the top of the housing 1 transmits the illumination light from the top into the housing 1 while diffusing the illumination light from the top.

The reference plate 131 reflecting the spectral distribution of the light source passes through the optical system 133, is reflected by the mirror 134, and forms an image on the image sensor.

The image that has been rotated by the rotating color filter 12 and passed through each band-pass filter is converted into digital data, as described later, and then image-processed by an illumination condition detection circuit 137 to obtain illumination condition data such as brightness and illumination direction. Then, light source data is obtained. The data on the direction of illumination will be described later in detail.

(4) The brightness data obtained by the illumination condition detection circuit 137 is sent to the control circuit 26 and used for exposure adjustment.

(4) The light source data is sent to the classification calculation circuit 28 and stored in the memory in the classification calculation circuit 28, and the data on the direction of illumination is also sent to the classification calculation circuit 28.

When measuring an object, the mirror 134 is located at a position 134 ′ shown by a broken line in FIG. 12, and reflected light from the object forms an image on the image sensor 14 through the optical system 10 and the rotating color filter 12. Is done.

A signal is sent from the control circuit 26 to the motor drive circuit 124, and the rotary color filter 12 is rotated by the motor 24 so that the position of each bandpass filter is synchronized with the imaging of the image sensor 14.

The captured image of the object that has passed through each band-pass filter is formed on an image sensor 14, converted into digital data by an A / D converter 16 via an amplifier 15, and stored in a frame memory 18. The data is sent to the classification operation circuit 28, measured in advance, stored in a memory in the classification operation circuit 28, corrected for the light source with the data of the reference plate 131, and subjected to the classification operation.

Here, a method of acquiring data on the direction of illumination will be described.

棒 Attach the rod 132 to the center of the reference plate 131 as shown in FIG.

This reference plate 131 is observed by the image sensor 14 as an image 135 as shown in FIG.

That is, since the shadow 136 of the bar 132 is reflected in the image 135 of the reference plate 131, the illumination condition detection circuit 137 detects the illumination condition from this image.

First, the direction of illumination at the time of measurement can be detected from the angle and length of the shadow 136 of the bar, and in the case of light cloudiness, the difference in brightness between the shadow 136 of the bar and the image 135 of the surrounding reference plate is reduced, and the complete In the case of cloudy weather, there is no shadow, so that the lighting conditions can be grasped in detail.

In the classification operation circuit 28 to which information of the illumination detected by the illumination condition detection circuit 137 is sent, learning data having different illumination conditions such as normal light, backward movement, and cloudy weather are prepared in advance to perform measurement based on the difference in illumination conditions. The classification operation can be performed while eliminating errors as much as possible.

As described above, in the fifth embodiment, the reference plate 131 is provided in the upper front part of the housing 1 so that the data of the reference plate 131 can be obtained only by switching the angle of the mirror 134. Even when the illumination light changes from fine to cloudy, the data of the reference plate 131 can be easily obtained, for example, the data of the reference plate 131 can be easily retaken.

The angle of the mirror 134 may be changed manually, or a signal may be sent from the control circuit 26 using a driving device to change the angle automatically.

(Sixth Embodiment) Next, a sixth embodiment in which measurement of a reference plate is simplified mainly when used outdoors will be described with reference to FIGS. 14, 15 and 16.

The color classification device of the present embodiment includes a housing 1, an optical system 10 including a lens and the like, as shown in FIG. 14, and a plurality of bandpass filters 12A, 12B,..., 12E as shown in FIG. Rotating color filter 12, filter position sensor 123, motor 24, motor driving circuit 124, image sensor 14, amplifier 15, image sensor driving circuit 22, A / D converter 16 for capturing images of an object and a reference plate , A frame memory 18, an illumination condition detection circuit 145 for detecting illumination conditions from data in the frame memory 18, a classification operation circuit 28, a control circuit 26, a reference plate 138 provided on an upper part of the housing 1, and a reference plate 138. Mirror 141 and mirror 142 arranged so as to form an image on the image sensor 14 through the optical system 10, and these mirrors 141 and 142 are fixed to the housing 1. An arm 143 and 144 Metropolitan for.

The present embodiment relates to a measurement apparatus that measures outdoors, that is, uses sunlight as illumination light.

場合 When measuring an object, the reflected light from the object passes through the optical system 10 and the rotating color filter 12 and forms an image on the image sensor 14.

同時 に Simultaneously, the image of the reference plate 138 is reflected by the mirrors 141 and 142 and is imaged on a part of the image sensor 14 through the optical system 10.

FIG. 15 shows this image formation. In the entire screen, an object 146 is shown, and in a lower part of the center of the screen, a reference plate 138 reflected by mirrors 141 and 142 is shown. ing.

An area 147 surrounded by a square in the center of the screen indicates a measurement area, which is a range for measuring a color for classifying an object and can be arbitrarily changed (position, size). .

When a signal is sent from the control circuit 26 to the motor drive circuit 124 and the rotating color filter 12 is rotated by the motor 24 so that the position of each band-pass filter is synchronized with the imaging of the image sensor 14, the image sensor 14 As shown in FIG. 15, the image of the object 146 and the image of the reference plate 138 are simultaneously formed.

(4) The captured image of the object 146 and the reference plate 138 that have passed through each band-pass filter is formed on the image sensor 14, A / D converted via the amplifier 15, and stored in the frame memory 18.

The illumination condition detection circuit 145 detects the illumination condition data from the data of the reference plate 138 in the image data stored in the frame memory 18 and sends it to the control circuit 26, and sends the data of the reference plate 138 to the classification operation circuit 28. And stored in an internal memory.

The control circuit 26 controls the shutter speed of the image pickup device 14 based on the brightness data from the illumination condition detection circuit 145 so as to obtain an appropriate exposure, and sends the data of the illumination direction to the classification calculation circuit 28.

The data of the object 146 in the image data stored in the frame memory 18 is sent to the classification operation circuit 28, and the data of the reference plate 138 stored in the memory in the classification operation circuit 28 is used to correct the light source and perform classification. It is calculated.

Also, a rod 139 is attached to the center of the reference plate 138 as shown in FIG. The direction of the illumination light at the time of measurement can be detected from the relative angle and the length of the illumination light.

輝 度 Also, in the case of light cloudiness, the difference in brightness between the shadow 149 of the bar and the image of the surrounding reference plate 138 becomes small, and in the case of perfect cloudy weather, the shadow disappears, so that the lighting conditions can be grasped in detail.

As described above, in the sixth embodiment, the reference plate 138 is provided on the upper portion of the housing 1, and the mirrors 141 and 142 are provided so that the reference plate 138 is focused on a part of the imaging device 14 through the optical system 10. Since the image of the reference plate 138 is included in the same image when measuring the object 146, the measurement of the reference plate 138 and the measurement of the object 146 can be performed simultaneously, and the labor of the measurement is greatly simplified, and Further, the illumination condition does not change between the time of measuring the reference plate and the time of measuring the object, and the reliability of correction is improved.

In addition, according to the present embodiment, the provision of the rod 139 on the reference plate 138 makes it possible to detect lighting conditions such as direct light, backlight, and cloudy weather from information on the shadow of the reference plate 138. By preparing the data, it is possible to perform the classification calculation while minimizing the measurement error due to the difference in the lighting conditions, and the classification accuracy is improved.

{Seventh Embodiment} Next, a seventh embodiment in which measurement of the reference plate is simplified mainly when used outdoors is described with reference to FIG.

As shown in FIG. 17, the color classification device of this embodiment includes a housing 1, an optical system 10 for imaging an object, and a plurality of bandpass filters 12A, 12B as shown in FIG. , 12E, a rotary color filter 12, a filter position sensor 123, a motor 24, a motor drive circuit 124, an image sensor 14 for capturing an image of an object, an amplifier 15, an image sensor drive circuit 22, an A / D converter 16, a frame memory 18, a classification operation circuit 28, a control circuit 26, a reference plate 151 provided in the front part of the housing 1, an image sensor 154, and an optical system 152 for forming an image of the reference plate 151 on the image sensor 154. An A / D converter 156 for converting an image signal from the image sensor 154 into digital data via an amplifier 15a; Consisting illumination light detection circuit 137 Metropolitan for detecting the matter.

実 施 The present embodiment relates to a measurement device that uses outdoor sunlight, ie, sunlight as illumination light.

場合 When measuring an object, the reflected light from the object passes through the optical system 10 and the rotating color filter 12 and forms an image on the image sensor 14.

A signal is sent from the control circuit 26 to the motor drive circuit 124, and the rotary color filter 12 is rotated by the motor 24 so that the position of each bandpass filter is synchronized with the imaging of the image sensor.

{Circle around (2)} An image signal is obtained from the image sensor 14 after passing the reflected light from the object through each band-pass filter, and is processed in the same manner as in the above-described embodiments.

On the other hand, the image pickup device 154 is disposed at a diagonal position with respect to the image pickup device 14 with respect to the rotation color filter 12, so that when the rotation color filter 12 makes one rotation, reflected light from the reference plate 151 is emitted from the image pickup device 154. An image signal that has passed through each bandpass filter is obtained.

