JPH04119411A - Joint conversion correlator - Google Patents

Abstract

PURPOSE:To simultaneously use many reference pictures to perform pattern recognition by changing the transmittance or the reflectivity of a part corresponding to each reference picture in linear or nonlinear relations to a correlation coefficient. CONSTITUTION:A coherent picture to be correlated and a coherent reference picture are simultaneously inputted to a means 4 for obtaining the joint Fourier transform and are subjected to joint Fourier transform, and the result is inputted to a space optical modulator 5, and the intensity distribution picture is displayed on the space optical modulator 5. This intensity distribution picture is read out by the coherent light from a second coherent light source 6 and is subjected to Fourier transform and is converted to a correlation signal by a means 7 which subjects the internsity distribution picture to Fourier transform to convert it to the correlate signal. In such a case, an image projection optical system can be built in as the preprocessing for inputting an input image, and the input intensity of a reference picture 3 having the same space frequency component as a picture 2 to be correlated is relatively increased. Thus, the number of reference pictures to be used at a time is increased.

Description

[Detailed Description of the Invention] [Field of Industrial Application] The present invention is applicable to CCD in the field of optical information processing and optical measurement.
For two-dimensional images obtained from imaging devices such as cameras,
The present invention relates to a device that automatically performs pattern recognition and measurement by performing optical correlation processing using coherent light.

[Summary of the invention]

The present invention utilizes a joint transform correlator (J
oint Transform Correlator
), as a means for converting at least one reference image containing a desired target into a cohesin 1-image, the newly input at least one correlated image and a plane wave having a predetermined light intensity and containing no image information are used. By using means for irradiating the reference image with a coherent image consisting of an image superimposed with the correlated image, the input intensity of the reference image having the same spatial frequency component as the correlated image is relatively increased, and By substantially increasing the spatial frequency of a reference image that has spatial frequency components similar to but different from the spatial frequency components of the correlated image, the correlation between the correlated image and the similar but different reference image is improved. By reducing the number of patterns, pattern recognition ability is improved. In addition, each two-dimensional correlation coefficient between at least one reference image and at least one correlated image is input to a linear or nonlinear transfer function, and each correlation coefficient is changed according to the output. By substantially changing the light intensity transmitted through each corresponding reference image, the light intensity of the correlation peak representing the correlation coefficient decreases and noise increases when the number of correlated images or reference images is changed. This prevents unrecognizability or erroneous recognition caused by such factors, and enables high-speed and accurate pattern recognition and measurement.

[Conventional technology]

Joint transform correlators have been used in conventional pattern recognition devices and correlation processing devices.
FIG. 4 shows an example using an optical writing type spatial light modulator in the rm Correlator. In this method, an image in which a reference image to be recognized and a correlated image are placed adjacent to each other at the same time is used as an input image on a liquid crystal television 28.
to be displayed. After the light beam emitted from the laser beam a10 is expanded to a predetermined beam diameter by the beam expander 11,
The beam splitter 12 splits the beam into two beams. The light beam transmitted through the beam splitter 12 illuminates the input image displayed on the liquid crystal television 28, converting the input image into a coherent image. This coherent image is Fourier transformed using the first Fourier transform lens 15, and an optical writing base liquid crystal light valve 29 is placed on the Fourier transform surface.
The light intensity distribution image of the join I-Fourier transform image of the reference image and the correlated image is written in .

Next, the light beam reflected by the beam block 12 is successively reflected by the first mirror 17, the second mirror 18, and the polarizing beam splitter 19, and illuminates the readout surface of the optical writing type liquid crystal light valve 29, where it is written. The light intensity distribution image of the joint Fourier transform image is converted into a coherent image.

This coherent image is read out as a negative image or a positive image by passing through a polarizing beam splitter 19 that acts as an analyzer, and is Fourier transformed by a second Fourier transform lens 20, and a CCD camera 30 disposed on the transformation surface. The light is received by In this way, a correlation peak representing a two-dimensional correlation coefficient between the reference image and the correlated image can be obtained.

The above example uses a reflective optically writable liquid crystal light valve, which is an optically writable spatial light modulator, but when used as a transmissive type, BSO crystal (Bi+zSiOz) is used.

Optical writing type spatial light modulators using optically written crystals (crystals) are well known.Also, instead of using optical writing type spatial light modulators, a joint Fourier transform image is captured with a COD camera and There is also a method of displaying on an electric writing type spatial light modulator such as a liquid crystal television.

