JPS62168290A - Image recognizing device - Google Patents

Abstract

PURPOSE:To decrease a memory capacity and to execute a high speed processing by executing the image pickup of plural circular objects, binary- coding an outputted video signal, and deriving the coordinate of the center of gravity of a single circular object, based on the counting output of the number of picture elements of a binary image in an area for surrounding the single circular object to be recognized. CONSTITUTION:As for connected circular objects brought to an image pickup by a television camera 6, only a hexagonal area for surrounding a single circular object among them is set. A CPU 20 makes a window memory store its set information. On the other hand, a video signal whose synchronizing signal is separated by a synchronization separating circuit 8 is binary-coded to black and white by a binarization circuit 9. In the binary image, the number of picture elements contained in an area shown by an area signal is counted by a picture element counter 17 and given to the CPU 20, and also, a clock signal count value counted by horizontal and vertical counters 24, 27 is also given to the CPU 20, and based on these count values, the coordinate of the center of gravity of the single circular object to be recognized is derived.

Description

[Detailed Description of the Invention] [Industrial Application Field] This invention is based on the first picture! ? The present invention relates to an image recognition device, which is applied to a robot's visual device, for example, and which recognizes a single circular object when a plurality of circular objects are connected.

[Prior Art] FIG. 9 J3 and FIG. 10 are diagrams showing examples of images recognized by conventional image recognition devices.

First, a conventional image recognition device will be briefly described with reference to FIG. 9 J3 and FIG. 10. As shown in FIG. 9, when a plurality of circular objects 2 such as bearings are arranged in a row in the horizontal slot 1, a method for recognizing a single circular object 2 among them will be explained. Conventional painting J&N
? , the identification device images a plurality of circular objects 2.2... connected in a slot 1 with a television camera, processes the video signal output from the television camera into 21 format, and temporarily stores it in an image memory. I was letting it happen. Then, by software processing the image data stored in the image memory, the contours of a plurality of circular objects 2, 2, . . . are extracted, and each circular object is classified and labeled, as shown in FIG.

[Problems to be Solved by the Invention] As described above, in the conventional image recognition device, the binarized image data is stored in the -H image memory, the outline of each circular object is extracted, and the lζth point of labeling is performed. Since processing is required, a large capacity buffer memory is required. Moreover, since it is necessary to first import image data into a buffer memory and then read and calculate each pixel, the processing time is reduced to zero.
．． There was a problem in that it took about 5 seconds to 1 second, resulting in a long processing time.

Therefore, the main object of the present invention is to provide an image recognition device that can recognize a single image of a circular object in an extremely short time without requiring a large capacity image memory.

[Means for Solving the Problems] The image recognition vJc device according to the present invention includes a retraction means that images a plurality of circular objects and outputs a video signal without displaying a gray scale image, and a damped image carrier. At the back of the area, an area signal generating means for generating a signal representing an area surrounding a single circular object to be recognized among the plurality of circular objects, and a video signal output from the static image means are binarized. A binary processing means that generates a binary image from a plurality of vertical and horizontal pixels, and a binary processing means that calculates the number of pixels included in an area of the binary image that is expressed by an area signal (F'') to 4.
The apparatus is comprised of a counting means for counting, and an arithmetic processing means for calculating the coordinates of the center of gravity of a single circular object to be recognized based on the counting results.

[Operation] In the image recognition device of the present invention, a plurality of circular objects can be
image, convert the output video signal 82, and recognize it.
The number of pixels of the image is counted within the area surrounding the circular object to be captured, and the coordinates of the center of gravity of the single circular object are calculated based on the counted output.

[Embodiment of the Invention] FIG. 2 is a diagram for explaining the invention in detail, and FIG. 3 is a time chart of a video signal in an embodiment of the invention.

First, the principle of the present invention will be explained with reference to FIG. 2 J3 and FIG. 3. In FIG. 2, each vertical and horizontal grid represents a pixel, and each pixel has a horizontal coordinate address X1.
X2...×2.6 and vertical coordinate address Y1. Y2
...Y256 is assigned.