The image signal of the reference plate 151 thus obtained is converted into digital data by the A / D converter 158 via the amplifier 15a, and then the illumination condition detection circuit 137 detects the illumination condition data and sends it to the control circuit 26. At the same time, the data of the reference plate 151 is sent to the classification operation circuit 28.

The illumination condition can be grasped in detail by configuring the reference plate 151 as shown in FIG. 16 similarly to the sixth embodiment.

The control circuit 26 controls the shutter speeds of the image pickup device 14 and the image pickup device 154 based on the brightness data from the illumination condition detection circuit 137 so as to obtain an appropriate exposure, and sends the data of the illumination direction to the classification calculation circuit 28. .

The data of the object stored in the frame memory 18 is sent to the classification operation circuit 28, where the data of the reference plate stored in the memory in the classification operation circuit 28 is used to correct the light source to perform the classification operation.

As described above, in the seventh embodiment, the reference plate 151 is provided at the front part of the housing 1, and the reference plate 151 is configured to capture data that has passed through each band-pass filter using the dedicated optical system 152 and the image sensor 154. As a result, the measurement of the reference plate and the measurement of the object can be performed at the same time, which greatly simplifies and speeds up the measurement, and furthermore, the lighting conditions between the reference plate measurement and the object measurement. Does not change, and the reliability of correction is improved.

Further, according to the present embodiment, by providing the reference plate 151 with a bar similar to the bar 139 of the sixth example, it becomes possible to detect the lighting conditions such as direct light, backlight, and cloudy from the shadow information. By preparing learning data having different lighting conditions, measurement errors due to differences in lighting conditions can be eliminated as much as possible.

In the case where data such as the direction of illumination is not required, there is no need to provide a bar or a scale on the reference plate 151, and it is sufficient to use a simple photodiode or CdS instead of the image sensor 154. Therefore, the configuration can be further simplified.

(Eighth Embodiment) Next, an eighth embodiment that does not use a rotating color filter will be described with reference to FIGS.

As shown in FIG. 18, the color classification apparatus according to the present embodiment includes a housing 1, an optical system 10 for imaging an object, and a fiber bundle having one end face aligned with an imaging surface on which the object is imaged. 161, a bandpass filter 162a, 162b, 162c, 162d coupled to the opposite end face divided into a plurality of parts, an image sensor 14, an amplifier 15, and an image sensor drive for capturing an image of an object passing through the bandpass filter The circuit 22, the A / D converter 16, the frame memory 18, the classification operation circuit 28, the control circuit 26, a transmissive reference plate 164 provided on the upper part of the housing, and a fiber bundle 165 having the end faces aligned with the reference plate 164 , And band pass filters 166a, 166b, 166c, and 166d provided between the opposite end of the fiber bundle 165 and the image sensor 14.

In the present embodiment, the image pickup surface of the image pickup device 14 is divided so that data that has passed through a plurality of band-pass filters can be simultaneously input for both the object and the light source. There are four types of bandpass filters.

Here, the configurations of the fiber bundle 161, the fiber bundle 165, the reference plate 164, and the band-pass filters 162a to 162d and 166a to 166d used in the present embodiment will be described with reference to FIGS.

First, the fiber bundle 165 is connected to the reference plate 164 with one end face aligned as shown in FIG.

Next, as shown in FIG. 20, bandpass filters 162a to 162d and 166a to 166d are attached to the imaging surface 167 of the imaging element 14 in an arrangement as shown in FIG.

う ち Of these, the band-pass filters 162a and 166a, 162b and 166b, 162c and 166c, and 162d and 166d have the same band-pass characteristics.

反 対 The other end of the fiber bundle 165 is attached to bandpass filters 166a to 166d.

In this case, the opposite ends of the fiber bundle 165 are aligned and placed on the image plane of the optical system 10 as a plane, but the opposite end is divided into four parts.

FIG. 21 shows a part of the cross section of the fiber bundle 165 on the image forming plane. In FIG. 21, the fiber indicated by the numeral 1 surrounded by a circle is connected to the bandpass filter 162a, and the fiber indicated by the numeral 2 surrounded by a circle is: In the band-pass filter 162b, the fiber indicated by the numeral 3 encircled by {circle around (4)} is connected to the band-pass filter 162c, and the fiber indicated by the numeral 4 encircled by {circle around (4)} to the band-pass filter 162d. Divide and bundle.

Therefore, the opposite end of the fiber bundle 165 becomes a reduced screen of each divided image plane, and is input to the image sensor 14 through each band-pass filter.

Thereby, the object 160 imaged as shown in FIG. 19 is observed as an image signal as shown in FIG.

This image signal is digitized by an A / D converter 16 through an amplifier 15 and then written into a frame memory 18. Illumination data for correction from a portion passing through a band-pass filter 166 a to 166 d from a reference plate 164. Then, data of a predetermined area of the object is sent to the classification operation circuit 28 for each of the band-pass filters 166a to 166d.

The classification operation circuit 28 corrects the data of the object with the illumination data, performs a classification operation, and outputs the result.

As described above, according to the eighth embodiment, the imaging surface of the imaging element 14 is divided, data of an object corresponding to a plurality of bandpass filters is simultaneously formed, and illumination data is simultaneously captured in the same frame. With this configuration, the measurement time is significantly reduced.

Since FIG. 18 is an example for outdoor measurement, the transmission type reference plate 164 is fixed to the upper part of the housing 1 so as to catch sunlight, but as shown in FIG. , The reference plate 164 can be drawn out of the housing 1 and arranged next to the object, so that it can be used for indoor measurement.

Further, the fiber bundle 165 in which the reference plate 164 is attached to the tip for acquiring illumination data is configured to be detachable from the housing 1 using the optical connector 169 as shown in FIG. .

In the present embodiment, the number of bandpass filters is four, but the number of bandpass filters is not limited to four, and it goes without saying that the number of bandpass filters may be an optimum number.

{Ninth Embodiment} Next, a ninth embodiment that does not use a rotating color filter will be described with reference to FIGS.

As shown in FIG. 25, the color classification device according to the present embodiment includes a housing 1, an optical system 10 for forming an image of an object, an optical path splitting prism 170a, and two split lights emitted from the optical path splitting prism 170a. The incident light path splitting prisms 170b and 170c and the image pickup element 14, the amplifier 15, and the light path splitting prisms 170b and 170c between the image pickup element 14 where the light split and output by the light path splitting prisms 170b and 170c form an image. Bandpass filters 162a to 162d arranged so that each of the divided lights reaches the image sensor 14 after passing through different bandpass filters, the image sensor drive circuit 22, the A / D converter 16, the frame memory 18, the classification operation It comprises a circuit 28, a control circuit 26, and a light source spectral sensor 180 provided on the upper part of the housing.

In the present embodiment, the image pickup surface of the image sensor 14 is divided so that an image of an object that has passed through a plurality of band-pass filters can be input at the same time. Here, the band-pass filters required for classification are as follows. There are four types.

First, the configuration of the optical path splitting prisms 170a to 170c used in the present embodiment, and the configurations of the optical path splitting prisms 170a to 170c, the band pass filters 162a to 162d, and the image sensor 14 will be described.

FIG. 26 shows the configuration of one of the optical path splitting prisms 170a to 170c. In FIG. 26, the optical path is shown by a dashed line, the half mirror is shown by a broken line, and the mirror is shown by a thick line.

26, the light incident from the left side in FIG. 26 is split into the half mirror 171 by 50% in two directions, and the light transmitted through the half mirror 171 is reflected by mirrors 162, 163, 164, and 165, respectively, and output to the right side in FIG. You.

On the other hand, the light reflected by the half mirror 171 is reflected by the mirrors 176, 177, and 178, and emerges on the right side in FIG.

That is, the light incident on the optical path splitting prism is split into two, and the split light is shifted in optical axis and output side by side from the exit of the prism.

３Three optical path splitting prisms 170a to 170c are prepared, and these and the image pickup device 14 are assembled as shown in FIG.

In this case, by configuring the two outputs of the optical path splitting prism 170a to be incident on the optical path splitting prisms 170b and 170c, the light incident on the optical path splitting prism 170a finally passes through the optical path splitting prisms 170b and 170c. It is divided into four parts and output in two rows and two columns.

Further, four bandpass filters 162a to 162d having different characteristics required for measurement are attached to the image sensor 14 as shown in FIG. 28, and each is divided into four by optical path dividing prisms 170a to 170c. Corresponding to the image.

With the above configuration, the reflected light from the object is divided into four parts and simultaneously imaged on the image sensor 14 through different bandpass filters.

The image signal output from the image pickup device 14 is digitized by an A / D converter 16 via an amplifier 15 and stored in a frame memory 18. Is read from the frame memory 18 and sent to the classification operation circuit 28.

On the other hand, the data from the light source spectral sensor 180 is also sent to the classification operation circuit 28 at the same time.

Here, as shown in FIG. 29, the configuration of the light source spectral sensor 180 includes a transmittance variable filter 181 disposed on the outermost side, a plurality of bandpass filters 182a to 182d disposed on the inner side thereof, A plurality of optical sensors 183a to 183d corresponding to the pass filters 182a to 182d; a transmittance control circuit 184 to which signals from the plurality of optical sensors 183a to 183d are input; a signal from the transmittance control circuit 184; And a signal from the optical sensors 183a to 183d.