FIG. 5 shows an example of an input image displayed on the liquid crystal television 28, and FIG. 6 shows a pair of correlation peaks representing the two-dimensional correlation coefficient between the reference image and the correlated image obtained from the CCD camera 30. An example is shown. As shown in FIG. 5, the input image consists of a transparent part 31 forming an image and a light shielding part 32 other than that,
A reference image serving as a reference for recognition and a correlated image serving as a recognition target are arranged adjacent to each other at the same time. Furthermore, if the reference image and the correlated image shown in FIG. 5 are placed a distance apart from each other, the positions of the pair of correlation peaks shown in FIG. 6 will appear 2b apart from each other. do. In this way, the position where the correlation peak appears is uniquely determined by the positions where the reference image and the correlated image are placed.

Furthermore, a join 1-transform correlator has been proposed that has a foot-hack system that can increase the number of reference images that can be used at once in the conventional joint transform correlator and enables high-speed pattern recognition.

Usually, in the feedback method, in the first step,
The light intensity of each correlation peak representing a two-dimensional correlation coefficient between each reference image and the correlated image is measured. In the second step, the light intensity of each obtained correlation peak is normalized by the maximum light intensity among all correlation peak intensities. In steps 3 and 7, the reference image is placed in front of or behind the reference image such that the light intensity transmitted through each corresponding reference image changes substantially in linear or non-linear relation to its normalized value. Drive the placed mask spatial light modulator. In the above method, the correlation coefficient is fed back to the light intensity transmitted through the corresponding reference image, and pattern recognition is performed by repeating this feedback many times.

[Problem to be solved by the invention]

However, even in the conventional joint transform correlator or the joint transform correlator having a feedback system, the number of reference images used at each time cannot be said to be sufficiently large. In other words, even when an optical writing type ferroelectric liquid crystal light valve, which is a high-resolution, high-contrast binary device, is used as a spatial light modulator that displays a joint Fourier transform image, it does not have a feed-hack system. A joint transform correlator has at most 11 correlators, and a joint transform correlator with a feedbank system has at most 1
This method has a problem in that only five reference images can be used at one time.

[Means to solve the problem]

In order to solve the above problems, the present invention provides means for converting at least one reference image including a desired target and at least one newly input correlated image into a coherent image; Fourier transforming the coherent image; means for obtaining a joint Fourier transform image of the reference image and the correlated image; means for converting the joint Fourier transform image into an intensity distribution image and displaying the intensity distribution image on a spatial light modulator; means for reading out an intensity distribution image displayed on a modulator using coherent light; Fourier transforming the read intensity distribution image to convert it into a correlation function; converting the correlation function into a correlation signal using a photodetector; In a joint transform correlator having means for converting at least one reference image containing a desired target and a new input at least one correlated image into a coherent image, the means for converting at least one newly input correlated image into a coherent image A joint transform correlator characterized by comprising means for irradiating the reference image with a coherent image consisting of a superimposed image of a correlated image and a plane wave having a predetermined light intensity and containing no image information, and means for processing the correlation signal to obtain two-dimensional correlation coefficients between the reference image and the correlated image; and at least one reference image including the desired target and at least one newly input object. A means for converting a correlated image into a coherent image or a masking spatial light modulator placed in front of or behind the reference image converts the transmittance or reflectance of a portion corresponding to each of the reference images into the correlation coefficient. The above-mentioned problem has been solved by providing a joint transform correlator having a feed-hack system comprising means for changing the relationship linearly or non-linearly.

[Effect]

By having the above configuration, the joint transform correlator of the present invention can incorporate a projection optical system as a preprocessing when inputting an input image, and can perform the same spatial frequency component as the spatial frequency component of the correlated image. It is possible to relatively increase the input intensity of a reference image having a frequency component, and to substantially increase the spatial frequency of a reference image having a spatial frequency component similar to but different from the spatial frequency component of the correlated image.
This makes it possible to reduce the correlation coefficient between the correlated image and a similar but different reference image, resulting in a joint transform correlator with high pattern recognition ability. This makes it possible to further increase the number of reference images that can be created.

