KR920003653B1 - Picture perceiving device - Google Patents

Abstract

No content.

Description

Image recognition device

1 is an exploded view showing an embodiment of an image recognition device in the case where an image is stopped.

Fig. 2 is a block diagram showing the addition compression means during the first.

3 is a block diagram showing the addition compression operation of the addition compression means during the first embodiment.

4 is a block diagram showing the power spectrum calculating means during the first step;

5 is a diagram showing the operation of the image recognition device when the target image is stationary.

6 is a block diagram showing a judging means during the first step.

7 and 10 are block diagrams showing embodiments of the image recognition device in the case where the target image is in motion.

8 is a block diagram showing the addition compression means during the seventh.

9 is a diagram showing the operation of the image recognition device when the target image is moving.

11 is a block diagram showing an image recognition device capable of recognizing an image regardless of the operation of the target image.

FIG. 12 is a block diagram in which an adder is provided in the determining means shown in FIG.

* Explanation of symbols for main parts of the drawings

10, 110, 210: Image input means 11, 111, 211: Add compression means

12, 112, 212: power spectrum calculation means 13, 113, 213: determination means

14, 114, 214: data output means 15, 115, 215: control means

[Industrial use]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image recognition apparatus for recognizing a desired image signal, and more particularly, to an image recognition apparatus for extracting and recognizing a desired signal from an acquired image of a stationary object or an image of a moving object.

[Traditional Technology and Problems]

In general, an image recognition device includes an imaging unit such as a CCD camera for imaging an object, and performs preprocessing such as gradation conversion on the image output acquired by the imaging unit, and then Binarization is carried out with the determined level as an economy.

The binarization output by this process corresponds to, for example, a character recorded on the target object or a damaged part present on the target object. The binarization output is compared with the reference data stored in advance, and the image desired by the binarization output. It is determined whether or not it corresponds to the signal, and the determination result is output to the data output means.

In such a device, however, if the object is printing a character on a metal plate, a slight tint may be generated in the image due to the finishing state of the metal surface or the arrangement relationship between the imaging unit and the lighting. When the density profile on the line across the factor is binarized to a predetermined level, the profile of the factor profile and the metal plate serving as the background become almost the same brightness (concentration), making it difficult to extract only the factor and correcting the target object. There is a problem that can not be recognized. This is true even when the object to be extracted is not limited to the printing of characters, and various object portions such as various printed matters and damaged portions become almost equal to the brightness of the object.

According to this, even when a target object is being transported, such as a manufacturing line in a manufacturing plant, the image of the target portion deteriorates in the input image. Therefore, when the extracted image is binarized by the binarization means, it becomes difficult to recognize the target portion. There is also.

As described above, in the conventional image recognition apparatus, there is a problem that there is little difference in density between the target object and the target portion, or when the target object is being conveyed, it is difficult to extract the target object from the input image, so that the target portion cannot be recognized reliably. .

[Purpose of invention]

SUMMARY OF THE INVENTION The present invention has been made in view of the above point, and an object thereof is to provide an image recognition device capable of more reliably extracting a desired image regardless of whether a target object is stationary or moving.

[Configuration of Invention]

In order to achieve the above object, the present invention provides an image input means (10) for capturing a stationary object and outputting image data, and the image data of the object in response to the image output of the image input means (10). Is added to one direction on the screen, the power compression means 11 for calculating the power spectrum by Fourier transforming the output signal of the addition compression means 11, and the power spectrum calculating means 12. A judging means 13 for judging whether or not the output of the target object exceeds the reference value, and data indicating whether a desired image is present in the picked-up image of the target object by receiving an output signal from the judging result by the judging means 13; A control signal is output to the output means 14 and the control signal input terminals of the image input means 10, the addition compression means 11, the power spectrum calculating means 12, the determination means 13 and the data output means 14. Authorized Characterized in that it comprises a control means 15 for controlling the drive of each of the means (10, 11, 12, 13, 14).