The plurality of bandpass filters 182a to 182d are the same as the bandpass filters 162a to 162d for classifying an object.

(4) The optical sensors 183a to 183d can use photodiodes, phototransistors, CdS, and the like, and the transmittance variable filters 181 can use liquid crystal filters and the like.

(4) The transmittance control circuit 184 controls the transmittance of the transmittance variable filter 181 so that each optical sensor is not saturated.

The illumination condition detecting circuit 185 detects the brightness of the light source from the output of the transmittance control circuit and outputs the spectral data of the light source from each optical sensor.

As described above, in the classification operation circuit 28 to which the data obtained by the light source spectral sensor 180 is input, the data of the object read from the frame memory 18 is converted into the illumination obtained by the light source spectral sensor 180. And then perform a classification operation.

According to the ninth embodiment, the imaging surface of the image sensor 14 is divided and the data of the object corresponding to the plurality of bandpass filters is formed simultaneously, thereby greatly reducing the measurement time. In addition, since there is no mechanical movable part, the operation is stable and the durability is improved.

Since FIG. 25 is an example for measurement outdoors, the light source spectroscopic sensor is fixed to the upper part of the housing so as to catch sunlight, but the sensor itself is not fixed to the housing but connected by a cable. With such a configuration that it can be arranged next to the object, it can be used for indoor measurement.

Of course, the configuration is such that the cable of the light source spectroscopic sensor can be detached from the housing using a connector, making it easier to carry.

Also, in the present embodiment, the number of bandpass filters is four, but it is needless to say that the number of bandpass filters may be increased or the number of bandpass filters may be reduced.

<< (Tenth Embodiment) >> Next, a tenth embodiment not using a rotating color filter will be described with reference to FIG.

As shown in FIG. 30, the color classification apparatus according to the present embodiment includes a housing 1, an optical system 10 for imaging an object, a band-variable bandpass filter 190, an image sensor (CCD) 14, an amplifier 15, and a CCD. It comprises a drive circuit 22, an A / D converter 16, a frame memory 18, a classification operation circuit 28, and a control circuit 26.

帯 域 The band variable band-pass filter 190 used in the present embodiment is capable of controlling the frequency of the pass band with an external signal.

画像 The image of the object passing through the optical system 10 is transmitted through the band variable bandpass filter 190 and formed on the CCD 14.

The image signal output from the CCD 14 is digitized by an A / D converter 16 via an amplifier 15 and stored in a frame memory 18. Then, data of a predetermined area of the object is stored in a film for each bandpass filter. The data is read from the memory 18 and sent to the classification operation circuit 28.

The control circuit 26 sets the pass band of the band variable band-pass filter 190 to a predetermined value, and sends a control signal to the CCD drive circuit 22 and the frame memory 18 so as to write data captured by the CCD 14 to the frame memory 18.

動作 This operation is performed over a predetermined number of bands, and the classification operation circuit 28 corrects the data of the object with the data of the reference plate measured in advance or simultaneously, and performs the classification operation.

According to the tenth embodiment, a variable band-pass filter 190 is used instead of a plurality of band-pass filters, and the pass band is controlled to take in data of an object. A high-precision approximation is also possible for the classification spectrum, and the operation is stable and the durability is improved because there is no mechanical movable part.

Furthermore, in the present embodiment, the degree of freedom in setting the number and band of the bandpass filters is large and the versatility is high. Therefore, even when limiting to a specific application, the optimization can be performed only by changing the control software of the control circuit 26. Becomes possible.

In this embodiment, the image data is read for each predetermined band. However, in the classification mode after the classification spectrum is obtained in the learning mode, the band characteristic of the band variable band-pass filter 190 is controlled during the exposure time of the CCD 14. Alternatively, the characteristics of the classified spectrum may be approximated.

In other words, the band of the band variable bandpass filter 190 is set during the exposure time, and after holding it for a certain period of time, once it is set to the next band, it is held for another period of time.

At this time, the holding time after the band setting of the band variable band-pass filter 190 is controlled so as to be proportional to the value of each band of the classified spectrum to be approximated (the band where the value of the classified spectrum is negative is: Exposure is performed collectively for only the negative bands, and subtraction is performed on the frame memory).

動作 Operating in this way is the same as performing the inner product operation with the classification spectrum, and the classification can be determined based on the amount of charge accumulated during the exposure time.

<< (Eleventh Embodiment) >> Next, an eleventh embodiment using a classification operation circuit for performing a multi-class classification operation in each of the above embodiments will be described with reference to FIG.

When performing a multi-class classification operation, for example, in the case of classification of four classes of classes 1 to 4, when the classes 2 and 3 are projected onto the space formed by the classification spectra calculated from the data of class 1 and class 4, the class 1 In some cases, the boundary between classes such as, class 2 or class 3 and class 4 is not clear, and the classification accuracy may be reduced.

In the present embodiment, even in such a case, classification can be performed effectively without lowering the classification accuracy.

The classification operation circuit 28 of the present embodiment includes a luminance component extraction unit 30, a classification operation unit 32, and a classification determination unit 34, as shown in FIG.

構成 The configuration of the classification operation unit 32 used in the present embodiment will be described with reference to FIG.

As shown in FIG. 31B, the classification operation unit 32 of the present embodiment includes a switch 202 for switching the output of the d1 memory 200 and the d2 memory 201, and the d1 memory 200 and the d2 memory 201 for storing the classification spectrum; , A multiplier 203 for taking the product of the spectrum data from the unknown object, an accumulator 206 comprising an adder 204 and a memory 205, a d1 memory 210, a d2 memory 211, and a d1 memory 213 for storing the classified spectrum. d2 memory 214, switch 212 for switching the output of d1 memory 216 and d2 memory 217, output of d1 memory 210 and d2 memory 211, switch 215 for switching the output of d1 memory 213 and d2 memory 214, d1 memory 216 and d2 memory A switch 218 for switching the output of the accumulator 217; A classification spectrum selection circuit 207 for selecting one of the three types of classification spectra based on the signal from the input unit, a multiplier 223 for multiplying the selected classification spectrum by the spectrum data from the unknown object, an adder 224, and a memory 225. And an accumulator 226 comprising:

Next, the operation of the classification operation unit 32 according to the present embodiment will be described. Here, the number of classes to be classified is four (1 to 4).

These four classes are assumed to be distributed in the order of numbers in a multidimensional space, and the d1 memory 200 and the d2 memory 201 store the classification spectra d1 _1-4 and d2 ₁₋ calculated from the learning data of the classes 1 and 4 respectively. ₄ and are respectively stored.

Also, d1 memory 210 and d2 class 1 and class classification spectrum d1 _1-2 calculated from the learning data 2 in the memory 211 and d2 _1-2, d1 memory 213 and d2 in the memory 214 Class 2 and Class 3 classification spectrum d1 _2-3 calculated from learning data with the d2 _2-3, d1 memory 216 and d2 in the memory 217 class 3 and class classification spectrum d1 _3-4 calculated from learning data 4 and d2 _{3 -4} are stored.

For the unknown data of the object from the luminance component extraction unit 30, the product of the classification spectrum d1 _1-4 from the d1 memory 200 and each component (dimension) is obtained by the multiplier 203 in the classification calculation unit 32.

積 The products of the components are added up by the accumulator 206 and input to the classification spectrum selection circuit 207.

出力 The output of the accumulator 206 is the inner product of the unknown data and the classification spectrum as a result.

Next, the switch 202 is switched to calculate the inner product value for d2 in the same manner and send it to the classification spectrum selection circuit 207.

The classification spectrum selection circuit 207 is composed of a classification judgment circuit 230 and a selector 231 as shown in FIG. 31C, and when the inner product value is input from the accumulator 206, the classification judgment circuit 230 roughly classifies the classification. A determination is made.

Here, the unknown data is projected onto the space formed by the classification spectra d1 _1-4 and d2 _1-4 calculated from the data of class 1 and class 4, and the class is determined by the determined classification boundary.

Here, the boundaries are determined so as to be classified into three new classes, such as "from class 2 to class 1", "from class 2 to class 3", and "from class 3 to class 4".

If the classification spectrum selection signal output from the classification determination circuit 230 is “from class 2 to class 1”, the selector 231 outputs the input from the switch 212 as a classification spectrum.

As a result, the classified spectrums d1 _1-2 and d2 _1-2 are selected, the inner product operation of the unknown data is performed by the multiplier 223 and the accumulator 226, and the inner product value is obtained by the classification spectrum from the classification spectrum selection circuit 207. The final class is sent to the classification judging unit 34 together with the selection signal.

When the classification spectrum selection signal output from the classification determination circuit 230 is “between class 2 and class 3”, the selector 231 outputs the input from the switch 215 as the classification spectrum, and outputs “class 3 to class 3”. If it is “close to 4”, the input from the switch 218 is output as a classification spectrum.

Thereby, the optimal classification spectrum is selected, and the inner product calculation is performed.

In the present embodiment, the first-stage classification spectrum obtained from the data of class 1 and class 4 is used. However, this does not mean that the data of the two classes at both ends of the distribution is used. A classification spectrum obtained from the data of class 2 and class 3 may be used.

Further, a new class 1 'in which classes 1 and 2 are combined and a new class 4' in which classes 3 and 4 are combined may be defined, and a classification spectrum calculated using data of these classes may be used. .

