JPS63245779A - Parallel extracting device for projection feature of image - Google Patents

Abstract

PURPOSE:To shorten a time required for parallel extraction of projection features by introducing optical parallel processing to image extraction. CONSTITUTION:An image on a display 1 is led into a lens array 3 through a projection lens 2. Respective images are simultaneously formed on the photodetecting faces of individual projecting optical sensors 5a in an optical sensor array 5 by individual image forming lenses 3a of a lens array 3. Electric signals corresponding to intensity distribution waveforms having projecting features specific in respective images formed on the photodetecting faces of the sensors 5a are outputted. The electric signals are compared with the projecting feature of a previously prepared reference pattern by an identification part to identify the images.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to a device that can extract multiple types of projection features of images such as characters and other figures in parallel. OCR (Optical C
It is most suitable for application to image recognition devices such as Haracter Reader.

[Summary of the invention]

The present invention provides an apparatus that is capable of extracting multiple types of projective features in parallel for images such as characters and other figures, which optically multiplies images from which projective features are to be extracted using a multiplication means. A plurality of optical images having substantially the same shape as each other are formed, and these plurality of optical images are individually supplied to a plurality of types of projection detection means, so that a plurality of types of projection features can be detected in parallel. With this configuration, the configuration is simple and low cost, and the time required for parallel extraction of projection features can be shortened, and various projection features can be easily obtained as needed. Is possible.

[Conventional technology]

2. Description of the Related Art Conventionally, image recognition devices that recognize images such as characters and other figures are configured to perform processing mainly using electronics technology.

An example of such a conventional image recognition device will be described below. First, an image pattern to be image recognized printed on a manuscript or the like is a CC image recognition device.
D, an image is formed by an optical lens on the light receiving surface of an image sensor such as a MOS sensor. A multi-value digital signal as image information is output from this image sensor, and this multi-value digital signal is binarized using an appropriate threshold value (if there are multiple threshold values, the above (after performing multilevel conversion different from the case of ), the data is stored in memory. Further, this binarized image information is subjected to preprocessing for image shaping as necessary, and then stored in the above memory or another memory. Note that the preprocessing described above performs noise removal processing, normalization processing of position, size, inclination, line width, etc.

Next, 4. is written in the memory in this manner. A projection processing unit extracts projection features necessary for image identification from the image information that has been processed by a.

That is, when projecting an image onto a directional axis (e.g., the
The above-mentioned image information is read out from this memory in chronological order or in parallel time-series by scanning, the read out image information is transferred to the projection processing section, and then the transferred image information is sequentially processed in the above projection processing section. Weigh. The amounts of electricity sequentially obtained by such measurements are stored at predetermined positions corresponding to the aging direction axis in the above-mentioned memory or another memory. Also, based on the electrical quantity stored in this way, the waveform of the intensity distribution from which the projection feature about the acidic direction axis is extracted is determined.

Since such an operation for extracting projection features of an image is normally performed on a plurality of directional axes, a plurality of intensity distribution waveforms can be obtained for the same image information. The image is identified by comparing the plurality of projection features of the image represented by these intensity distribution waveforms with the projection features of a standard pattern prepared in advance.

In the image recognition device as described above, in order to increase the image recognition rate, it is necessary to perform projection processing on the same image information about many directional axes and extract many types of projection features. In order to perform projection onto a large number of directional axes, (1) the above-described projection processing is sequentially repeated many times in a single projection processing section.

(2) A large number of projection processing sections each having a separate memory for projection processing are installed, and image information read from the above-mentioned memory in which image information is stored is transferred to each of the above-mentioned large number of projection processing sections. , the projection processing as described above is performed in parallel in these many projection processing units.

It is necessary to adopt one of the following methods.

[Problem to be solved by the invention]

However, since the conventional image recognition apparatus described above is configured to perform processing mainly using electronics technology, there is a problem in that the processing time becomes long.