In addition, the present invention, the image input means 110 for picking up the moving object and outputting the image data, and the image output of the image input means 110 is input to Fourier transform the image data of the target object to power Power spectrum calculating means 112 for calculating a spectrum, addition compressing means 111 for adding the output of the power spectrum calculating means 112 in one direction on the screen, and the output of the adding compressing means 111 exceeds a reference value; Determination means (113) for determining whether or not the camera is supplied; data output means (114) for receiving an output signal according to the determination result of the determination means (113) to indicate whether or not a desired image is found in the photographed image. And the image input means (110) and addition compression means (111). Control to control the driving of the respective means (110, 111, 112, 113, 114) by applying a control signal to the control signal input terminal of the power spectrum calculating means (112), the determining means (113), and the data output means (114). It is characterized by consisting of means (115).

The present invention also provides an image input means (210) for photographing a target object and outputting image data, an addition compression means (211) for adding image data of the target object in one direction on the screen, and the image input means (210). Output spectrum from the power spectrum calculation means 212, the addition compression means 211 or the power spectrum calculation means 212 to Fourier transform the output signal of the output signal or the output signal of the addition compression means 211 to calculate a power spectrum A judging means 213 for judging whether or not the level of the power spectrum signal to be exceeded the reference value is determined, and whether or not a desired image is present in the picked-up image of the target object by receiving the output signal of the judging means 213; If the output image data of the data output means 214 or the image input means 210 is a still image, the addition compression means 211 outputs the output image data. Is added and compressed for each line on the screen, and then the switching unit 227 is switched to output the addition compression processing output to the power spectrum calculating means 212 via the data bus 211a, and the power spectrum calculating means. At 212, the power spectrum of the added-compressed image input data is calculated, and the power spectrum is output to the determination means 213 via the switch 246 and the data bus 212a, and the image input means ( When the output image data of 210 is a moving image, the switching unit 227 is switched to supply the output image data directly to the power spectrum calculating means 212 via the data bus 211a, and the power spectrum The calculating means 212 calculates a power spectrum of the uncompressed image input data, and then converts the power spectrum by the changer 246 and the data bus 212b. 211, the addition compression means 211 adds and compresses the power spectrum signal and outputs the result of the addition compression process to the determination means via the switch 227 and the data bus 211b. Characterized in that it comprises a control means 215 for controlling the addition compression means 211 and the power spectrum calculation means 212.

[Action]

According to the present invention configured as described above, in the case where the image data is stopped, the image data is added in one direction by using an addition compression means, and then Fourier transformed to calculate a power spectrum, and the calculated value and the reference value are compared. By doing this, the image is recognized.

Therefore, since the image data is handled in the spatial frequency region, the target portion can be surely grasped even when there is little difference in density between the target object and the target portion.

In addition, when the image data is moving, first, the image data is Fourier transformed, the image is treated as a spatial frequency, and it is added and compressed to emphasize the signal of the target portion, thereby making it possible to reliably grasp the target portion.

EXAMPLE

Hereinafter, one embodiment according to the present invention will be described in detail with reference to the accompanying drawings.

FIG. 1 is a view showing an image recognition device for performing image recognition when the object is stationary, and a case of detecting whether or not there is a character or the like on the object will be described as an example.

In FIG. 1, the image input means 10 is an imaging device such as a CCD camera. The image input means 10 performs A / D conversion of an image signal obtained by capturing a target object, thereby generating a digital shade image. It outputs to the addition compression means 11 via the bus 10a. The addition and compression means 11 is constituted of, for example, a digital signal processing apparatus (DSP). The addition and compression means 11 adds and compresses an image signal input via the data bus 10a and then performs the data compression through the data bus 11a. The addition compression process result is supplied to the power spectrum calculation means 12. As shown in FIG. That is, the adding and compressing means 11 adds the image signal from the data bus 10a in one direction on the screen and adds the addition result via the data bus 11a as one line of data to the power spectrum calculating means 12. Output to The power spectrum calculating means 12 calculates the power spectrum in the frequency domain of the image signal by performing, for example, a one-dimensional Fourier transform on the image signal from the data bus 11a with, for example, a digital signal processing apparatus (DSP), The power spectrum signal from the power spectrum calculating means 12 is outputted to the determining means 13 via the data bus 12a. In the determination means 13, the reference level is stored in advance. When there is a power spectrum signal exceeding the reference level by comparing the power spectrum signal level with the reference level from the data bus 12a, a character or the like is recorded on the object. It is determined that there is, and the determination signal is outputted to the data output means 14 via the data bus 13a.