Further, in the present embodiment, the classification is determined in two stages, but the same effect can be obtained by performing the determination in three stages or more stages in accordance with the number of classes to be classified. Not even.

According to the eleventh embodiment, since the classification is determined in multiple stages, effective classification can be performed without deteriorating the classification accuracy even in multi-class classification.

<< (Twelfth Embodiment) >> Next, a twelfth embodiment using a classification operation circuit for performing a multi-class classification operation in each of the above embodiments will be described with reference to FIG.

When performing multi-class classification, for example, in the case of three classes of classes 1 to 3, when class 3 is projected onto the space formed by the classification spectra calculated from the data of class 1 and class 2, FIG. As shown in (2), the distributions of class 2 and class 3 overlap, the boundary between the classes may not be clear, and the classification accuracy may decrease.

In the present embodiment, even in such a case, classification can be performed effectively without lowering the classification accuracy.

(C) The classification operation circuit of this embodiment is shown in FIGS. 32 (C) and (D). First, the configuration of the luminance component extraction unit 30 will be described with reference to FIG. 32 (C).

As shown in FIG. 32C, the luminance component extraction unit 30 used in the classification operation circuit 28 of the present embodiment includes a measurement region extraction unit 36 that cuts out the measurement region data from the image data in the frame memory, and obtains the data over the measurement region. A luminance component averaging unit 38 for obtaining an average value of the obtained data, an illumination data memory 40W for storing data of the reference plate, a measurement data memory 40A for recording data of the target, and a data for the reference plate. It comprises a correction circuit 42 for correcting with data, and correction data memories 241, 242, 243 for storing data of the corrected object.

Next, the configuration of the classification operation unit 32 will be described with reference to FIG.

The classification operation unit 32 used in this embodiment includes a switch 251 controlled by the control circuit 26 as shown in FIG. 32D, a class selection circuit 252 connected to one output of the switch 251, and a switch 251. The classification circuit calculator 253 connected to the same output of 251, two d1 memories 254 and 255 for storing the classification spectrum calculated by the classification spectrum calculation unit 253, and two d1 memories 254 and 255 from the control circuit 26. And a scalar product operation circuit 257 for calculating the scalar product of the output data and the other output of the sWitcher 251.

Next, the operations of the luminance component extraction unit 30 and the classification calculation unit 32 according to the present embodiment will be described separately for the learning mode and the classification mode.

In the learning mode, the data of a known target is measured. However, the measurement of the reference plate is performed before (or simultaneously with) the measurement of the target, and the measured data of the reference plate is described as image data. This is stored in the frame memory 18 as described above.

The data of the measurement area is read from the frame memory 18 by the measurement area extraction unit 36, the average value is calculated over the entire measurement area by the luminance component averaging unit 38, and then stored in the illumination data memory 40W.

Similarly, the data of the known object is also read out of the measurement area by the measurement area extraction unit 36, the average value is calculated over the entire measurement area by the luminance component averaging unit 38, and then stored in the measurement data memory 40A. Is done.

The correction circuit 42 corrects the data in the measurement data memory 40A using the data in the illumination data memory 40W, and writes the data in the correction data memories 241, 242, and 243 of the respective classes.

The known data is measured a predetermined number of samples (N) for each class and stored in each of the correction data memories 241 to 243 as corrected learning data.

In the learning mode, in the classification operation unit 32, the switch 251 is set to the side b by the control circuit 26.

Data is sequentially read from the correction data memories 241, 242, and 243 of the luminance component extraction unit 30 by the signal from the control circuit 26, and the average spectrum of each class, the difference spectrum between the classes, and the norm thereof are calculated in the class selection circuit 252. Is calculated.

The class selection circuit 252 informs the control circuit 26 of the maximum class combination of the norm of the difference spectrum (here, class 1 and class 3).

Signals are sent from the control circuit 26 to the correction data memories 241 and 243, and the data is sequentially read out and sent to the classification spectrum calculation unit 253 via the switch 251.

The classification spectrum calculation unit 253 calculates the classification spectrum by using the FS conversion and writes the classification spectrum in the d1 memory 254.

That is, a classification spectrum that maximizes the fisher ratio (Fisher ratio) between the most distant classes of the average spectrum is obtained and written to the d1 memory 254.

Next, the class selection circuit 252 informs the control circuit 26 of the combination of the minimum classes of the norms of the difference spectrum (here, classes 2 and 3).

Signals are sent from the control circuit 26 to the correction data memories 242 and 243, data is sequentially read out, and sent to the classification spectrum calculation unit 253 via the switch 251.

The classification spectrum calculation unit 253 calculates the classification spectrum by using the FS conversion and writes the classification spectrum into the d1 memory 255.

That is, a classification spectrum in which the fisher ratio (Fisher ratio) between classes having the closest average spectrum is maximized is obtained and written to the d1 memory 255.

Next, the switch 251 is set to the a side, and the inner product calculation circuit 257 takes the inner product value between the data of each class and the classification spectra d1 _1-3 and d1 _2-3 and sends it to the classification determination unit 34.

The classification determination unit 34 obtains the distribution of the data of each class and sets the boundaries of each class.

That is, the data of each class is projected on the space formed by the classification spectra d1 _1-3 and d1 _2-3 , and a boundary is set as shown in FIG.

学習 This is the learning mode.

Next, the operation of each unit in the classification mode will be described.

参照 The reference plate is also measured in the classification mode, and is stored in the illumination data memory 40W in the luminance component extraction unit 30 as in the learning mode.

On the other hand, the unknown object data is also averaged in the measurement area as in the learning mode and stored in the measurement data memory, and corrected by the reference plate data stored in the illumination data memory 40W by the correction circuit 42. It is stored in the data memory 241.

In the 演算 classification operation unit 32, the switch 251 is set to the a side, and unknown data is sent from the correction data memory 241 of the luminance component extraction unit 30 to the inner product operation circuit 257.

と 同時 に At the same time, the switch 256 is set to the d1 memory 254 side.

Accordingly, the inner product value of the inner product calculation circuit 257 classification spectrum d1 _1-3 and unknown data are calculated.

Next, the switch 256 is set on the d1 memory 255 side, and the inner product calculation circuit 257 calculates the inner product value of the classification spectrum d1 _2-3 and the unknown data, and sends the inner product value of d1 _1-3 to the classification determining unit 34. Then, the unknown data is classified according to the classification boundary set in the learning mode.

According to such a twelfth embodiment, the classification boundary used for the classification determination is obtained from a combination of a set of classes (for example, d1 _1-3 and d2 _1-3 obtained from class 1 and class 3). Instead, the classification spectra (d1 _1-3 obtained from classes 1 and ₃ and d1 _2-3 obtained from classes 2 and 3 as in the present embodiment) obtained from a combination of a plurality of classes are used. By using the determination, the distribution between the classes does not overlap beyond the boundary as shown in FIG. 32A, and the boundary where the distribution of each class can be well separated as shown in FIG. 32B. Can be set, so that classification can be performed with higher accuracy in multi-class classification.

In this embodiment, the classification spectrum obtained from the largest difference spectrum between classes and the classification spectrum obtained from the smallest difference spectrum are used, but the classification spectrum obtained from the largest difference spectrum and the next largest difference spectrum are used. Alternatively, a classified spectrum obtained from the smallest difference spectrum and the next smallest difference spectrum may be used.

Furthermore, a second classification spectrum may be obtained from a combination of the minimum classes of Fisher ratio when projected onto the classification spectrum d1 obtained first.

Also, in the present embodiment, two classification spectra are used, but when the number of classes to be classified is large, the classification accuracy can be further improved by increasing the number of classification spectra.

<< (Thirteenth Embodiment) >> Next, a thirteenth embodiment using a classification operation circuit for performing a multi-class classification operation in each of the above embodiments will be described with reference to FIG.

When performing multi-class classification, for example, in the case of three-class classification of classes 1 to 3, when class 3 is projected onto the space formed by the classification spectra calculated from the data of class 1 and class 2, FIG. As shown in (2), the distribution of class 2 may be wider than the other two classes.

In this case, if the boundary between the classes is a perpendicular bisector of a straight line connecting the centers of the distributions of the classes, as shown in FIG. 33A, even if a sample of the class 2 is measured, it is determined that the sample is the class 1 or the class 2. May be performed, and the classification accuracy may be reduced.

In the present embodiment, even in such a case, classification can be performed effectively without lowering the classification accuracy.

FIG. 33C shows a classification operation circuit of this embodiment.

The classification determination unit 34 used in the classification calculation circuit 28 of the present embodiment includes a switch 280 for switching the data from the classification calculation unit 32 according to a mode (learning mode, classification mode), and calculates an average value and a variance value of each class. A statistic calculation unit 281 to be obtained, a boundary calculation unit 282 for calculating a boundary from the obtained statistic, a boundary memory 64 for storing the obtained boundary, and unknown data obtained by comparing the data of the boundary memory 64 with the measured unknown data. And a classification determination unit 66 for performing classification of.

Next, the operation of the classification determination unit 34 of the present embodiment will be described separately for the learning mode and the classification mode.

In the learning mode, the data of the known object is measured. The data of the object is corrected for the illumination in the luminance component extraction circuit 30, and then the inner product value of the object with the classification spectrum is calculated in the classification calculator 32. Is sent to the classification determination unit 34.