Image a30 is a 1lfa image of object 7.
2IfJi obtained by binarizing the J' fir number
iii image, for example, the object 7 is a black pixel,
The Gyoda portion is represented as a white pixel. In FIG. 2, G indicates the center of gravity of the image 30, and its coordinates (XG, YG) can be expressed as a function of Σ as shown in the following equations (1) to (3).

Note that since the image 30 is a binary image, the density weighting function r(Xl, Y,) described above is either 11 Q II or II i 11 as expressed by the following m(3) formula. Takes a value.

As mentioned above, in equations (1) to (3),
ΣY, -N, is the first moment regarding the vertical coordinate address, 4...N,,42;
- In the embodiment shown in FIG. 2, specific examples of the moments are shown on both the left and right sides of the figure.

For the embodiment shown in FIG. 2, 2ffl! for each horizontal scan line of the video signal! The O-th moment N regarding the horizontal coordinate address of the image 30, ;! '3 and the first moment Σx1 are determined by hardware, and the first moments Y, N, regarding the vertical coordinate address are determined in the next horizontal scanning line.
Assuming that Σ is descended in a soft F~ manner, the cumulative addition value ΣN,, Σ over the effective vertical scanning period for each horizontal scanning line is
Y, N・, ΣΣX+, ΣYjJ JJ
J Σ

Yc) is calculated using software.

FIG. 3 is a time chart of the video signal in an embodiment of the present invention, in particular, FIG. 3(a) shows the video signal VD in the vertical scanning period, and FIG. 3(b) shows the video signal VD in the horizontal scanning period. A3 video signal HD is shown. In addition, in FIG. 3, VD indicates a vertical synchronization signal,
1-ID indicates a horizontal synchronization signal, and one vertical scanning period (16
, 6 μsec> is made up of a vertical blanking period of 201-1 (where 11-1 means one horizontal scanning period, which is 635 μsec) and the remaining effective vertical scanning period. This effective vertical scanning period includes 242 horizontal scanning lines, and each horizontal scanning line includes 256 pixel data. In addition, in Fig. 3, Y
l, Y2, Y-...Y24□ correspond to the vertical coordinate address in FIG. 2, Xl, X2, X-...×2
5G supports horizontal coordinate addresses.

FIG. 1 is a concrete block diagram of an embodiment of the present invention. First, the configuration of an embodiment of the present invention will be described with reference to FIG. The television camera 6 conveys an image of the object 7, for example from above, and outputs a video signal representing a grayscale image of the object 7. This video signal is given to a sync separation circuit 8, which separates a horizontal sync signal HD and a vertical sync signal VD from the video signal, generates a clock signal CK, and generates a video signal from which the sync signal is separated. is given to the 21a conversion circuit 9. The binarization circuit 9 converts the video signal into black and white for each field based on a predetermined threshold level.
A binary image pattern signal is outputted to a video monitor 11 via a video synthesis circuit 10.
and one input terminal of the AND gate 16.

The address counter 12 is reset by the vertical synchronizing signal VD, counts the clock signal GK, and divides the horizontal address. The count output of this address counter 12 is given to the window memory 14 via the multiplexer 13, and specifies the address of this window memory 14. This window memory 14 stores V area information indicating a window 34 surrounding a single circular object 32 among a plurality of circular objects 31 to 33 shown in FIG. 4, which will be described later.

The area information output from the window memory 14 is applied to the other input terminal of the AND game 1-16, and is also applied to the video monitor 11 via the video synthesis circuit 10. Therefore, the images of the plurality of circular objects 31 to 33 and the image of the window 34 are displayed on the video monitor 11.

The water 1L counter 24 is for 1) allocating the horizontal coordinate address mentioned in 0, and starts counting the clock signal CK after a time corresponding to, for example, 10 pixels has elapsed from the fall of the horizontal synchronization signal HD. . horizontal counter 24
The counting output of is given to an adder 25, and the output of the adder 25 is given to a buffer 26. The output of the buffer 26 is I
It is given to the adder 25 as well as to the adder 25. The adder 25 and buffer 26 calculate the cumulative addition E[lff
This is for calculating 4X1 and outputting it to the CPU 20, and it uses the count 1fix+ of the horizontal counter 24 and the buffer 2.
The contents of 6 are cumulatively added.