〔Example〕

Embodiments of the joint transform correlator of the present invention will be described below with reference to the drawings. FIG. 1 is a block diagram showing the configuration of a joint transform correlator according to the present invention, in which 1 is a first coherent light source, 2 is a correlated image display means, and 3 is a block diagram showing the configuration of a joint transform correlator according to the present invention.
4 is a reference image display means, 4 is a means for obtaining a joint Fourier transform, 5 is a spatial light modulator, 6 is a second coherent light source, 7 is a means for Fourier transforming an intensity distribution image to convert it into a correlation signal, and 8 is a means for performing a Fourier transform on an intensity distribution image. The computer 9 is a mask spatial light modulator. In FIG. 1, there is provided a means for irradiating the reference image with a coherent image consisting of a superimposed image of at least one newly input correlated image and a plane wave having a predetermined light intensity and containing no image information. The means for converting at least one reference image containing the desired target with a first coherent light source 1 into a coherent I-image
The correlated image display means 2 and the reference It (image display means 3 and mask spatial light modulator 9), and the means for converting at least one newly manually correlated image into a coherent image are the first coherent light source. 1 and correlated image display means 2
The means for Fourier transforming the coherent image and obtaining a joint Fourier transform image of the reference image and the correlated image is means 4 for obtaining joint Fourier transform, and converts the joint Fourier transform image into an intensity distribution image. The means for converting the intensity distribution image into the spatial light modulator and displaying the intensity distribution image on the spatial light modulator is the spatial light modulator 5, the means GJ for reading out the intensity distribution image displayed on the spatial light modulator using coherent light, the second The means for Fourier transforming the read intensity distribution image into a correlation function and converting the correlation function into a correlation signal using a photodetector is a coheren I light source 6 of means 7 for converting the correlation signal into a correlation signal, processing the correlation signal,
Means for obtaining two-dimensional correlation coefficients between the reference image and the correlated image is a computer 8, and means for converting the reference image into a coherent image, or a means for converting the reference image into a coherent image, or a mask disposed in front of or behind the reference image. Spatial light modulation for mask 3! is a means for changing the transmittance or reflectance of the portion corresponding to each of the reference images in a linear or non-linear relationship with respect to the correlation coefficient using a spatial light modulator. ii9
It is.

The coherent light emitted from the first coherent light source 1 illuminates the correlated image displayed on the correlated image display means 2 via the computer 8, and converts the correlated image into a coherent image.

A portion of the coherent image of the correlated image is input to means 4 for obtaining a joint Fourier transform. on the other hand,
The remainder of the coherent image of the correlated image passes through the mask spatial light modulator 9 and is irradiated with each reference image displayed on the reference image display means 3 via the computer 8 to be combined with each reference image. The image obtained by the projection of is converted into a coherent reference image. At this time, each center of the coherent correlated image is located at the reference image display means 3.
The image is irradiated to match the center of the reference image displayed on the screen. At the same time, the reference image display means 3 is irradiated with coherent plane waves of a predetermined light intensity that do not include an image from the first coherent light source 1 in a superimposed manner. The light intensity of this plane wave is controlled by the correlated image display means 2, the mask spatial light modulator 9, and the reference image display means 3.

The coherent correlated image and the coherent reference image are simultaneously input to a means 4 for obtaining a joint Fourier transform, subjected to joint Fourier transform, and input to a spatial light modulator 5, which transmits the intensity distribution image to the spatial light modulator 5. indicate. The intensity distribution image is based on the second coherent light #6.
The intensity distribution image is read out by coherent light from the source, and is Fourier-transformed and converted into a correlation signal by means 7 for Fourier-transforming the intensity distribution image and converting it into a correlation signal. The converted correlation signal is taken into the computer 8, and a correlation coefficient corresponding to the reference image and the correlated image is calculated.

When converting the reference image into a coherent image, a mask spatial light modulator 9 is configured so that the intensity of the converted coherent I image corresponds to the correlation coefficient.
masking the coherent correlated image and the coherent plane wave illuminating the reference image. f
(x) is a predetermined increasing function with a domain [0,1] and a value range [0,1], and the correlation coefficient is C4(j-1, . . .
nun is a natural number), the input intensity to the means 4 for obtaining the joint Fourier transform of the reference image is f(Cj)
Thus, the spatial light modulator 9 for masking masks the coherent correlated image and the coherent I/plane wave that illuminate the reference image. Further, in the reference image display means 3, in a state in which only a plane wave having a predetermined light intensity and containing no image information is irradiated, the transmitted light amount of the display portion of the reference image is calculated from the portion corresponding to the reference image. change according to the area ratio or amount of transmitted light,
These Fourier transform intensities are normalized to be equal.