That is, in order to increase the image recognition rate, projection features are extracted for a large number of directional axes. The image information is sequentially read out from this memory by scanning in the direction of , and transferred to the projection processing section, and then the transferred image information is sequentially measured in the projection processing section, and then the electricity obtained by such measurement is It is necessary to find the intensity distribution waveform based on each amount.

Furthermore, it is necessary to sequentially repeat such projection processing many times. Therefore, in this case, the time required for the projection process becomes extremely long, resulting in poor efficiency.

Also, in the case of the 0 term, it is necessary to transfer and measure the image information as described above and then obtain the intensity distribution waveform, so in the case of the 0 term, the time required for the projection process is still The disadvantage was that it was long. Furthermore, in this case, it is necessary to provide a large number of projection processing units, which has the disadvantage that the configuration becomes complicated.

Furthermore, in the conventional image recognition device as described above,
If non-linear projection processing such as circumferential projection and radial projection as described below is to be performed, the start and end positions of each linear scan in a predetermined direction are determined using a special function, It is necessary to determine the range of the non-linear shape to be projected using a large number of such linear scans. Therefore, it is not easy to perform non-linear projection on image information, and therefore it is very difficult to improve the image recognition rate.

The present invention has been made in view of the above problems, and by introducing optical parallel processing to extract projection features of images, a large number of projection features can be extracted at high speed with a simple configuration. It was made so that it could be done.

[Means to solve the problem]

The present invention is substantially the same as a multiplication means that optically multiplies images of characters, other figures, etc. from which projected features are to be extracted to form a plurality of optical images having substantially the same shape. and a plurality of types of projection detection means each capable of detecting projections different from each other from the optical images of the shape, and detecting the plurality of optical images having substantially the same shape formed by the multiplication means. A device for parallel extraction of projection features of an image, characterized in that a plurality of types of projection features are obtained in parallel from a plurality of types of projection detection means by individually supplying the projection features to different types of projection detection means. It is something.

[Effect]

According to the present invention configured as described above, a plurality of optical images having substantially the same shape obtained by optically multiplexing images from which projection features are to be extracted by a multiplexing means are multiplexed. The plurality of types of projection detection means are supplied individually and simultaneously, and a plurality of types of projection features are obtained in parallel from these projection detection means.

〔Example〕

Next, an embodiment in which the present invention is applied to an image recognition device will be described with reference to the drawings.

1 to 5 show an embodiment of the present invention, in which an image pattern on a document is converted into image information by an image sensor in the same way as in the case of the conventional image recognition device as described above. The image is converted, then binarized and preprocessed, stored in a memory as necessary, and then displayed as an optical image on the screen of a display 1 made of a cathode ray tube or the like. The image on the screen is then directed to a lens array 3 via a projection lens 2 placed in front of the screen and separated by its focal length.

A photosensor array 5 including a large number of projection photosensors 5a is arranged in front of the lens array 3. Then, an optical image having substantially the same shape as the optical image displayed on the screen of the display 1 is transmitted to these optical sensors 5a.
The lens array 3
are each provided with an imaging lens 3a at a position corresponding to the optical sensor 5-a. Note that as the lens array 3, a flat plate microlens can be used, in which the same number of microlenses (for example, gradient index lenses) as the optical sensors 5a are formed in a flat plate made of glass, synthetic resin, or the like.

Further, as the lens array 3, the SLA (
Trademark: Nippon Sheet Glass Co., Ltd.) can be used, and in this case, resin is filled between the lenses so that each cylindrical gradient index lens is placed at the position of the imaging lens 3a. All you have to do is turn it into

The above-mentioned projection lens 2 and lens array 3 constitute the above-mentioned multiplication means 4. and projection lens 2
The image on the screen of the display 1 guided to the lens array 3 via the individual imaging lenses 3 of the lens array 3
images are simultaneously formed on the light receiving surfaces of the individual projection optical sensors 5a of the optical sensor array 5. Note that the number of projection optical sensors 5a can be appropriately determined depending on the type of projection feature to be extracted, and it goes without saying that the number of imaging lenses 3a can also be appropriately determined accordingly. .