The aforementioned AND gate 16 outputs the pixel signal included in the window 34, and supplies the output to the pixel counter 7 and the adder 25.

The pixel counter 7 is for counting the number of pixels (black pixels in this embodiment) constituting the circular objects 31 to 33 surrounded by the window 34 for each horizontal scanning line, and the count value N , is sent to CPU2 via I1018
given to 0. Note that the output ΣX of the buffer 26 mentioned above
For every 141 horizontal scan lines, C,
Based on these data, the CPU 20 executes a predetermined calculation using 98 principles to obtain the coordinates (XG, Ya) of the center of gravity G of the image.

Note that one setting switch sets the window 34 stored in the window memory 14, and the area information set by this setting switch 19 is given to the CPU 20 via l1018. The CPU 20 provides area information representing the area set by the setting switch 19 to the window memory 14 via the P/S converter 15. In addition, F
ROMg2 stores a series of programs for determining the coordinates of the center of gravity, and RAM23 stores various data.
It consists of a work area for executing processing. Further, the communication control unit 21 is for transmitting the calculation results by the CPU 20 to a higher-level controller.

4 and 5 are diagrams showing an example of a circular image to be recognized as J: in an embodiment of the present invention, and FIG. 6 illustrates a specific operation of the embodiment of the present invention. It is a flow diagram of the process.

Next, with reference to FIGS. 1 to 6, a specific operation of an embodiment of the present invention will be described.

First, as shown in FIG. 4, an object 7 consisting of a series of circular objects 31 to 33 is imaged by the television camera 6. At this time, the setting switch 19 is operated to set an area representing a hexagonal window 34 surrounding the single circular object 32. Setting information set by the setting switch 19 is sent to the CPU 20 via l1018. The CPU 20 provides area information to the P/'S converter 15 based on the setting information. The P/S converter 15 converts the area information from parallel to serial and stores it in the window memory 14.

On the other hand, the video signal output from the television camera 6 is given to the synchronization separation circuit 8, where the horizontal synchronization signal 1-ID and vertical synchronization signal VD are separated, and the
will be issued. The video signal from which the synchronization signal has been separated is sent from the synchronization separation circuit 8 to the binarization circuit 9 (L7), which converts the video signal into black and white 21a, and converts it into black and white 21a.
A fi image Bakun signal is output. This binary image pattern signal is displayed on the video monitor 1 via the video synthesis circuit 1o, and is also applied to one input terminal of the AND gate 16.

On the other hand, the address counter 2 receives the clock signal CK and passes it through the multiplexer 13 to the 8! Sends a number output to window memory 4. As a result, area information representing the window 34 (not shown in FIG. 4) is output from the window memory 4. This area information is given to the video monitor 11 via the video synthesis circuit 10, and
Open gate 16. Therefore, a series of circular objects 31 to 33 and a window 34 surrounding a single circular object 32 are displayed on the video monitor 1, as shown in FIG. The AND gate 16 outputs a pixel signal 8 surrounded by a window 34 for each horizontal scanning line and supplies it to the pixel counter 7 and the accelerator 25. The pixel counter 7 calculates the number of pixels surrounded by the window 34 as 4, and outputs the 51 pixels (direct N) to the CPU 2 via the l1018.
Give to 0. The horizontal counter 24 also receives a clock signal C.
K is counted and the count jiff X + is given to the adder 25. Then, the output of the adder 25 is stored in the buffer 26, and the output copy x1 is given to the adder 25 and also to the CPU 20 via l1018.

Further, the vertical counter 27 counts the horizontal synchronizing signal 1-ID, and transmits the counted value Y to the CPU via l1018.
Give 20.