Next, a first embodiment of the join I/transform correlator of the present invention will be described. FIG. 2 is a block diagram of one embodiment of the joint transform correlator according to the present invention, in which 10 is a laser light source, I
I is a beam expander, 12 is a beam splitter, 1
3 is the first LCD TV, 14 is the second LCD TV, 15
is a first Fourier transform lens, 16 is an optical writing type spatial light modulator, 17 is a first mirror, 18 is a second mirror,
19 is a polarizing beam splinter, 20 is a second Fourier transform lens, 21 is a photodiode array, 22 is a computer, 23 is a first D/A converter, and 24 is a second D/A converter.
/A converter. In FIG. 2, means for converting at least one reference image including a desired target and at least one newly input correlated image into coherent images include a laser light source 10, a beam expander 11, and a first liquid crystal television 13. and a second LCD television 14 and a computer 2
2, the first D/A converter 23, and the second D/A converter 24
That is. Further, means for Fourier transforming the coherent image to obtain a joint Fourier transforming image of the reference image and the correlated image is a first Fourier transforming lens 15, which transforms the joint Fourier transforming image into an intensity distribution image. The means for displaying the intensity distribution image on the spatial light modulator is the optical writing type spatial light modulator 16. Further, the means for reading out the intensity distribution image displayed on the spatial light modulator using coherent light includes a laser light source 10, a beam expander 11, a beam splitter 12, a first mirror 17, a second mirror 18, and a polarized light source. Beam splitter 1
9. The read-out intensity distribution image is Fourier-transformed into a correlation function, and the correlation function is converted into a correlation signal using a photodetector. It is 21. The means for converting at least one reference image containing a desired target into a coherent image includes superposition of the newly input at least one correlated image and a plane wave having a predetermined light intensity and containing no image information. The means for irradiating the reference image with coherent light consisting of an image is
They are a first liquid crystal television 13, a second liquid crystal television 14, a computer 22, a first D/A converter 23, and a second D/A converter 24.

A beam emitted from a laser light source 10 is expanded to a predetermined beam diameter by a beam expander 11, and then split into two beams by a beam splitter 12. The light flux that has passed through the BIMSUBURINOTA 12 converts the correlated image displayed on the first liquid crystal television 13 into a coherent image, and then illuminates the reference image displayed on the second liquid crystal television 14. An example of a correlated image displayed on the first liquid crystal television 13 at this time is shown in FIG. 7, and an example of a reference image displayed on the second liquid crystal television 14 is shown in FIG. As can be seen from FIG. 7 while referring to FIG. 8, the correlated images are arranged at the center and at positions around it corresponding to the same positions as the reference images in FIG. There is. Further, the reference images in FIG. 8 are arranged with the correlated image at the center and on the circumference around the correlated image. At this time, the centers of each correlated image in FIG. 7 and the reference image in FIG. 8 coincide with each other, and the orientations of these images are the same and the thicknesses are also the same.

Such an arrangement of correlated images and reference images is created by determining the coordinates of the reference image stored in the preliminary computer 22 and the newly input correlated image on the computer 22, and The image is displayed on the first liquid crystal television 13 and the second liquid crystal television 4 via the A converter 23 and the second D/A converter 24. However, in the example of the correlated images shown in FIG. 7, the correlated image located at the center may be omitted and left in a completely light-transmitting state.

The cohesin l-correlated image and coherent reference image thus obtained are transferred to the first Fourier transform lens 15.
The optical writing type spatial light modulator 16
These joints form a Fourier transform image on the writing surface of '2'. This joint Fourier transform image is displayed on the optical writing type spatial light modulator 16 as an intensity distribution image of the joint Fourier transform image.