The data output means 14 displays in a suitable display method whether there is a character or the like on the object according to the determination signal from the determination means 13, which is a visual display means (e.g., LCD, LED, CRT) or printing means (for example, thermal printer).

Further, the image input means 10, the addition compression means 11, the power spectrum calculating means 12, the determining means 13, and the data output means 14 are connected to the control signal input terminal, for example, a general purpose calculator or the like. Control means 15 are connected via a control bus 15a, and the drive means 15 control the drive as described above.

FIG. 2 is a block diagram of the addition and compression means 11 of the image recognition device shown in FIG. 1. The image memory 21 shown in the same drawing stores image data applied from the data bus 10a. It has a capacity to store data for one screen (e.g., 512 x 512 pixels). Of the image data stored in the image memory 21, the data addressed from the address generator 22 via the address line 25a is buffered through the data buses 21a, 21b, 21c. 25 is stored in the buffer memory 23, 24, 25 via the address line 25b of the address generator 22. The address designation of the buffer memory 23, 24, 25 is performed.

The adder 26 receives the data of the buffer memories 23, 24, and 25 via the data buses 23a, 24a, and 25a, and adds the result of adding the data of the two buffer memories to the buffer memory 23. Or 25), data corresponding to a plurality of lines of the data screen stored in the image memory 21 can be added. In the image memory 21, the address generator 22, the buffer memories 23, 24, 25, and the adder 26, control signals from the control means 15 shown in FIG. And supplied to control the addition operation of the image data.

FIG. 3 schematically shows the addition operation of the addition compression means 11 shown in FIG. 2. First, the addition operation is performed by the image data of the next line from the image memory 21 as shown in FIG. 24 is added, and the data stored in the buffer memories 24, 25 are added by the adder 26, and the addition result is output to the buffer memory 23 for storage.

Further, in the next addition operation, as shown in FIG. 3C, the image data of the next line from the image memory 21 is taken into the buffer memory 24, and the data stored in the buffer memories 23 and 24 in the adder 26. FIG. After the addition is made, the addition result is output to the buffer memory 25 and stored. Thereafter, this operation is repeated to accumulate and add image data for a predetermined line, and the final result is output to the power spectrum calculating means 12 via the data bus 11a.

4 is a block diagram of the power spectrum calculating means 12. The image data from the data bus 11a is once stored in the input memory 41, and then supplied to the FFT processor 42 and Fourier transformed. That is, the FFT processor 42 has as many points as the number of pixels existing in one line, and outputs the conversion result to the output memory 43 in the form of a complex signal. As a result, real data R and imaginary data I are stored.

을 구하는 장치로서, 이 합(P)이 파워스팩트럼의 값으로 되고, 이와같은 자승합 연산기(44)는 디지탈신호 처리기(DSP)로 구성하여도 된다. 그리고, 상기 자승합 연산기(44)에서 얻어진 파워스펙트럼은 출력메모리(45)에 기억 유지되고, 데이터버스(12a)를 매개하여 판정수단(13)에 공급된다.On the other hand, the square sum operator 44 shown in FIG. 4 is, for example, a sum of squares of real data R and imaginary data I, each composed of a multiplier and an adder, stored in the output memory 43.

As the device for calculating the sum, this sum P is the value of the power spectrum, and such a square sum calculator 44 may be configured as a digital signal processor DSP. The power spectrum obtained by the square sum calculator 44 is stored in the output memory 45 and supplied to the determination means 13 via the data bus 12a.

FIG. 5 shows the concept of image data addition compression and power spectrum calculation operation when the object is stationary. Character ABC is recorded on the object, and the captured image 51 is as shown in FIG. 5A. Assume In the image memory 21 shown in FIG. 2, for example, 512 lines of data on the captured image are stored, and the data of these 512 lines are added as described in FIG. Here, the data to be added does not have to be 512 lines. That is, the data of the pixels corresponding to each line are sequentially added in the direction of the arrow 51R, and finally, the compressed data 52 is added for one line.