Next, referring to FIGS. 4 and 6, the single circular object 3
A method for calculating the center of gravity of 2 will be explained. Figure 4<a
>If we count the number of pixels included in the 21 self-portrait pattern of the single circular object 32 surrounded by the hexagonal window 34 and parts of the circular objects 31 and 33 on both sides of it for each horizontal scanning line, The distribution will be as shown in Figure 4 (b). As shown in FIG. 4(b), 1 is a threshold value for removing image noise, and K2 is a threshold value for determining a peak value. Further, points 0 to A and points 8 to D are the number of pixels included in the single circular object 32, and A
From the point [3 points is the sum of the number of pixels of the single circular object 32 and some of the circular objects 31 and 33 surrounded by the window 34, and between the 0 point and the A point are the points J3 and 8. Compared to point D, the rate of increase and decrease in the number of pixels is greater in some areas.

Therefore, in the present invention, when the rate of increase or decrease in the number of pixels is greater than the increase or decrease in the number of pixels in the previous one horizontal scanning line a3, that is, between point A and point 8, The number of bars for the number of pixels is prohibited, and only the number of pixels between the 0 point and the A point J3 and between the 8 point and the D point are counted.

Now, assuming that the pixel counter 7 has completed counting black pixels for the Y4th horizontal scanning line, the CPU 20 first reads the count value N of the pixel counter 7 in step (abbreviated as SP in the figure). . Then, in step SP2, the read count III N
, exceeds a threshold value of 1 or not. If N.Crab exceeds 1, it is determined in step SP3 whether or not the end flag is set. This end flag counts the number of pixels for each horizontal line, and is calculated by counting the number of pixels for each horizontal line.
) After counting the number of pixels up to point D shown in
It is not set in the initial state.

Therefore, the process advances to step SP4.

At step SP4, the difference ΔN between the number of pixels NF in the previous horizontal scanning line and the number N of pixels in the next horizontal scanning line is calculated. Number of pixels N F IL O
Therefore, ΔN becomes N, -3.

Then, in step SP5, it is determined whether the difference ΔN in the number of pixels is smaller than or equal to the threshold l1lJ1<2 at the peak point of the number of pixels shown in FIG. 4 (1) by J. If ΔN≦]ku2, total @ direct N.・ is determined to be the peak value of the number of pixels. However, since peak 1 has not yet been reached at this point, the process proceeds to step SP6.

At step SP6 a3, it is determined whether the difference ΔN in the number of pixels is greater than or equal to the difference ΔN F in the number of pixels immediately before. When ΔN is larger than ΔNF, it becomes 0 point as shown in FIG. Determine whether it is set.

This I N l-1 flag is set when scanning between point A and point 8, as shown in FIG. It represents. In the initial state J3, the IN 1 flag is not hit, so the process advances to step SP8. A start flag is set in step SP8, and the process proceeds to step SP9.

At step SP9, J3 accumulates the count value N,
A cumulative addition value N (corresponding to iiff notation ΣN) is calculated and stored in the RAM 23. Roughly, in step 5P10, the CPU 20 calculates the count value Y of the center of gravity counter 27.
, read. At this point, the count value of the vertical counter is Y
It has become l.

The CPU 20 goes to step 3P11, performs the product addition value NT· (corresponding to the above-mentioned 3Y, NJ), and stores the calculation result in the RAM 23.

At step SP12, the contents of the buffer 26 ΣX
, t (Xi, Y, > is read out, and in step SPυ13, the cumulative addition value NT2 (the above ΣΣ

- Crush f(corresponding to Xi, Y, )). and,
Stella 7SP 14 ni J-; i T, pixel r! IN,
@The previous number of pixels NF is set, and the difference ΔN in the number of pixels is set as the difference ΔNF in the previous number of pixels. Furthermore, step SP
15, the count value Y of the vertical counter 27 is '2
Step SP for the 242nd horizontal scanning line of Hiiragi
It is determined whether the processing from SP1 to SP14 is completed. Here, since the processing of steps SPI to SPl 4 has only been completed for the first horizontal scanning line, the process waits until the next interrupt lNT1, and when that interrupt occurs, the steps SP1 to 5P1 described above are completed. The operation of ti is not returned.