On the other hand, the other beam reflected by the beam splitter 12 is successively reflected by the first mirror 17, the second mirror 18, and the polarizing beam splitter 19, and illuminates the readout surface of the optical writing spatial light modulator 16. Then, the intensity distribution image of the joint Fourier transformed image is read out. The intensity distribution image of the read joint Fourier transform image is Fourier transformed by the second Fourier transform lens 20, and the correlation function of the reference image and the newly input correlated image is displayed on the photodiode array 21 surface. form a light intensity distribution corresponding to The light intensity distribution corresponding to this correlation function has a correlation peak at a position corresponding to the position of the reference image and the correlated image, as explained in the above-mentioned conventional technique. Since the positions where these correlation peaks appear are known in advance, if the light detection section of the photodiode array 21 is placed at the position where these correlation peaks appear, the light intensity of these correlation peaks will be detected, and the light intensity of the photodiode array 22 will be detected. It can be converted into an electrical signal using its photoelectric conversion function. Here, the second liquid crystal television 14 is located on the front focal plane of the first Fourier transform lens 15, and the optical writing type spatial light modulator 16 is located on the back focal plane of the first Fourier transform lens 16 and on the second Fourier transform lens 15. A photodiode array 21 is arranged on the front focal plane of the transformation lens 20 and on the back focal plane of the second Fourier transformation lens 20 . Furthermore, the first LCD television 1
3 is placed immediately in front of the second liquid crystal television 14.

In the second liquid crystal television 14, the optical writing type spatial light modulator 16 is placed at any position from just before the first Fourier transform lens 15 to the front focal plane, or immediately after the first Fourier transform lens 15. It goes without saying that it may be placed at any position from just before the second Fourier transform lens 20 to the front focal plane.

If the joint transform correlator of the present invention described above is used, only the common spatial frequency components of the correlated image and the reference image are used for recognition, so that extremely high-speed and accurate pattern recognition is possible. A particular problem arises when a correlated image and a reference image that are similar to each other and completely include each other in the spatial frequency domain are used. For example, consider a case where correlated image F is input when reference images include E, F, and B. At this time, the image displayed on the reference image input means is as shown in FIG. 9(a),
A set of a correlated image and a reference image that is actually subjected to joint Fourier transformation is as shown in FIG. 9(b). In this way, if the light transmittance of a portion where no image is displayed is zero, the reference image that is actually subjected to joint Fourier transform will be an image that is exactly the same as or extremely similar to the correlated image. Therefore, the joint transform correlator of the present invention may perform incorrect pattern recognition.

At this time, in the example of the correlated image shown in Fig. 7, by appropriately determining the light transmittance of each dark area where the image is not shown, a joint Fourier transform as shown in Fig. 10 is performed. Can create images. for example,
In the correlated image shown in Figure 7, the light transmittance of the dark areas where no image is shown is 50%, and in the reference image shown in Figure 8, the light transmittance of the dark areas where no image is shown is 0%.
Then, the light transmittance of the image part in Figure 10 is 10
0%, the light transmittance of the non-common parts of the correlated image and the reference image indicated by diagonal lines is 50%, and the light transmittance of the parts other than the above is 0%.
It can be done. In other words, even when using multiple correlated images and reference images that are similar to each other, it is possible to extract and strengthen the common image parts, and at the same time to perform joint Fourier transform on the non-common image parts with a predetermined intensity. can. If there are only 4 to 6 correlated images in the reference image, performing this kind of processing in advance can significantly improve the pattern recognition ability of the joint transform correlator. Understood.

Further, in the reference image display means 3, in a state where only a plane wave having a predetermined light intensity and containing no image information is irradiated, the amount of transmitted light of the display portion of the reference image is determined corresponding to those reference images. If it is changed according to the area ratio of the parts or the amount of transmitted light and normalized so that the Fourier transform intensities are equal, it can be used in conjunction with the reference image input means as described above to match the correlated image. Since the input intensity of images other than the reference image can be substantially weakened, pattern recognition with a better S/N ratio can be performed.

For example, if E is used as the correlated image in the example of the reference image shown in FIG. 8 and pattern recognition is performed using the joint transform correlator of the present invention, the magnitude of the resulting correlation coefficient will be the highest If the large one is 1, E=1, B
=0.8, R-0°6, G=0.5, and the correlation coefficients with respect to other reference images were 0.3 or less. Furthermore, when other correlated images are used in Fig. 8, the S
/N was good, and accurate pattern recognition was possible.

The inventors have found that by using the joint transform correlator of the present invention, when performing alpha vent pattern recognition as shown in FIG. 8, pattern recognition can be performed using 13 to 18 reference images simultaneously. confirmed. this is,
This can be said to be an extremely superior result compared to the case where the number of reference images that can be used in a conventional joint transform correlator without a feed hack is 8 to 11.