The data 52a portion of the data for one line corresponds to the character portion 51a of the captured image 51, and similarly, the data 52b portion of the data corresponds to the background portion 51b. The added compression data 52 is one-dimensional Fourier transformed in the direction of the arrow 52R to calculate the frequency spectrum and sum the squares to obtain the power spectrum as shown in FIG. 5B.

When the object is stopped in this manner, when the image data is first added and compressed, and then the power spectrum is calculated, the specific portion 53 shown in FIG. 5B in the power spectrum appears on the predetermined frequency component. This is because by adding and compressing the image data, the change in the density profile of the random data in the background portion is canceled and made smaller, and the emphasis is because the change in the density profile of the character portion data is at about the same position on one line. Therefore, when the level of the power spectrum in the predetermined frequency portion exceeds the reference level, it can be seen that there is a feature portion (i.e., a character) on the image pickup screen.

The addition compression operation before calculating the power spectrum is not necessarily required when the feature portion also appears in one line of data in the captured image. In addition, the level data of the power spectrum in the range of the band B near the frequency where the feature appears in the power spectrum may be added, and the addition result may be compared with a reference value to detect the presence or absence of the feature.

6 is a block diagram of the determination means 13 shown in FIG. 1, in which the power spectrum data supplied from the data bus 12a is stored in the power spectrum memory 71 shown in FIG. That is, the power spectrum data stored in this power spectrum memory 71 corresponds to FIG. 5B. The reference value for determination is stored in the variable memory 72, which corresponds to the reference level L shown in FIG. 5B. The data stored in the power spectrum memory 71 and the variable memory 72 are supplied to the subtractor 72 via the data buses 71a and 72a, respectively, which subtracts the reference value data from the power spectrum data. It subtracts and outputs the result to the determiner 74 via the data bus 73a. The data output from the subtractor 73 may be only code data of the subtraction result.

Next, when the sign of the subtraction result is positive, the determiner 74 judges that a character is recorded on the captured image, i.e., the object, on the picked-up image. The data of the determination result is output to the data output means 14 shown in FIG. On the other hand, if the sign of the subtraction result is negative, it is determined that there is no feature part on the picked-up image, and a signal indicating that no signal is output or that there is no feature part is output to the data output means 14.

Next, referring to FIG. 7 for an image recognition device for detecting a character or the like recorded on the object when the object to be imaged moves, that is, when the relative positional relationship between the object and the image recognition device changes. Will be explained.

As shown in FIG. 7, the image signal obtained by the image input means 110 is supplied to the addition compression means 111 via the data bus 110a and then the power spectrum calculating means via the data bus 111a. And the output of the power spectrum calculating means 112 are supplied to the addition and compressing means 111 again via the data bus 112a.

Here, the image data supplied from the image input means 110 to the addition compression means 111 is not added and compressed by the addition compression means 111 and directly to the power spectrum calculating means 112 via the data bus 111a. Is output.

8 shows a block diagram of the adding and compressing means 111. Each component is similar to the image recognition apparatus for recognizing the stationary object shown in FIG. 2, but the signal flow is somewhat different. .

The image data stored in the image memory 121 of the addition compression means 111 is supplied from the buffer memory 123 or 125 to the power spectrum calculating means 112 via the data bus 111a. Further, the output of the power spectrum calculating means 112 is again supplied to the adding and compressing means 111 via the data bus 112a and then subjected to the addition compression process as in the operation shown in FIG.

Subsequently, this addition-compressed data is supplied to the judging means 113 via the data bus 111b and compared with a reference level stored in advance, so that a judging signal is outputted through the data bus 113a. By supplying to the data output means 114, an indication of whether a character or the like is recorded on the object is performed.

Further, the image input means 110, the addition compression means 111, the power spectrum calculating means 112, the determining means 113, and the data output means 114 are connected to the control signal input terminal such as a general purpose calculator or the like. The control means 115 is connected via the control bus 115a, and the drive means 115 controls the drive as described above. Here, the power spectrum calculating means 112, the determining means 113, and the data output means 114 shown in FIG. 7 are the same as those shown in FIG.