The operations of steps SP1 to SP15 described above are repeated to count the number of pixels from point 0 to point A shown in FIG. 4(b), and in processing the next horizontal scanning line, the number of pixels is counted in step SP6. When it is determined that the difference ΔN is larger than the previous difference ΔNF in the number of pixels, the process goes to step SP16 J3 and sets the I N I-1 flag. Then, proceed to step SP1/l, set the number of pixels N- in that horizontal scanning line as the immediately preceding number of pixels NF, and proceed to step S
Proceed to P15. Then, in the next horizontal scanning line, it is determined in step SP7 that the I N H flag is set, and the processing in steps 8 to 5P13 cannot be executed, and the process directly proceeds to step 5P14. However, steps SP8 to SPI3 are not executed from point A to beak J shown in FIG. 4(b).

When it is determined in step SP5 that the pixel number difference ΔN has become smaller than the peak point threshold value 2, step SP
Similar to I7, the difference ΔN in the number of pixels is the difference ΔN in the number of pixels immediately before.
It is determined whether or not it has become smaller than F. Since the difference ΔN in the number of pixels in the water XIL scan line exceeding the peak point is not smaller than the difference ΔNF in the number of pixels immediately before, the process immediately proceeds to step 5P14. Repeat this operation and
When reaching point B (not shown in figure (b)), the number of pixels XΔN becomes smaller than the difference ΔNF between the previous numbers of pixels, so step S
It is determined in P17, and I is determined in step 5P18.
Reset the Nl-1 flag.

That is, once point B is exceeded, the distance between point B and point D is only the number of pixels of the single circular object 32, so the operations of steps SP9 to 5P13 are executed.

When the point D is reached, the number of pixels N becomes smaller than the threshold value K1, so this is determined in step SP2, the start flag is determined to be set in step 5P22, and the end flag is flagged in step 5P23. When the end flag is set to 1-, in the processing of the next horizontal scanning line, it is determined in step SP3 that the end flag is set, and the process proceeds to step SP14, and the process proceeds to step SP4 of the previous step (E). Or 5P13 paste" is not executed. Then, when the vertical counter 27)11 value becomes 242, step 5P15
This is determined in step SP19, and the process proceeds to step SP19.

In step 5P19, previous) flea step 5P11
The Y coordinate Yao of the center of gravity G is calculated based on the cumulative addition value NT obtained in step SP9 and the cumulative addition value N obtained in step SP9, and is stored in the RAM 23.

Furthermore, in step 5P20, step 5P13
Cumulative addition n1 direct NT2 obtained in step SP9 and 1 obtained in step SP9
．． : Based on the cumulative addition value N, the X coordinate of the center of gravity G is XGO.
is calculated and stored in the RAM 23. Furthermore, in step 5P21, each cumulative addition value N, NT+, NT
2, etc., and complete the series of operations.

FIG. 5 is a V diagram showing an example of an image recognized by another embodiment of the present invention.

Although the window 34 shown in FIG. 4 previously had a hexagonal area surrounding the single circular object 32, in the embodiment shown in FIG. 111- and most of the circular objects 31, 33 on both sides. By using such a window 35, as shown in FIG. 5(b), the number of pixels increases more slowly from point Δ than the change in the number of pixels from point 0 to point A. +11
1 L, and since the number of pixels decreases rapidly at 8 points,
This has the effect of making it easier to make the judgments in steps SP6 and 5P17 shown in FIG. 6 described above more accurately. Note that the window 35 consisting of a cross-shaped area can be set in any shape by using one setting switch shown in FIG.

Figure 7 is recognized by this other embodiment of Brahma'#
8 is a diagram showing an example of J@, and FIG. 8 is a flowchart for explaining the processing operation for recognizing the image shown in FIG. 7.

The embodiment shown in FIGS. 7 and 8 attempts to recognize a single circular object 82 in the center when circular objects 81 to 83 such as light emitting diodes are arranged in a vertical direction. . For this purpose, a window 84 consisting of a rectangular area is set so as to surround a portion of the medium-heat circular object 82 and the circular objects 81J3 and 83 on both sides. When the number of pixels surrounded by the window 84 is multiplied by the number of layers, the number of pixels m becomes as shown in FIG. 7(b). That is, point A where circular objects 81 and 82 touch and circular objects 82 and 8
The B points where 3 touch each other become minimum points.