Next, an embodiment in which the present invention is applied to a joint transform correlator having a feed hack will be described. FIG. 3 is a block diagram of one embodiment of a joint conversion correlator with feedback according to the present invention, in which 25 is an A/D converter, 26 is a third D/A converter, and 27 is a mask liquid crystal. It's television. Note that the same or corresponding parts as in FIG. 2 are given the same numbers, and their explanations are omitted. In FIG. 3, means for converting at least one reference image including a desired target and at least one newly input correlated image into a coherent image include a laser light source] 0, a beam expander 11, and a first
A liquid crystal television 13, a second liquid crystal television 14, a computer 22, a first D/Ai converter 23, and a second D/A converter 24. Further, the means for performing Fourier transform on the coherent image to obtain a joint Fourier transform image of the reference image and the correlated image includes a first Fourier transform lens 1.
5, the means for converting the joint Fourier transform image into an intensity distribution image and displaying the intensity distribution image on the spatial light modulator is an optical writing type spatial light modulator 16. Further, the means for reading out the intensity distribution image displayed on the spatial light modulator using coherence I light includes a laser beam a] 0, a beam Aegis bander 11, a beam subrick 12, a first mirror 17, and a first mirror 17. 2
mirror 18 and a polarizing beam splinter 19. The means for Fourier transforming the read intensity distribution image and converting the image into a correlation signal using a photodetector is the second Fourier transform lens 20 and the claw I/diode array 21. In the means for converting at least one reference image containing a desired target into a coherent image, the newly input at least one correlated image;
The means for irradiating the reference image with coherent light consisting of an image superimposed with a plane wave having a predetermined light intensity and containing no image information includes the first liquid crystal television 13, the second liquid crystal television 14, and the computer 22. First D/A converter 23
and a second D/A converter 24. Further, means for processing the correlation signal to obtain two-dimensional correlation coefficients between the reference image and the correlated image includes an A/D converter 2.
5 and a computer 22, the transmittance or reflection of a portion corresponding to each reference image is determined by a means for converting the reference image into a coherent image, or by a spatial light modulator for a mask placed in front of or behind the reference image. The means for changing the ratio in a linear or non-linear relationship with respect to the correlation coefficient includes a third D/A converter 26 and a mask liquid crystal screen I.
/Bi27.

In the initial process of recognizing an image, at least the portion of the mask liquid crystal television 27 through which the light irradiating the correlated image and the reference image is transmitted is in a state where light is most transmitted. In this state, the operation for determining the correlation coefficient between the correlated image and the reference image is the same as that shown in the embodiment of the joint transform correlator according to the present invention shown in FIG.
A detailed explanation thereof will be omitted. Electric signals corresponding to the light intensities P1 to P of n correlation peaks between the correlated image and the n reference images obtained in this way are sent to the A/D converter 2.
After being converted into a digital signal in step 5, the signal is input into the computer 22. The electrical signals corresponding to the correlation peak intensities taken into the computer 22 are normalized by their maximum value, and the two-dimensional correlation coefficient Cj of the correlated image and the reference image is calculated.
(j=1...nun is a natural number).

This correlation coefficient CJ is calculated by processing Mj = g (Cj
), (j=1.2...,n). This 01
The transmittance MJ (j=1.2
．． ..., n) are determined. However, 0≦M, ≦1.
0≦CJ≦1.

Then, the third
A mask liquid crystal television 27 is controlled via a D/A converter 26 . As a result, the 2 obtained by correlation processing
The light intensity with which each reference image is irradiated can be changed depending on the value of the dimensional correlation coefficient. However, the transmittance Ms of the part corresponding to the correlated image is always maX (CJ), (
j-], 2. ..., n).

With the transmittance of the mask LCD TV 27 changed in this way, write light is again irradiated onto the optical writing type spatial light modulator] 6 and correlation processing is performed in the same manner as above to obtain a two-dimensional correlation coefficient. . The result is expressed as the feedhack transfer function g
feed-hack the transmittance of the mask LCD TV 27 again through
Configure a feed hack loop.

With the configuration as described below, if the light intensity of a certain correlation peak is weaker than the others, the light intensity that irradiates the corresponding reference image will be small due to feedback.If the correlation process is performed again in this state, the light Writeable spatial light modulator 1
Among the joint Fourier transform images displayed in 6, the joint Fourier transform image of the reference image corresponding to the weak correlation peak becomes weak and unclear. As a result, the light intensity of the correlation peak corresponding to that reference image becomes even smaller. Therefore, by repeating this feed hack, the reference image that has a small correlation with the correlated image will gradually be exposed to the masking liquid crystal television 27, and the light intensity irradiating the reference image will become weaker, so that the correlation peak corresponding to the correct reference surface image will be only becomes very strong. Therefore, in the initial state, even if there are many correlation peaks on the correlation output surface and the light intensity of each correlation peak is small and buried in noise, making accurate recognition impossible, repeat the feed hack described above. This allows for accurate recognition.