Next, the concept of the addition compression operation and the power spectrum calculation operation for the case where the object is moving will be described in detail with reference to FIGS. 9A to 9D.

First, image data obtained by photographing a moving object stored in the image memory 121 shown in FIG. 8 is supplied from the buffer memory 123 or 125 to the power spectrum calculating means 112 via the data bus 111a. do. Then, the power spectrum calculating means 112 performs a Fourier transform for each line of the input data, calculates the sum of squares, and transmits the power spectrum for each line to the addition compression means 111 via the data bus 112a. The power spectrum calculating means 112 performs a Fourier transform on the captured image 61 shown in FIG. 9A in the direction of the arrow 61R, and calculates the power spectrum by obtaining the sum of the squares of the complex data. do. This power spectrum is roughly represented by frequency 0 to + f, -f to -0, as indicated by reference numeral 62 in FIG. 9B. That is, in the case of an image of 512 x 512 pixels, the highest spatial frequency expressed in the image is 256 (the frequency is 256 in 256 pixels because the frequency is 1 in one set of pixels). Therefore, when the image is one-dimensional Fourier transformed, the highest spatial frequency appears at the 256th point, so that frequency information of 0 to the highest frequency f is included in the half-width region of 512 x 512 pixels. This area is the "real" area, and the other half is the "vacant" area. In the power spectrum display, the spectrums of both regions are symmetrical, and as shown in FIG. 9B, the "real" region is a positive frequency region of 0 to + f, and the "period" region is -0 to -f. It is common to think of the negative frequency domain.

FIG. 9C shows the power spectrum for each line of the " real " regions 0 to + f in FIG. 9D, where 512 power spectrums corresponding to n = 512 lines are obtained. In the case of a moving image, the feature is not clearly shown in the power spectrum of each line shown in FIG. 9C. Here, when these power spectra 63 ₁ to 63 _n are added by the adding and compressing means 111, the specific portion 64 becomes clear as shown in FIG. 9D. Therefore, by comparing the level of the power spectrum with the reference value L, it is possible to detect the presence or absence of a feature part.

Also in this case, the level data of the power spectrum in the range of the band B near the frequency at which the feature appears may be added, and the addition result may be compared with a reference value to detect the presence or absence of the feature. In addition, in the embodiment of Fig. 7, the output of the image input means 110 is input to the addition compression means 111. In this addition compression means 111, the image input data is stored in the buffer memories 123 and 125. In addition, since the addition compression operation is not performed in particular, as shown in FIG. 10, the image input data may be directly supplied to the power spectrum calculating means 112, and the output may be supplied to the adding compression means 111. FIG.

Next, an image recognition apparatus capable of recognizing a character or the like recorded on an object regardless of whether the object is stationary or moving will be described.

11 shows a block diagram of an image recognition apparatus capable of image recognition irrespective of the state of a target object. First, an image signal obtained by the image input means 210 is added and compressed by means of a data bus 210a. Is supplied. In the case where the image input data is stopped, the addition compression means 211 performs addition compression processing, then switches the switching unit 227, and outputs the addition compression processing output via the data bus 211a to the power spectrum calculation means ( 212), when the image input data is moving, the switching unit 227 is switched to directly supply the image input data to the power spectrum calculating means 212 via the data bus 211a.

Further, the power spectrum calculation means 212 calculates the power spectrum of the image input data. When the image input data is subjected to the addition compression process, the switcher 246 is switched to determine the data spectrum via the data bus 212a. The output is output to the means 213, and if it is still being subjected to addition compression, the switch 246 is switched and supplied to the addition compression means 211 to the determination means 213 via the data bus 212b. In this case, the addition compression means 211 adds and compresses the power spectrum signal, switches the switching unit 227, and outputs the result to the determination means 213 via the data bus 211b. The determination means 213 makes a comparison between the power spectrum signal level from the data buses 212a or 211b and the reference level stored in advance, and according to the comparison result, whether or not a character or the like exists on the object. Determination is made, and the determination signal is output to the data output means 214, and the data output means 214 displays the determination result according to the determination signal.