Therefore, by detecting the minimum point of point A and Brm, calculating the number of pixels between point Δ and point B, and executing a predetermined calculation, the coordinate (Xa) of the center of gravity G of the circular object 82 can be obtained.

Yc) is required. In addition, Fig. 7(b)
KO at is the threshold value at the minimum point Δ,B.

Further, the hardware configuration in this embodiment can be the same as that shown in FIG. 1 described above, and a program as shown in FIG. 8 may be stored in the FROM 22 in advance.

Next, the specific operation of this embodiment will be explained with reference to FIG. First, in step 5P31, the 51 number output Y of the vertical counter 27 is Y of the minimum point Δ, Pi
! ! It is determined whether it is greater than or equal to the target address YA. When Y is smaller than YA,
Proceeding to step 5P32, count value N of the pixel counter 17
, is set as the number of pixels NF of the straight curve. Step 5P3
3, the count value Y of the vertical counter 27 is the minimum point B.
It is determined whether the Y coordinate address Ya is larger than the Y coordinate address Ya.

In this state, since Y is smaller than Ya, the process waits until the next interrupt INT2 occurs. When the next interrupt INT2 is input, the operations of steps 5P31 to 5P33 described in the description are not repeated. Then, when the count value Y of the vertical counter 27 matches the Y coordinate address Y4 of the minimum point A, step 5P3
Proceed to step 4, read the count 1N of the pixel counter 7,
In step S P35, the difference ΔN between the number of pixels N· which is the count output of the pixel counter 7 and the number of pixels NF of the straight curve is determined.

J3 is at step 5P36 and the end flag is Sera]~
When it is determined that the end flag has not been set, it is determined in step 5P37 whether or not the edge flag is set. Since the edge flag is not set here, it is determined in step 5P38 whether the number of pixels N is smaller than the threshold value KO of the minimum point A or if it is a cup. And N
, is smaller than KO, it is determined that the number of pixels at the minimum point Δ is the argument, and step 5P39
Set the edge flag at .

Furthermore, execute the child operations of steps 5P32 and 5P33, and when the next interrupt [NT2 occurs, execute the operations of steps 5P31 to 5P36. And step 5P
In step S37, it is determined whether the edge flag has been set, and in step SP40, it is determined whether or not the start flag has been set. Since the start flag is not set here, it is determined in step SP/11 whether the difference ΔN in the number of pixels is equal to or larger than the threshold value 1. After determining that ΔN lfi K is 5 greater than 1, step SP/I
2 and hit the start flag. In other words,
After determining γ to the minimum point, steps SP9 to 5P13 shown in FIG. 6 described above are performed.
Run it with 43 or 5P47. J Nara, CPU2
0 is the cumulative ait stop of the cumulative addition values N, NT, and NT2.

In the next horizontal scanning line, since the start flag has already been set, it is determined in step S P /I O that the start flag has been set, and in step 5P48 it is determined whether the top flag has been set or not. Determine. Since the 1 to Tsubu flag has not been set yet, the difference ΔN in the number of pixels is determined to be O in step 5P49.
Determine whether it is equal to or less than. ΔN/)
When tO is equal to or smaller, the number of pixels at the center of the circular object 82 has been counted, so a top flag is set in step 5P50.

When the top flag is set, in the next horizontal scanning line, it is determined in step 5P48 that the top flag has been set, and in step 5P51, it is determined whether the difference ΔN in the number of pixels is equal to the threshold value 1 or Daiki'
Determine whether or not it has occurred. If ΔN is larger than 1, the processing in steps 5P43 to 5P47 described above is executed. When ΔN is smaller than 1, it is determined that the minimum point has been reached, and the end is executed in step 5P52. Set the flag. Then, step 5P
Proceed to step 32, where the number of pixels N is set as the previous number NF of pixels and let J' is executed.