The inner product value of the known data is two-dimensional of d1 and d2, and a predetermined number of samples (N) is sent for each class.

In the classification determination unit 34, the switch 280 is set to the b side, and the statistic calculation unit 281 calculates the average and variance of N samples for each class with respect to the inner product value of the known data sent.

In the classification mode, the inner product value of unknown data d1 and d2 is sent to the classification determination unit 34.

In the classification determination unit 34, the switch 280 is set to the a side, and the boundary calculation unit 282 draws an equal dispersion line for each class as shown in FIG. 33 (B), and a line connecting the intersections of the equal dispersion lines. Is written to the boundary memory 64 as a boundary.

According to the thirteenth embodiment, since the statistic calculation unit 281 is used, the classification boundary used for the classification determination is determined using the variance value of each class. As shown in FIG. 33 (B), it is possible to set a boundary where the distribution of each class can be separated well without causing the class distribution to exceed the boundary as described above. Separation can be performed well.

(Fourteenth Embodiment) Next, in each of the above embodiments, a fourteenth embodiment using a luminance component extraction unit that optimally sets a measurement area when measuring a glossy object is described with reference to FIG. Will be explained.

When measuring a glossy object, if the glossy part (specular reflection component) is included in the measurement area, the color of the object cannot be measured correctly, and the classification accuracy may be reduced. I do. The luminance component extraction unit of the present embodiment will be described with reference to FIG.

The luminance extraction unit of the present embodiment includes a measurement region extraction unit 36, a luminance component averaging unit 38, a pixel counting unit 293, a measurement data memory 40A that temporarily stores measurement data of an object, and a measurement of a reference plate. It comprises an illumination data memory 40W for storing data, a correction circuit 42 for correcting the data of the object with the data of the reference plate, and a correction data memory 44 for storing the corrected data of the object for a plurality of classes.

The measurement area extraction unit 36 includes a specific area detection unit 290 and a non-measurement area setting unit 291.

FIG. 34B shows an example of a glossy object.

The data of the predetermined measurement area 300 is read into the measurement area extraction unit 36 from the frame memory by the control circuit 26.

(4) The read data enters the unique area detection unit 290, and an area that is abnormally bright compared to the surrounding area (specular reflection portion 301) is detected.

The non-measurement area setting unit 291 sets a range (non-measurement area 302) that is not used for measurement by giving a margin of a predetermined number of pixels around the pixels of the regular reflection detected by the regular reflection detection unit 290, and 26 to change the measurement area.

(4) The signal from the control circuit 26 is controlled by the signal from the control circuit 26 so that the output of the luminance component averaging unit 38 at the subsequent stage is not valid until the specular reflection in the measurement area disappears.

(4) The control circuit 26 skips the pixels specified by the non-measurement area setting section 291 and reads the data in the predetermined area of the frame memory again.

At this time, the regular reflection detection circuit 290 does not find any regular reflection pixel, and the non-measurement area setting unit 291 notifies the control circuit 26 that there is no non-measurement area.

新 し い The new measurement area at this time is shown in FIG.

Since the data of the object from which the non-measurement area has been skipped has a smaller number of pixels than the predetermined measurement area, the pixel count unit 293 obtains the total number of pixels. The average value is obtained from the data of the object and the number of pixels, and is temporarily stored in the measurement data memory 40A.

(4) The correction circuit 42 corrects the data of the object from the measurement data memory 40A using the data of the reference plate from the illumination data memory 40W and stores the corrected data in the correction data memory.

According to the fourteenth embodiment, since the specific region detection unit 290 and the non-measurement region setting unit 291 are provided in the measurement region extraction unit, the target object has a glossy specular reflection portion. In this case, the specular reflection portion can be removed from the measurement area, so that good measurement can be performed without lowering the classification accuracy.

In the present embodiment, a measurement area is set before measurement, and if there is a specular reflection portion in the measurement area, the operation is performed to remove it. An example is described below.

If the object or the portion to be measured of the object has a spherical shape or a shape close to it, and it is known that the surface is glossy, the specific region detection unit 290 detects a regular reflection region from the entire screen. In the non-measurement area setting section 291, the area not to be used for measurement (the non-measurement area 302) and the area surrounding the regular reflection pixel detected by the unique area detection section 290 with a predetermined number of pixels are provided. A measurement area 304 having a predetermined width is set and sent to the control circuit 26.

Until the measurement area is determined, the luminance component averaging unit 38 at the subsequent stage is controlled so that the output is not valid.

(4) The control circuit 26 reads the data of the measurement area determined by the non-measurement area setting unit 291.

新 し い The new measurement area at this time is shown in FIG.

Since the number of pixels differs depending on the size and shape of the regular reflection area, the total number of pixels is obtained by the pixel counting unit 293, and the average value is obtained by the luminance component averaging unit 38 from the data of the object and the number of pixels. , Are temporarily stored in the measurement data memory 40A.

(4) The correction circuit 42 corrects the data of the object from the measurement data memory 40A using the data of the reference plate from the illumination data memory 40W and stores the corrected data in the correction data memory.

According to this modification, the specific region detection unit 290 and the non-measurement region setting unit 291 are provided in the measurement region extraction unit to automatically extract and measure the peripheral part of the specular reflection part of the object. Even when the measurement area of the object is a sphere or a shape close to it, the intensity of the reflected light is stable, good measurement is possible, and the measurement area can be automatically set, making measurement easy and reducing the time required for measurement. can do.

(Fifteenth Embodiment) Next, in each of the above embodiments, a fifteenth embodiment for speeding up the classification operation will be described with reference to FIGS. 35, 36, and 37.

In a color classification device using a plurality of bandpass filters, there is a possibility that some of the bandpass filters having little influence on the classification may appear depending on an object to be measured, but such bandpass filter data is omitted. Thus, the calculation speed can be increased.

FIG. 35A is a diagram showing a configuration of the classification operation circuit of the present embodiment.

The classification operation circuit 28 of the present embodiment includes a luminance component extraction unit 30, a classification operation unit 32, and a classification determination unit 34. The classification operation unit 32 includes a switch 310, a classification spectrum calculation unit 311 and a fisher ( It is composed of a Fisher ratio calculation unit 312, a classification spectrum inspection unit 313, a d1 memory 314 and a d2 memory 315 for writing a classification spectrum, a switch 316, and an inner product calculation circuit 317.

FIG. 35 (B) shows a flow of processing for calculating a classification spectrum having as few dimensions as possible by removing data from unnecessary band-pass filters having little influence on classification among a plurality of band-pass filters.

Here, the operation of the classification operation circuit of this embodiment shown in FIG. 35A as a color classification device having five bandpass filters will be described with reference to FIG.

When data is input using five bandpass filters, the data is treated as five-dimensional data.

Here, assuming that a classification spectrum is created by FS conversion from the data of five-dimensional N samples in each class, it is also five-dimensional data, and the components of each dimension correspond to five bandpass filters.

FIG. 35B shows a method of calculating a classification spectrum in the learning mode.

In the learning mode, first, the switch 310 is set to the b side.

In the control circuit 26, the measurement data corrected by the reference plate data is set to read all five-dimensional data in both the class 1 and the class 2 (step 320), and the data of the class 1 and the class 2 are read from the luminance extracting unit 30 (step 320). Step 321).

The classification spectrum calculation unit 311 calculates a classification spectrum based on the read data (step 322), and the fisher ratio (Fisher ratio) calculation unit 312 uses the classification spectrum and the measurement data from the luminance component extraction unit 30 to change the classification spectrum. ) Is calculated (step 323).

The classification spectrum inspecting unit 313 checks whether or not the Fisher ratio (Fisher ratio) is larger than a predetermined value (Step 324). At the same time as writing, the control circuit is informed of the number of the band-pass filter used among the five filters (step 325). Then, the absolute value of the component corresponding to each band-pass filter of the classified spectrum is checked. Is obtained (step 326) and sent to the control circuit 26.

If the fisher ratio (Fisher ratio) is smaller than a predetermined value, the classification spectrum inspecting unit 313 does not write the classification spectrum at that time to the d1 memory 314 and the d2 memory 315, but stores the classification spectrum in the control circuit 26. A signal indicating the end of calculation is sent (step 328).

The control circuit 26 is set to read the filter data other than the filter number sent from the classification spectrum inspection unit 313 (step 327), reads the data again for the set filter, creates a classification spectrum, and (Fisher @ ratio) is obtained and compared with a predetermined value (steps 321 to 324).

That is, while the fisher ratio (Fisher @ ratio) is equal to or more than a predetermined value, the data of the filter having the smallest absolute value of the classification spectrum is further omitted. Exit without saving.

On the other hand, in the classification mode, by controlling the control circuit to input only the necessary data from the band-pass filter, the amount of data handled for both input and classification operations is reduced, so that high-speed classification can be determined.

(4) When reducing the data of the band-pass filter, deleting components having smaller classification spectrum components means deleting components having less difference between the two classes. Therefore, even if this is deleted, the classification result is not significantly affected.