The joint transform correlator with feed hack of the present invention converts the correlated image displayed on the first LCD television 13 into a coherent image as described above, and then converts the correlated image displayed on the second LCD television 14 into a coherent image. By illuminating the reference image, the image recognition ability in the initial state is enhanced, so the number of feed hacks required to recognize the correct image is reduced compared to the conventional joint transform correlator with feed hacks. At the same time,
More reference images can be used at once. Further, in the reference image display means 3, in a state where only a plane wave having a predetermined light intensity and containing no image information is irradiated, the amount of transmitted light of the display portion of the reference image is determined corresponding to those reference images. By changing the area ratio of the parts or the amount of transmitted light and normalizing them so that the Fourier transform intensities are equal, it is possible to , it is possible to substantially weaken the input strength of images other than the reference image that matches the correlated image, so the S/
N good pattern recognition can be performed.

In this way, when the joint transform correlator with feedback according to the present invention was used to recognize the same type of alpha head as the reference image shown in FIG.
Reference images of 24 or more characters could be used simultaneously as reference images.

In addition, when alpha head recognition was performed using images that were the same as the reference images shown in FIG. 8 but differed only in the correlated images, 1 to Accurate recognition was achieved with an extremely small number of feedbacks (three times).

As the feedhack transfer function used in the embodiment shown in FIG. 3, not only linear functions but also nonlinear functions such as sigmoid functions, logarithmic functions, trigonometric functions, and step functions can be used. It is known that faster pattern recognition can be achieved if a nonlinear function is used as the feedback transfer function used in this embodiment.

The optical writing type spatial light modulator 16 used in the embodiment shown in FIGS. It is preferable to use an optical writing type ferroelectric liquid crystal light valve using a ferroelectric liquid crystal having bistable memory properties during applied voltage. The structure of this optically written ferroelectric liquid crystal light valve includes, for example, a pair of transparent substrates having a silicon monoxide alignment film formed by oblique vapor deposition on their surfaces, and a spacer on the side of the -silicon oxide alignment film. The ferroelectric liquid crystal is sandwiched between the ferroelectric liquid crystal and the ferroelectric liquid crystal. In addition, the transparent substrate on the write side includes a transparent electrode,
A photoconductive film, a light-shielding film, and a dielectric mirror are laminated between the alignment film and the readout-side substrate, and a transparent electrode is formed between the alignment film and the alignment film. By using such a ferroelectric liquid crystal light valve, it is possible to display the join I-Fourier transform intensity distribution image at an extremely fast display speed on the order of microseconds, making it possible to perform pattern recognition at a higher speed. Since the displayed joint Fourier transform intensity distribution image is binarized and displayed, pattern recognition with a better S/N ratio can be performed.