Next, electrical signals corresponding to intensity distribution waveforms having projection characteristics specific to the image formed on the light-receiving surface of the projection optical sensor 5a are outputted from these optical sensors 5a as described below. Therefore, in the same way as in the case of the conventional image recognition device as described above, by comparing these electrical signals with the projection features of a standard pattern prepared in advance in the identification section, image identification is performed electrically. can be done.

Next, an example of the optical sensor used as the projection optical sensor 5a shown in FIG. 1 will be described in detail based on FIGS. 2 to 5. Note that these figures show the case where the character "Koto" is displayed on the screen of the display 1, and therefore, the character "Koto" is also imaged on the light receiving surface of the optical sensor 5a.

The optical sensor shown in FIG. 2A is for performing projection in the X-axis direction, and includes a large number (16 in the illustrated case) of optical sensor units each having a rectangular light-receiving surface having a sensor width Δ. It consists of 10. These optical sensor units 10 are arranged so that their side ends are adjacent to each other in order, and constitute a substantially square light-receiving surface as a whole.

Then, according to the shape of the image formed on the light-receiving surface of the optical sensor, electric signals corresponding to the projection time characteristics are obtained from each optical sensor unit 10, and from these electric signals, as shown in FIG. 2B, An intensity distribution waveform in the X-axis direction can be obtained.

The optical sensor shown in FIG. 3 has a pattern obtained by rotating the optical sensor pattern shown in FIG. 2A by an arbitrary angle θ around the origin. Furthermore, the optical sensor shown in FIG.
As in the case of the optical sensor shown in the figure, each sensor width Δ
It consists of a large number of optical sensors 11 each having a light-receiving surface.

In the case of the optical sensor shown in FIG. 3, projection onto the X' axis is performed by rotating the X axis by an angle θ. In this case,
Three types of optical sensors of θ-45°, 90°, and 135° can be used, but it goes without saying that various values of θ can be selected as required.

The optical sensor shown in FIGS. 4 and 5 is for performing non-linear projection. That is, the optical sensor shown in FIG. 4 is for performing projection in the circumferential direction, and is composed of a large number (16 in the illustrated case) of optical sensor units 12 arranged concentrically with each other. Note that the central optical sensor unit among these optical sensor units 12 has a circular light receiving surface with a small diameter. The remaining optical sensor units are provided with annular light-receiving surfaces having different diameters, and their widths are equal to 1' and equal to the radius of the central optical sensor unit. These optical sensor units 12 are arranged so that the outer peripheral edge of the inner optical sensor unit is adjacent to the inner peripheral edge of the outer optical sensor unit in order from the central optical sensor unit to the outer peripheral optical sensor unit. , which forms a circular light-receiving surface as a whole. In the case of this optical sensor, even if the image formed on the light-receiving surface is rotated, it is possible to obtain an intensity distribution waveform that removes the influence of this rotation. Features can be extracted from images.

The optical sensor shown in FIG. 5 is for performing projection in the radial direction, and is composed of a large number (16 in the illustrated case) of optical sensor units 13, each having a sector-shaped light receiving surface and having the same shape as each other. ing. These optical sensor units 13 are arranged around the one point so that the vertices of their central angles are almost all gathered at one point. In this state, the side edges thereof are adjacent to each other in sequence, and the whole forms a circular light-receiving surface. In the case of future sensors, even if the image formed on the light-receiving surface is slightly shifted from the center position, this movement of the center will not have much of an effect on the intensity distribution waveform, so we will extract projection features that do not change the center movement. can do.

Note that in the optical sensors shown in FIGS. 2 to 5, a large number of optical sensor units 10 to 13 can be arranged in a plane on a common substrate. In this case, these units 10
Since a transparent electrode is usually formed on the plane of ~13,
It is preferable to provide a slight gap between adjacent units so that these electrodes are not electrically connected to each other.

Furthermore, it goes without saying that various changes can be made to the optical sensors shown in FIGS. 2 to 5.