In the fifteenth embodiment, a classification spectrum inspecting unit and a fisher ratio (Fisher ratio) calculating unit are provided, and a minimum band pass required in a range where the fisher ratio (Fisher ratio) in the obtained classification spectrum does not fall below a predetermined value. Since the configuration is such that data from the filter is obtained, the number of dimensions of the classification spectrum used for the classification operation can be reduced to the minimum without lowering the classification accuracy, so that higher-speed classification judgment can be performed in the classification mode.

In the present embodiment, the components of the classified spectrum are deleted from the data of the band-pass filter corresponding to the component having the small absolute value. However, the data having the small absolute value of the component of the difference spectrum may be deleted. .

FIG. 36 shows a modification of the present embodiment. The classification calculation circuit 28 shown in FIG. 36A adds a difference spectrum calculation unit 331 to the classification calculation unit 32 shown in FIG. A difference spectrum inspection unit 332 is provided in place of the unit 313. The flowchart shown in FIG. 36B has a step 341 for obtaining two classes of difference spectra instead of the step 326 in FIG. The step 342 for obtaining the number of the band-pass filter having the smallest absolute value by examining the component of is applied.

Thus, in the difference spectrum between the two classes, the one having a value close to 0 is a component having a small difference between the classes, and can be omitted.

Similarly, a spectrum component whose correction is not close to 0 may be deleted.

FIG. 37 is also a modification of the present embodiment, and the classification operation circuit 28 shown in FIG. 37 (A) is provided with a spectrum inspection unit 351 instead of the classification spectrum inspection unit 313 of the classification operation unit 32 of FIG. The output of the luminance data extraction unit 30 is connected to the output of a not-shown measurement data memory (memory in which data before correction with reference plate data is stored), and the flowchart shown in FIG. In this embodiment, a step 361 for finding the number of the band-pass filter having the minimum spectrum component before correction is applied instead of the step 326 in FIG.

Since most of the components of the spectrum before the correction are close to 0 are noises of the image sensor, even if they are deleted, the classification result is hardly affected. Therefore, the range in which the fisher ratio (Fisher ratio) is maintained at a predetermined value or more. And the number of dimensions of the classification spectrum (ie, the number of bandpass filters) can be reduced.

In the present embodiment, the data of the band-pass filter corresponding to the one having the small absolute value of the component of the classified spectrum is deleted, and as a modified example, the one having the small absolute value of the component of the difference spectrum or the component of the spectrum before correction is used. Although the method of deleting the data of the band-pass filter corresponding to the small one has been described, not only each of these is performed separately, but also the dimension of the classification spectrum (that is, the number of band-pass filters) is considered in consideration of all of them. Even if the number is reduced, it is possible to speed up the classification operation without lowering the classification accuracy.

<< (Sixteenth Embodiment) >> Next, a sixteenth embodiment capable of more accurately classifying various objects will be described with reference to FIG.

{The color classification device of this embodiment will be described with reference to FIG.

The color classification apparatus according to the present embodiment includes a housing 1, an optical system 10, an imaging device 14 on which reflected light from an object is imaged by the optical system, and five imaging devices between the optical system 10 and the imaging device 14. , A motor 24 for driving the rotary color filter 12, a motor drive circuit 124, and a filter for detecting the rotational position of the rotary color filter 12. A position sensor 123, an image sensor driving circuit 22, an A / D converter 16 for converting a video signal from the image sensor 14 into a digital signal via an amplifier 15, and a frame memory 18 for storing the digitized video signal. An image processing circuit 371 for performing various processes on the image data stored in the frame memory 18 and outputting image feature values; A classification operation circuit 28 for classifying the object from the data from the road 371, and a control circuit 26 for controlling these.

The classification operation circuit 28 of the present embodiment includes a luminance component extraction unit 30, a spectrum expansion unit 372, a classification operation unit 32, and a classification determination unit 34 as shown in FIG.

The control circuit 26 sends a signal to the motor drive circuit 124 so that the rotary color filter 12 synchronizes with the exposure of the image sensor 14 while monitoring the output of the filter position sensor 123.

The image of the object formed on the image sensor 14 is photoelectrically converted and output as a video signal, and then converted to digital by an A / D converter 16 via an amplifier 15 and temporarily stored in a frame memory 18. You.

(4) The image signal stored in the frame memory 18 is subjected to various kinds of processing on the image in the measurement area by the image processing circuit 371, and the feature amount of the image is extracted and sent to the classification operation circuit 28.

As the feature value of the image, it is possible to use a differential value, a spatial frequency component, a coefficient of orthogonal transformation, each feature value obtained using a co-occurrence matrix, a density histogram, a line or an outline, a correlation value with a specific pattern, and the like. it can.

An image signal from the frame memory 18 is input to the classification operation circuit 28, and an average value in the measurement area is obtained for each bandpass filter (five) by the luminance component extraction unit 30 to obtain five-dimensional spectrum data. .

At this time, the illumination light is corrected and normalized as necessary.

The next-stage spectrum expansion unit 372 combines the five-dimensional spectrum data obtained from the five band-pass filters with the P-dimensional data that is the feature amount of the image obtained from the image processing circuit 371, and expands the data by 5 + P dimensions. The obtained spectrum is transferred to the classification calculation unit 32.

In the learning mode, the classification calculation unit 32 creates a 5 + P-dimensional classification spectrum from the data of the number of samples N for each of the two classes transferred as described above, and sends it to the classification determination unit 34 together with the data.

At this time, if necessary, the data is normalized so that the weight of the data becomes a predetermined value between the feature amount of the image obtained from the spectrum image processing circuit obtained from the band-pass filter.

The classification determination unit 34 determines and stores a classification boundary.

In the classification mode, the classification operation unit 32 performs an inner product operation of the 5 + P-dimensional data of the object transferred as described above and the classification spectrum obtained in the learning mode, and sends the result to the classification determination circuit 34. .

The classification judgment circuit 34 outputs the classification judgment result of the object based on the classification boundaries stored in the learning mode.

According to the sixteenth embodiment, since the image processing circuit 371 is provided at the subsequent stage of the frame memory 18 and the spectrum expansion unit 372 is provided in the classification operation circuit 28, not only the color of the object but also the surface Since the shape and texture are reflected in the classification spectrum, more accurate and favorable measurement can be performed.

<< (Seventeenth Embodiment) >> Next, a seventeenth embodiment in which the classification result can be visually confirmed on the monitor display will be described with reference to FIG.

{The color classification device of this embodiment will be described with reference to FIG.

The color classification apparatus according to the present embodiment includes a housing 1, an optical system 10, an imaging device 14 on which reflected light from an object is imaged by the optical system, and five imaging devices between the optical system 10 and the imaging device 14. , A motor 24 for driving the rotary color filter 12, a motor drive circuit 124, and a filter for detecting the rotational position of the rotary color filter 12. A position sensor 123, an image sensor driving circuit 22, an A / D converter 16 for converting a video signal from the image sensor 14 into a digital signal via an amplifier 15, and a frame for temporarily storing the digitized video signal A memory 18, a color synthesizing circuit 381 for synthesizing image data to be output to a monitor from image data stored in the frame memory 18, and a frame memory for monitor output And 82, a classification operation circuit 28 for classifying the object from the data from the frame memory 18, and a control circuit 26 for controlling these.

FIG. 39B shows the configurations of the frame memory 18, the color synthesis circuit 381, and the frame memory 382 of this embodiment.

The frame memory 18 includes a BPF1 plane 401, a BPF2 plane 402, a BPF3 plane 403, a BPF4 plane 404, and a BPF5 plane 405 corresponding to each bandpass filter.

The color synthesizing circuit 381 stores a coefficient memory 420 storing coefficients (BPF coefficients 421 to 425) for multiplying outputs of the BPF1 planes 401 to 405 of the frame memory 18, and outputs and coefficient memories of the BPF planes 401 to 405. It comprises multipliers 411 to 415 each having a corresponding set of BPF coefficients 421 to 425 stored in 420 as inputs, and an adder 410 for adding outputs of the multipliers 411 to 415.

The frame memory 382 is a frame memory for outputting to the TV monitor, and includes RGB planes 431 to 433, and a video signal generation circuit 434 for reading data from the memory and converting the data into a video signal that can be input to the TV monitor.

Next, the operation of the color classification device of the present embodiment will be described.

The control circuit 26 sends a signal to the motor drive circuit 124 so that the rotary color filter 12 synchronizes with the exposure of the image sensor 14 while monitoring the output of the filter position sensor 123.

The image of the object formed on the image sensor 14 is photoelectrically converted and output as a video signal, and then temporarily stored in a frame memory 18 as image data converted digitally by an A / D converter 16 via an amplifier 15. Is stored in

The image data from the frame memory 18 is subjected to data conversion for output to a TV monitor by a color synthesizing circuit 381, written to a frame memory 382 for displaying a TV monitor, and input to the TV monitor by a video signal generating circuit 434. It is converted to a possible video signal and output to the TV monitor.

On the other hand, the image signal from the frame memory 18 is input to the classification operation circuit 28, and the average value in the measurement area is obtained for each bandpass filter (five) by the luminance component extraction unit 30, and the five-dimensional spectrum data is obtained. And

At this time, the illumination light is corrected and normalized as necessary.

In the learning mode, the classification operation unit 32 creates a five-dimensional classification spectrum from the data of the number of samples N for each of the two classes transferred as described above, and sends it to the classification determination unit 34 together with the data.