Here, the switching of the data bus is performed by the switchers 227 and 246 as follows when performing processing when the object is stationary or moving. That is, the output of the determination means 213 is supplied to the detector 216, and a switching control signal in the stationary or moving state is supplied to each of the switchers 227 and 246 in accordance with the presence or absence of an output signal from the detector 216. The switchover is made at the appropriate time.

Incidentally, the operations of the addition compressing means 211, the power spectrum calculating means 212, the determining means 213, and the data output means 214 are the same as in the case of the above-mentioned FIG.

On the other hand, the control of the switchers 227 and 246 according to the state in which the object is stopped or moving may be performed manually, and the compression level and the reference level of the power spectrum signal may be determined and controlled according to the result. do.

The image recognition device according to the present embodiment can also be applied to the printing of the stationary object, and of course, even if the moving object takes an image almost at a still state using a high speed camera (a camera having a shutter speed of about 1/1000 second). You can.

In the device according to the present invention, when a character is imprinted relatively clearly on each input image, it is added to one direction on the screen to be taken as image data corresponding to one line. There is some knowing. Therefore, when the one-dimensional Fourier transform is performed on this one-line data, characteristic peaks appear in the power spectrum. Therefore, by determining the level of the peaks, it is possible to determine the presence or absence of printing on the object.

On the other hand, when a specific peak does not appear in the one-dimensional Fourier transform output and is dispersed, a predetermined band of the one-dimensional Fourier transform output (for example, the band B of FIG. 5B) is added to determine the level of the addition output. useful. Of course, if the data is compressed in the first spatial region when the noise is large or the characters imprinted on the object are inclined, the characteristic spectrum may be canceled by the data addition of each line and the characteristic spectrum may not appear. . In such a case, the method shown in FIG. 9 that uses Fourier transform for each line of data and then adds it is useful. In this case, although each power spectrum shown in FIG. 9B is not necessarily constant, the spatial frequency of the horizontal line 61a corresponding to the character shown in FIG. 9A is included in common. Therefore, when each power spectrum shown in FIG. 9B is added, characteristic peaks appear as shown in FIG. 9D in the addition result.

Furthermore, even in the addition result, a characteristic peak may not appear due to an illumination state or a driving state of an object. In this case, a specific band centered on a frequency estimated to represent the peak value in a frequency domain, for example, in FIG. The spectrum of the band B shown is added and the level of the addition result is compared with the reference level. Such a method can be appropriately set in accordance with the characteristics of the target image, and can provide an apparatus capable of reliably recognizing a print on the target object.

As described above, the image recognition device according to the present invention adds the input image data directly or in one direction through an addition compression means, and then Fourier transforms the power spectrum, and adds it in a predetermined band to add a stable spectrum. After the calculation is made, the calculated value is compared with the reference value to recognize the image, so that the image data is handled in the spatial frequency region, so that the target portion can be surely grasped even when there is little difference in density between the target object and the target portion. In addition, according to such an image recognizing apparatus, by adding image data directly or in one direction and then processing, high-speed processing can be realized almost constant according to the data amount.

In addition, since there is almost no difference in concentration between the target object and the target portion, even when the power spectrum is dispersed in a predetermined region, the extracted image can be grasped reliably, thereby making it possible to ensure recognition. In addition, even in the case where the target object moves so that the power spectrum is indistinctly distributed, the target portion can be reliably grasped, and almost certain recognition can be performed.

Incidentally, each of the image recognition apparatuses described above is configured to direct the image data from the image input means directly to the addition compression means. However, the present invention is not limited to this, but the image data is applied to the front end of the addition compression means according to the type of image signal. The preprocessing means for performing gradation conversion or the like may be configured integrally or separately.

Further, in the above embodiment, the case where the analog signal is configured to be handled by A / D conversion has been described. However, the present invention is not limited thereto, and the analog signal can be configured to be handled as it is.