If it is determined in step S P33 that the Y coordinate address of the vertical counter 27 has reached the Y coordinate address Ya of the minimum point B, the Y coordinate address of the center of gravity G is determined in step S P53.
The coordinate address Yc is calculated, and the X address XG of the center of gravity G is calculated in step 5P54. Furthermore, in step 5P55, each cumulative addition value N, NTI stored in the RAM 23.

NT2 etc. are cleared.

[Effects of the Invention] As described above, according to the present invention, a plurality of circular objects, each of which is connected to another, is converted to The number of pixels in the area surrounding one of the circular objects is counted for each horizontal scanning line, and based on the output, the barycentric coordinates of the circular object to be recognized are calculated. Therefore, there is no need to store image data captured in the conventional manner in an image memory, and the coordinates of the center of gravity of a single circular object can be calculated in one field. Therefore, there is no need to provide a large-capacity image memory, and the 111-circular object can be recognized in a short time.

[Brief explanation of drawings]

FIG. 1 is a concrete block diagram of an embodiment of the present invention. FIG. 2 is a diagram for explaining the invention in detail. FIG. 3 is a time chart of a video signal in one embodiment of the present invention. FIG. 4 is a diagram showing an example of an image to be recognized by an embodiment of the present invention. Fifth
The figure is a diagram showing an example of an image to be recognized by another embodiment of the present invention. FIG. 6 is a flowchart for explaining the specific operation of one embodiment of the present invention. FIG.
FIG. 3 is a diagram showing an example of an image. FIG. 8 is a single diagram for explaining the specific /J operation of another embodiment of the present invention. FIGS. 9 and 10 are diagrams showing examples of images recognized by a conventional image recognition device. In the figure, 6 is a television camera, 7 is a circular object, 8 is a synchronization separation circuit, 9 is a binarization circuit, 10 is a video synthesis circuit,
11 is a video monitor, 12 is an address counter, 13 is a multi-branch, 14 is a window memory, and 15 is a P/S.
Converter, 16 is an AND gate, 17 is a pixel counter, 1
8 is 1.10.19 is setting switch, 20 is CPU, 2
2 is PROM, 23 is RAM, 2/l is horizontal counter,
25 is a stimulator, 26 is a buffer 1, and 27 is a vertical counter. (2 others) Figure 1 Figure 3 PX Figure 4 Figure 5 (a) (b) Figure 6 Figure 9 Figure 10

Claims (4)

[Claims] (1) An image recognition device for recognizing a single circular object among a plurality of circular objects when they are connected, which captures images of the plurality of circular objects and generates a video signal representing a grayscale image. an area signal generating means for generating a signal representing a region surrounding a single circular object to be recognized among the plurality of circular objects in an imaging area imaged by the imaging means; Binary processing means for binarizing the video signal output from the means to generate a binary image consisting of a plurality of vertical and horizontal pixels; and a binary image generated by the binary processing means;
a counting means for counting the number of pixels in the area represented by the area signal generated by the area signal generating means; and calculating the coordinates of the center of gravity of the single circular object to be recognized based on the count output of the counting means. An image recognition device comprising an arithmetic processing means that performs. (2) The arithmetic processing means selects a measurement range in the vertical direction of the single circular object based on the count output of the counting means, and calculates the center of gravity coordinates. The image recognition device according to item 1. (3) The arithmetic processing means selects and measures a vertical region from the minimum point to the minimum point of the number of pixels for each horizontal scanning line based on the count output of the counting means,
The image recognition device according to claim 1, wherein the image recognition device calculates the coordinates of the center of gravity of the single circular object. (4) The arithmetic processing means includes means for calculating a zero-order moment and a first-order moment regarding the horizontal coordinate address of the binary image for each horizontal scanning line of the video signal, and for each horizontal scanning line of the video signal. means for calculating a first-order moment with respect to a vertical coordinate address of a binary image; means for calculating a cumulative addition value in a vertical scanning period for each of the moments; 4. The image recognition device according to claim 1, further comprising means for calculating the center of gravity of a circular object.