(4) The classification determination unit 34 determines and stores classification boundaries.

In the classification mode, the classification operation unit 32 performs an inner product operation of the five-dimensional data of the object transferred as described above and the classification spectrum obtained in the learning mode, and sends the result to the classification determination circuit 34. .

The classification judgment circuit 34 outputs the classification judgment result of the object based on the classification boundaries stored in the learning mode.

Next, the operation of the color synthesizing circuit 381, which is a feature of the present embodiment, will be described in more detail.

The data of the five bandpass filters are sequentially stored in the BPF planes 401 to 405 of the frame memory 18.

The coefficient memory 420 stores coefficients (BPF coefficients 421 to 425) for multiplying the data of each BPF plane.

BPF coefficients 421 to 425 are multiplied by data of the corresponding BPF plane by multipliers 411 to 415.

The data of each plane multiplied by the coefficients in the multipliers 411 to 415 are added in the adder 410, and the data is added to the plane designated by the control circuit 26 among the R plane 431, the G plane 432, and the B plane 433 of the frame memory 382. Written.

(4) The data of each plane is converted by the video signal generation circuit 434 into a signal that can be input to the TV monitor and output to the TV monitor.

In the learning mode, a coefficient obtained by approximating the spectral characteristic of the R component of the monitor from the spectral transmission characteristic of each bandpass filter is written in the coefficient memory 420, and the data of each plane is multiplied by the multipliers 411 to 415 and then added by the adder 410. The added data is written to the R plane of the frame memory 382.

Next, the coefficient approximating the spectral characteristic of the G component of the monitor is written to the coefficient memory 420, and the output of the adder 410 is written to the G plane 432.

Next, the coefficient approximating the spectral characteristic of the B component of the monitor is written to the coefficient memory 420, and the output of the adder 410 is written to the B plane 433.

(4) The data of each plane is converted by the video signal generation circuit 434 into a signal that can be input to the TV monitor and output to the TV monitor.

As a result, the object being measured is displayed on the monitor with good color reproducibility.

In the classification mode, the data of the classification spectrum is transmitted from the classification calculation circuit 28 to the control circuit 26, and the control circuit 26 normalizes the components corresponding to each band-pass filter of the classification spectrum to 0 to 1 and writes them to the coefficient memory 420. The output of the adder 410 is written to the R plane 431.

Next, the coefficient obtained by subtracting the coefficient written in the coefficient memory from 1 from 1 is written in the coefficient memory 420.

The output of the adder 410 is written to the G plane 432.

The data of each plane is converted into a signal that can be input to the TV monitor by the video signal generation circuit 434 and output from the TV monitor.

(5) Thus, on the monitor, for example, an image of an object classified into class 1 is displayed in red, an object classified in class 2 is displayed in green, and an intermediate object is displayed in yellow.

According to the seventeenth embodiment, the frame memory 382 for TV monitor display is provided with a plane corresponding to each band-pass filter inside the frame memory 18, and the color synthesis circuit 381 assigns a coefficient to the data of each plane. Is written in each plane of the frame memory, the classification result can be displayed in a different color on the monitor, and the result can be easily visually confirmed.

Further, in the present embodiment, in the learning mode, color display like a normal TV image is also possible, and the confirmation of the object can be performed naturally.

In the present embodiment, the R plane 431 and the G plane 432 of the frame memory 382 are used in the classification mode. However, the present invention is not limited to this, and it goes without saying that the same effect can be obtained by using any plane. Absent.

FIG. 1 is a block diagram showing a color classification device according to a first embodiment of the present invention. FIG. 2 is a schematic diagram of a rotating color filter used in the first embodiment. FIG. 3 is a flowchart illustrating the operation of the first embodiment. FIG. 4 is a block diagram illustrating a color classification device according to a second embodiment of the present invention. FIG. 5 is a schematic view of a rotary color stop plate used in the second embodiment. FIG. 6 is a block diagram illustrating a color classification device according to a third embodiment of the present invention. FIG. 7 is a schematic diagram of a rotating ND filter used in the third embodiment. FIG. 8 is a schematic view of a rotary slit plate used in the third embodiment. FIG. 9 is a block diagram illustrating a color classification device according to a fourth embodiment of the present invention. FIG. 10 is a flowchart illustrating the operation of the fourth embodiment. FIG. 11 is a block diagram illustrating a color classification device that is a modification of the fourth embodiment. FIG. 12 is a block diagram illustrating a color classification device according to a fifth embodiment of the present invention. FIG. 13 is a schematic view of a reference plate according to a fifth embodiment. FIG. 14 is a block diagram illustrating a color classification device according to a sixth embodiment of the present invention. FIG. 15 is a diagram illustrating an example of an image formed on a CCD according to the sixth embodiment. FIG. 16 is a schematic view of a reference plate according to a sixth embodiment. FIG. 17 is a block diagram illustrating a color classification device according to a seventh embodiment of the present invention. FIG. 18 is a block diagram illustrating a color classification device according to an eighth embodiment of the present invention. FIG. 19 is a view showing a connection between a CCD and an optical fiber according to the eighth embodiment. FIG. 20 is a layout diagram of a bandpass filter on a CCD according to the eighth embodiment. FIG. 21 is a view for explaining how to bundle optical fibers according to the eighth embodiment. FIG. 22 is a diagram illustrating an example of an image formed on a CCD according to the eighth embodiment. FIG. 23 is a diagram illustrating a modification of the eighth embodiment. FIG. 24 is a view for explaining another modification of the eighth embodiment. FIG. 25 is a block diagram illustrating a color classification device according to a ninth embodiment of the present invention. FIG. 26 is a diagram illustrating a configuration of an optical path splitting prism used in the ninth embodiment. FIG. 27 is a view for explaining a combined state of an optical path splitting prism and a CCD according to the ninth embodiment. FIG. 28 is a layout diagram of a bandpass filter on a CCD according to a ninth embodiment. FIG. 29 is a diagram illustrating a configuration of an illumination sensor used in the ninth embodiment. FIG. 30 is a block diagram illustrating a color classification device according to a tenth embodiment of the present invention. FIG. 31A is a block diagram illustrating a classification operation circuit according to an eleventh embodiment of the present invention, (B) is a block diagram illustrating a classification operation unit of the eleventh embodiment, and (C) is an eleventh embodiment. FIG. 2 is a block diagram showing a classification spectrum selection circuit of FIG. FIG. 32A is a diagram showing the distribution of learning data according to the conventional method, (B) is a diagram showing the distribution of learning data according to the twelfth embodiment of the present invention, and (C) is the diagram of the twelfth embodiment. FIG. 21 is a block diagram showing a luminance extracting unit, and (D) is a block diagram showing a classification calculating unit of the twelfth embodiment. FIG. 33 (A) is a diagram showing the distribution of learning data according to the conventional method, (B) is a diagram showing the distribution of learning data according to the thirteenth embodiment of the present invention, and (C) is the diagram of the thirteenth embodiment. FIG. 3 is a block diagram illustrating a classification determination unit. FIG. 34A is a block diagram illustrating a classification operation circuit according to a fourteenth embodiment of the present invention, (B) is a diagram illustrating a measurement target for explaining the operation of the fourteenth embodiment, and (C) is a diagram illustrating a measurement target. And FIG. 14D is a block diagram illustrating a shape of a measurement region according to a modification of the fourteenth embodiment. FIG. 35A is a block diagram illustrating a classification operation circuit according to a fifteenth embodiment of the present invention, and (B) is a flowchart illustrating the operation of the fifteenth embodiment. FIG. 36A is a block diagram illustrating a classification operation circuit according to a first modification of the fifteenth embodiment. FIG. 36B is a flowchart illustrating the operation of the first modification of the fifteenth embodiment. FIG. 37A is a block diagram showing a classification operation circuit according to a second modification of the fifteenth embodiment, and (B) is a flowchart for explaining the operation of the second modification of the fifteenth embodiment. FIG. 38A is a block diagram illustrating a color classification device according to a sixteenth embodiment of the present invention, and (B) is a block diagram illustrating a classification calculation circuit according to the sixteenth embodiment. FIG. 39A is a block diagram illustrating a color classification device according to a seventeenth embodiment of the present invention, and (B) is a block diagram illustrating a color combining circuit according to the seventeenth embodiment. FIG. 40 is a diagram illustrating classification boundaries. FIG. 41 is a diagram showing a configuration of a conventional color identification device. FIG. 42 is a diagram illustrating an example of the classification spectrum.

Explanation of reference numerals

1 ... housing,
10 optical system,
101 ... Aperture,
126 ... diaphragm control circuit,
12A to 12E: bandpass filters,
12 ... rotating color filter,
123 ... Filter position sensor,
24 ... motor,
124 ... motor drive circuit,
14 ... image sensor,
15 ... amplifier,
16 ... A / D converter,
18 ... frame memory,
28 Classification arithmetic circuit,
29: Exposure value memory,
26 ... Control circuit,
22 ... Imaging element drive circuit.