For example, the shape and number of the optical sensor units 10.11 can be changed so that the light-receiving surface of the optical sensor shown in FIGS. 2A and 3 becomes circular, rectangular, etc. as a whole. In addition, the optical sensor unit 1 is arranged so that the light receiving surface of the optical sensor shown in FIGS. 4 and 5 is square, rectangular, etc. as a whole.
2.13 shape and number can also be changed. In this case, for example, in the optical sensor shown in FIG. 4, the central optical sensor unit is formed into a square or rectangular shape, and the remaining optical sensor units are changed from an annular shape to a rectangular annular shape of a 1' square or a 4' rectangle. By changing to , the light-receiving surface can be changed to a square or a rectangle, and in this case as well, the influence of image rotation can be eliminated to a certain extent. Furthermore, in the optical sensor shown in FIGS. 2A, 3, and 5,
The shapes of the optical sensor units do not necessarily have to be the same, and in the optical sensors shown in FIGS. 2A, 3, and 4, the widths of the optical sensor units do not necessarily have to be the same.

As described above, the optical sensor array 5 includes various projection optical sensors, for example, one optical sensor for X-axis direction projection shown in FIG. 2A, and one optical sensor for X-axis direction projection shown in FIG. 3. Three optical sensors at θ-45°, 90°, and 135°, one optical sensor for circumferential projection shown in Fig. 4, and one optical sensor for radial projection shown in Fig. 5', for a total of 6 pieces. The light-receiving surface of each of these optical sensors is made up of a large number of optical sensor units 10 to 13.Therefore, different signals are detected from the electrical signals output from each of these optical sensor units 10 to 13. Six types of intensity distribution waveforms having projection characteristics can be obtained.

Note that in the above embodiment, the second projection detection means is
In addition to the optical sensors shown in FIGS. 5 to 5, a sensor partially shown in FIG. 6 may be used. This projection detection means includes a mask 7 having a projection slit 8 and a simple surface light sensor 9 having a single light receiving surface disposed corresponding to the mask 7.
It is equipped with a large number of mask-sensor pairs consisting of. Each projection slit 8 has a shape that substantially corresponds to any one of the optical sensor units 10 to 13 shown in FIGS. 2 to 5. Note that each simple surface optical sensor 9 may have approximately the same shape as the mask 7 as shown in the figure, or may have approximately the same shape as the slit 8 and be arranged to correspond to the slit 8. The structure may be such that the light that is imaged by the imaging lens 3a and passes through the slit 8 is substantially received by the simple surface optical sensor 9.

The projection detection means, a part of which is shown in FIG. 6, is for projection in the X-axis direction, and its specific configuration will be explained as follows. That is, this projection detection means for projection in the X-axis direction includes 16 masks 7 in which projection slits 8 each having a shape approximately corresponding to the 16 optical sensor units 10 shown in FIG. 2A are formed; It consists of 16 simple surface optical sensors 9 arranged in one-to-one correspondence with these masks 7. Therefore, the projection detection means for projection in the X-axis direction is composed of 16 mask/sensor pairs. Each mask/sensor pair is connected to the imaging lens 3 shown in FIG.
They are arranged in one-to-one correspondence with a. That is, the optical image on the screen of the display 1 is formed on the light receiving surface of the simple surface optical sensor 9 via the projection lens 2, the imaging lens 3a, and the projection slit 8 of the mask 7. Although FIG. 6 shows a portion of the projection detection means for the X-axis projection, it is quite obvious that the optical sensors shown in FIGS. 3 to 5 can also be modified in the same way. . However, in this case, since 16 mask/sensor pairs are required in place of each optical sensor, it is necessary to provide 16 times more imaging lenses 3a than in the case shown in FIG.

Furthermore, in FIG. 6, instead of the simple surface sensor 9,
A light concentrator may also be used. In this case, since the light passing through the projection slit 8 is collected by the condenser,
Intensity distributions with projective features can be displayed optically. For example, a lens or a prism provided at one end with a flat light-receiving surface having substantially the same shape as any one of the optical sensors 10 to 13 shown in FIGS. 2 to 5 can be used as the light condenser. In this case, if the lens or prism provided with the light-receiving surface is configured so that its diameter gradually decreases from one end to the other end, the light received by the light-receiving surface will be focused into a spot. The intensity distribution can be optically displayed as the intensity of the light spot.