In addition, as shown in FIG. 12, adders 75 and 76 are placed at the front end of the determination means 213 shown in FIG. 11 so that the features appear in the power spectrum shown in FIGS. 5B and 9D, respectively. The level data of the power spectrum in the range of the band B in the vicinity of the frequency is added, and the addition result is compared with a reference value to detect the presence or absence of the feature portion, so that even when the level of the feature portion is small, discrimination can be made.

[Effects of the Invention]

As described above, according to the present invention, when the image data of a target object is stopped, the image data is added and compressed for each line, and then Fourier transformed to obtain a power spectrum signal. Fourier transforms are performed for each line to calculate a power spectrum signal, and then these compression outputs are added and compressed, so that in any case, image data can be treated as a spatial frequency. It can be determined.

Therefore, the target object or the target part in the target object can be stably detected with less influence of the imaging condition, illumination condition, travel condition, and the like.

Claims (3)

Image compression means 10 for picking up a stationary object and outputting image data, and an addition compression for receiving the image output of the image input means 10 and adding the image data of the object in one direction on the screen; Means 11, power spectrum calculating means 12 for Fourier transforming the output signal of the adding and compressing means 11 to calculate a power spectrum, and determining whether the output of the power spectrum calculating means 12 exceeds a reference value; A data output means 14 for receiving the output signal of the determination result by the determination means 13, and for indicating whether or not a desired image is found in the picked-up image of the target object; and the image input. A control signal is applied to the control signal input terminals of the means 10, the addition compression means 11, the power spectrum calculating means 12, the determining means 13, and the data output means 14, and the respective means 10, 11, 12, 13, 14 drive Image recognition apparatus, characterized in that consists of the control means (15) for air. Image input means 110 for imaging the moving object and outputting the image data; and power for calculating the power spectrum by Fourier transforming the image data of the target object by receiving the image output of the image input means 110; The spectrum calculating means 112, an addition compressing means 111 for adding the output of the power spectrum calculating means 112 in one direction on the screen, and determining whether the output of the adding compressing means 111 exceeds a reference value; Determination means 113, data output means 114 for receiving an output signal of the determination result by the determination means 113 and indicating whether or not a desired image is found in the picked-up image of the object; and the image input means. A control signal is applied to the control signal input terminal of the 110 and the addition compression means 111, the power spectrum calculating means 112, the determining means 113, and the data output means 114, respectively. , 113, 114) Image recognition apparatus, characterized in that consists of the control means (115) for controlling the drive. Image input means 210 for photographing a target object and outputting image data, an addition compression means 211 for adding the image data of the target object in one direction on the screen, an output signal of the image input means 210, or Power spectrum calculation means 212 for calculating the power spectrum by Fourier transforming the output signal of the addition compression means 211, the power spectrum signal output from the addition compression means 211 or the power spectrum calculation means 212 A judging means 213 for judging whether or not the level is above a reference value and determining whether or not the reference value is exceeded, and outputting data indicating whether a desired image is present in the picked-up image of the object by receiving an output signal from the judging means 213 When the output image data of the means 214 and the image input means 210 is a still image, the addition compression means 211 sets the output image data to 1 on the screen. After each addition and compression process, the switching unit 227 is switched to output the addition compression processing output to the power spectrum calculating unit 212 via the data bus 211a, and the power spectrum calculating unit 212 adds the addition. The power spectrum calculation means 212 of the compressed image input data is output to the power spectrum calculation means 212, and the power spectrum calculation means 212 calculates the power spectrum of the addition-compressed image input data to convert the power spectrum into a switch 246 and When the output image data of the image input means 210 is a moving image by outputting to the determination means 213 via a data bus 212a, the switch 227 is switched to convert the output image data into data. The power spectrum of the image input data which is directly supplied to the power spectrum calculating means 212 via a bus 211a and not added and compressed in the power spectrum calculating means 212. After calculating the rum, the power spectrum is supplied to the adding and compressing means 211 through the switch 211 and the data bus 212b. The adding and compressing means 211 adds and compresses the power spectrum signal. Control means for controlling the addition compression means 211 and the power spectrum calculating means 212 to output the addition compression processing result to the determination means 213 via the switch 227 and the data bus 211b ( 215, characterized in that the image recognition device.

Images