Claims (32)

Imaging means for imaging light from the object;
Optical means for forming an image of light from the object on the imaging means,
A plurality of bandpass filters that pass bands of different light wavelengths disposed between the object and the imaging unit,
Classification means for calculating a classification spectrum for classification using a statistical method from the spectral spectrum of the object imaged by the imaging means, and classifying the object using this classification spectrum,
Exposure condition control means for controlling exposure conditions at the time of imaging of each of the plurality of band-pass filters,
With
The color classification apparatus according to claim 1, wherein the exposure condition control unit includes an exposure amount limiting unit that limits an exposure amount corresponding to the plurality of bandpass filters. 2. The color classification apparatus according to claim 1, wherein the exposure amount limiting unit is configured to be detachable and replaceable. The color classification apparatus according to claim 1, wherein the exposure condition control unit includes an exposure condition storage unit that stores exposure conditions at the time of imaging in the plurality of bandpass filters. 4. The color classification apparatus according to claim 3, wherein the exposure condition control unit includes an exposure condition correction unit configured to correct data captured based on the exposure condition. 5. 5. The exposure amount limiting unit includes a plurality of apertures corresponding to the plurality of bandpass filters, and includes an aperture plate that can be switched in synchronization with the plurality of bandpass filters. The color classification device according to any one of the above. The exposure amount limiting unit includes a plurality of filters having different transmittances corresponding to the plurality of bandpass filters, and includes a filter unit that can switch in synchronization with the plurality of bandpass filters. Item 5. The color classification device according to any one of Items 1 to 4. The exposure amount limiting unit includes a light shielding plate having a plurality of non-light shielding portions having different sizes corresponding to the plurality of band pass filters, and includes a light shielding unit that can be switched in synchronization with the plurality of band pass filters. The color classification device according to any one of claims 1 to 4, wherein: Imaging means for imaging light from the object;
Optical means for forming an image of light from the object on the imaging means,
A plurality of bandpass filters that pass bands of different light wavelengths disposed between the object and the imaging unit,
Classification means for calculating a classification spectrum for classification using a statistical method from the spectral spectrum of the object imaged by the imaging means, and classifying the object using this classification spectrum,
Exposure condition control means for controlling exposure conditions at the time of imaging of each of the plurality of band-pass filters,
With
The color classification apparatus according to claim 1, wherein the exposure condition control unit includes an exposure condition storage unit that stores an exposure condition at the time of imaging of each of the plurality of bandpass filters. 9. The color classification apparatus according to claim 8, wherein the exposure condition control unit includes an exposure condition correction unit that corrects data captured based on the exposure condition. Imaging means for imaging light from the object;
Optical means for forming an image of light from the object on the imaging means,
A band-variable bandpass filter that is arranged between the object and the imaging unit and that can change a wavelength band of light to be passed,
Classification means for calculating a classification spectrum for classification using a statistical method from the spectral spectrum of the object imaged by the imaging means, and classifying the object using this classification spectrum,
With
A color classification apparatus comprising: exposure condition control means for separately controlling exposure conditions at the time of imaging of passband characteristics of different wavelengths set in the band variable bandpass filter. 11. The color classification apparatus according to claim 10, wherein the exposure condition control unit includes an exposure condition storage unit that stores an exposure condition at the time of imaging in a passband characteristic of each wavelength of the band variable bandpass filter. . 12. The color classification apparatus according to claim 11, wherein the exposure condition control unit includes an exposure condition correction unit that corrects data captured based on the exposure condition. 13. The exposure condition control unit according to claim 8, further comprising an aperture control unit that controls an aperture adjustment unit of the imaging unit based on data of the exposure condition storage unit. The color classification device according to 1. The exposure condition control means,
A transmittance variable filter that is arranged between the object and the imaging unit and that can change transmittance;
At the time of each imaging of the plurality of band-pass filters, transmittance control means for controlling the transmittance of the transmittance variable filter based on the data of the exposure condition storage means,
The color classification device according to any one of claims 8, 9, 11, and 12, wherein: The imaging unit is capable of controlling an exposure time by a driving pulse,
9. The image forming apparatus according to claim 8, wherein the exposure condition control unit includes a unit that controls an exposure time of the imaging unit based on data of the exposure condition storage unit when each of the plurality of bandpass filters is imaged. 13. The color classification device according to any one of 9, 11, and 12. Imaging means for imaging light from the object;
Optical means for forming an image of light from the object on the imaging means,
A plurality of bandpass filters that pass bands of different light wavelengths disposed between the object and the imaging unit,
Classification means for calculating a classification spectrum for classification using a statistical method from the spectral spectrum of the object imaged by the imaging means, and classifying the object using this classification spectrum,
Normalizing means for normalizing at least one of the spectral characteristic and the luminance of the light source illuminating the object,
With
The color classification device, wherein the normalizing means includes a light source observing means for obtaining data of the light source. Imaging means for imaging light from the object;
Optical means for forming an image of light from the object on the imaging means,
A band-variable band-pass filter arranged between the object and the imaging means, which can change a band of a wavelength of light to be passed,
Classification means for calculating a classification spectrum for classification using a statistical method from the spectral spectrum of the object imaged by the imaging means, and classifying the object using this classification spectrum,
Normalizing means for normalizing at least one of the spectral characteristic and the luminance of the light source illuminating the object,
With
The color classification device, wherein the normalizing means includes a light source observing means for obtaining data of the light source. The light source observation means,
A reference plate that reflects or transmits the light of the light source and scatters the light,
Second optical means for guiding the reflected light of the reference plate to the imaging means,
The color classification device according to claim 16, further comprising: The reference plate is provided on an upper part of the housing,
The second optical means includes:
An optical system for imaging the reference plate on the imaging unit,
A mirror for guiding the optical axis of the optical system to the imaging means,
With
19. The color classification device according to claim 18, wherein the position of the mirror is displaceable. The reference plate is provided on an upper part of the housing,
19. The color classification apparatus according to claim 18, wherein the second optical unit has a mirror for guiding a light image of the reference plate to a field of view of the imaging unit. The light source observation means,
A reference plate arranged at the front of the housing and irregularly reflecting light from the light source;
An optical sensor for detecting the intensity of the reflected light from the reference plate,
A plurality of bandpass filters disposed between the reference plate and the optical sensor,
Optical means for imaging the reflected light from the reference plate on the optical sensor,
17. The color classification device according to claim 16, comprising: 22. The plurality of bandpass filters arranged between the reference plate and the optical sensor also serve as a plurality of bandpass filters arranged between the object and the imaging unit. Color classification device. 22. The color classification device according to claim 21, wherein the optical sensor uses an image sensor. 24. The color classification apparatus according to claim 19, wherein the light source observation unit includes a light source direction detection unit that detects a direction of a light source. The light source direction detection means,
Provided on the reference plate, according to a change in the direction of the light source, a rod-shaped or plate-shaped member so that the shape or position of the shadow changes,
Processing the image data of the reference plate obtained by the imaging means, the direction of the light source, the illumination condition detection means for obtaining the intensity and the spectral spectrum,
The color classification device according to claim 24, comprising: 19. The color classification device according to claim 18, wherein the second optical unit is formed of an optical fiber. 27. The color classification apparatus according to claim 26, wherein the reference plate can be used separately from a housing. 17. The color classification apparatus according to claim 15, wherein the light source observation unit includes a light source spectral sensor unit that detects a spectrum of light from the light source. The light source spectral sensor means,
A transmittance variable filter capable of changing the transmittance,
A plurality of bandpass filters having the same characteristics as the plurality of bandpass filters;
As many optical sensors as the number of bandpass filters for detecting the intensity of light transmitted through each bandpass filter,
A transmittance control circuit that monitors the outputs of these optical sensors and controls the transmittance of the transmittance variable filter;
An illumination condition detection circuit that receives signals from the plurality of optical sensors and the transmittance control circuit, and outputs brightness and a spectrum of a light source,
The color classification device according to claim 28, comprising: 29. The color classification apparatus according to claim 28, wherein the light source spectral sensor means can be used separately from a housing of a main body. Imaging means for capturing an image composed of light from an object,
A band-variable band-pass filter arranged between the object and the imaging unit, capable of changing a pass band characteristic of a wavelength,
In a plurality of classes defined for the classification of the object, using the spectral spectra of a plurality of types of first objects, each of which is a known object, each of the classes is a classification spectrum for the classification. Classification spectrum calculating means for calculating
Means for defining a space for performing the classification of the object in each of the classes using the classification spectrum,
The characteristic of the band variable bandpass filter is changed a plurality of times, and the corresponding class is obtained from a plurality of images obtained by imaging the second object, which is the unknown object, in a plurality of bands. Means for projecting the spectral data of the object into the space;
A classification unit configured to classify the second object in each of the classes based on the result of the projection. Imaging means for capturing an image composed of light from an object,
A plurality of bandpass filters, each having a different passband characteristic of wavelength, disposed between the object and the imaging unit,
In at least three classes defined for the classification of the object, each spectral spectrum of two first objects, which are the objects that are known to correspond to any two classes of the classes, respectively. Using, a classification spectrum calculation means for calculating a classification spectrum for the classification,
Means for defining a space for performing the classification of the object in each of the classes using the classification spectrum,
Using each of the bandpass filters, the spectral spectrum of the second object obtained from a plurality of images obtained by imaging the second object whose class is the unknown object. Means for projecting the data of
A classification unit configured to classify the second object in each of the classes based on the result of the projection.

Images