〔Effect of the invention〕

As explained above, the joint transform correlator of the present invention includes means for converting at least one reference image including a desired target and at least one newly input correlated image into a coherent image, and a means for converting the coherent image into a coherent image. converting the joint Fourier transform image to obtain a joint Fourier transform image of the reference image and the correlated image; converting the joint Fourier transform image into an intensity distribution image and displaying the intensity distribution image on a spatial light modulator; means for reading out the intensity distribution image displayed on the spatial light modulator using coherent light; and Fourier transforming the read intensity distribution image and converting the image into a correlation signal using a photodetector. In the joint transform correlator having means for converting, the means for converting at least one reference image containing a desired target into a coherent image, the means for converting at least one reference image containing a desired target into a coherent image, the at least one newly input correlated image having a predetermined light intensity. A means for irradiating the reference image with a coherent image consisting of an image superimposed with a plane wave that does not contain image information, further comprising a means for converting at least one reference image containing the desired target into a coherent image. , in a state where only a plane wave with a predetermined light intensity and no image information is irradiated, the conversion intensity of the reference image to a cohesin 1-image is calculated as the area ratio or transmitted light amount of the part corresponding to the reference image. The structure has means for normalizing the coherent images so that the Fourier transform intensities of these coherent images are equal, and in particular, processing the correlation signal to generate a two-dimensional image between the reference image and the correlated image. and means for converting the reference image into a cohesin 1-image, or a spatial light modulator for a mask placed in front of or behind the reference image, to determine a portion corresponding to each of the reference images. By using means for changing the transmittance or reflectance of This has the effect that pattern recognition can be performed using many reference images simultaneously. Alternatively, if pattern recognition is performed using the same number of reference images as in the conventional joint transform correlator, there is an effect that a higher S/N ratio or more accurate pattern recognition can be performed. Or again,
If pattern recognition is performed using the same number of reference images as in the conventional joint transform correlator with feed hack, there is an effect that faster pattern recognition can be performed.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing the configuration of a joint transform correlator according to the present invention, FIG. 2 is a diagram showing the configuration of one embodiment of the joint transform correlator according to the present invention, and FIG. FIG. 4 is a block diagram of an example of a conventional joint transform correlator, and FIG. 5 shows an example of an input image of the joint transform correlator. FIG. 6 is an example of a correlation peak in a joint transform correlator, and FIG. 7 is a diagram showing an example of a correlated image according to the present invention.
FIG. 8 is a diagram showing an example of a reference image in the present invention, and FIG. 9 is a diagram showing an example of an input image when using mutually detected images. The image displayed on the image input means, FIG. 9(b), is an image subjected to join 1-Fourier transformation. Further, FIG. 10 is a diagram showing an example of an image to be subjected to joint Fourier transform when images similar to each other are used in the present invention. ■... First coherent light source 2... Correlated image display means 3... Reference image display means 4... Means for obtaining joint Fourier transform 5... Spatial light modulator 6... Coherence I of No. 2 - Light source 7... Means for Fourier transforming the intensity distribution image and converting it into a correlation signal 8... Computer 9... Spatial light modulator for mask 10... Laser light source 11... Beam Expander] 2...Beam splinter 13...First liquid crystal television 14...Second liquid crystal television 15...First Fourier transform lens 16...Optical writing type spatial light modulator 7... ...First mirror 8...Second mirror 9...Polarizing beam splitter 0...Second Fourier transform lenst...Photodiode array 2...Computer 3...First D/ A converter 4...Second D/A converter 5...A/D converter 6...Third D/A converter 7...Liquid crystal television for mask 8...Liquid crystal television 9... Optical writing type liquid crystal light valve 0... CCD
Camerad...transparent part 2...light-shielding part and above Applicant: Seiko Electronic Industries Co., Ltd. Agent Patent Attorney: Takayuki Hayashi 31 ? (Stop [Shii Zoumazume's entering Kai Elephant 1 Iku 1 Figure wing wing Niwaichi + 76 correlation beef 14 rows 6 Figure Tako] Ko

Claims (3)

[Claims] (1) means for converting at least one reference image including a desired target and at least one newly input correlated image into a coherent image; and a means for Fourier transforming the coherent image and a joint of the reference image and the correlated image. means for obtaining a Fourier transform image; means for converting the joint Fourier transform image into an intensity distribution image and displaying the intensity distribution image on a spatial light modulator; and converting the intensity distribution image displayed on the spatial light modulator into a coherent light and means for Fourier transforming the read intensity distribution image into a correlation function and converting the correlation function into a correlation signal using a photodetector. means for converting at least one reference image including a target and at least one newly input correlated image into a coherent image, the means for converting the at least one newly input correlated image and an image having a predetermined light intensity into a coherent image; A joint transform correlator comprising means for irradiating the reference image with a coherent image consisting of an image superimposed with a plane wave that does not contain information. (2) In the means for converting at least one reference image including the desired target and at least one newly input correlated image into a coherent image, only a plane wave having a predetermined light intensity and containing no image information is irradiated. , the conversion strength of the reference image to a coherent image is
2. The joint transform correlator according to claim 1, further comprising means for normalizing the coherent images so that the Fourier transform intensities of the coherent images are equal, by changing the area ratio or the amount of transmitted light of the portions corresponding to the reference images. (3) Means for performing signal processing on at least one correlation signal containing the desired target and at least one newly input correlated image to obtain two-dimensional correlation coefficients between the reference image and the correlated image, respectively. and a means for converting the reference image into a coherent image, or a mask spatial light modulator placed in front of or behind the reference image, to change the transmittance or reflectance of the portion corresponding to each of the reference images to the coherent image. 3. The joint transform correlator according to claim 1, further comprising means for changing the relational coefficient in a linear or nonlinear manner.