Further, in the above-described embodiment, the image pattern on the document is read by the image sensor, and this is displayed on the display 1.
The displayed image is multiplied by the multiplication means 4. However, it is also possible to arrange the document directly at a position corresponding to the screen of the display 1 and to directly multiply the image pattern of this document by the multiplication means 4.

Further, in the above embodiment, an appropriate gap is provided between the imaging lens 3 and the light receiving surface of the optical sensor 5a, but the image on the screen of the display 1 is
By forming an image on the end face of this lens on the optical sensor 5a side, the imaging lens 3a and the light receiving surface of the optical sensor 5a can be brought into contact with each other.

〔Effect of the invention〕

The present invention utilizes a plurality of types of projections to obtain a plurality of optical images having substantially the same shape, which are obtained by optically multiplexing images from which projection features are to be extracted using a multiplication means, such as in a double series. The projection features are supplied to the detection means individually and simultaneously, and a plurality of types of projection features are obtained in parallel from these projection detection means.

Therefore, according to the present invention, since optical parallel processing is introduced for image extraction, the configuration is simple and low cost, and the time required for parallel extraction of projective features can be shortened. It is also possible to easily obtain various projection features as needed.

[Brief explanation of drawings]

1 to 5 show an embodiment in which the present invention is applied to an image recognition device, in which FIG. 1 is a schematic diagram of a parallel extraction device for image projection features, and FIG. 2A is an X-axis A front view of the optical sensor for directional projection, FIG. 2B is an intensity distribution waveform diagram obtained from the optical sensor for X-axis direction projection shown in FIG. 2A, and FIG. 3 is an X'
FIG. 4 is a front view of the optical sensor for axial projection, FIG. 4 is a front view of the optical sensor for circumferential projection, and FIG. 5 is a front view of the optical sensor for radial projection. FIG. 6 is a schematic perspective view showing a portion of another example of the projection detection means for projection in the X-axis direction. In addition, in the symbols used in the drawings, 1-------------Display 2-----
−〜−−−−−−−−−−−−−−Projection lens 3−−−−
−−−−−−−−−−−−−Lens array 4−−−−−−
----------Multification means 5 a-----1.-----1 is a projection optical sensor.゛Agent Masaru Saturn

Claims (1)

[Scope of Claims] 1. Multiplication means for optically multiplication of images from which projection features are to be extracted to form a plurality of optical images having substantially the same shape; and optical devices having substantially the same shape. a plurality of types of projection detection means each capable of detecting mutually different projections from the target image; 1. A parallel extraction device for image projection features, characterized in that a plurality of types of projection features are obtained in parallel from the plurality of types of projection detection means by individually supplying them to the detection means. 2. An image sensor for detecting an image, a preprocessing means for preprocessing the image signal from the image sensor for image shaping, and displaying the image signal preprocessed by the preprocessing means as an optical image. 2. The apparatus according to claim 1, further comprising a display means for displaying an optical image, wherein the optical image displayed by the display means is optically multiplied by the multiplication means. 3. Each of the projection detection means has a light receiving surface formed by combining a plurality of optical sensor units, and the combinations of the plurality of optical sensor units on these light reception surfaces are different from each other for each projection detection means. 3. The apparatus according to claim 1, wherein the apparatus is configured such that a plurality of optical images formed by the multiplication means are respectively formed on the light receiving surfaces of the plurality of types of projection detection means. 4. At least one of the projection detection means includes a large number of annular optical sensor units having sequentially different diameters,
4. A device according to claim 1, 2 or 3, wherein these optical sensor units are arranged successively adjacent in the radial direction. 5. At least one of the projection detection means is provided with a large number of substantially fan-shaped optical sensor units, and these optical sensor units are arranged around the one point so that the vertices of their respective central angles are gathered at approximately one point. 4. Apparatus according to claim 1, 2 or 3, wherein the apparatus is arranged one after the other.