JPS6145438B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to a device for matching an input image with a pre-registered image, and particularly to an image matching device that is effective for matching seal stamps.

An object of the present invention is to provide an image matching device that can automatically match two images with a simple configuration.

Another object of the present invention is to provide an image matching device that can not only automatically match two images, but also manually match them if necessary.

Hereinafter, the present invention will be explained with reference to the drawings.

FIG. 1 is a schematic explanatory diagram of an image matching device according to the present invention. The general operation of this device will be explained using FIG. 1. For the first time, the changeover switch 118 is turned to the contact 132 side, and the television camera 114 (hereinafter simply referred to as the camera) is connected to the video disk 108 (hereinafter simply referred to as the disk). camera 11
4 and the disk 108 receive the same horizontal drive signal HD and vertical drive signal VD (hereinafter simply referred to as HD and VD, respectively) from a timing signal generator 120 (hereinafter simply referred to as the generator). The camera 114 captures the stamp image on the stamped document 112, such as a check, and stores it on the disk 1 as image information.
Recorded on 08. In this way, disk 108
For example, many types of seal images necessary for business are recorded in advance.

In this state, the switch 118 is moved to the contact 131 side, and the switch 116 is moved to the contact 121 side. Next, a document stamped with a seal to be compared with a previously registered or recorded seal is set in a predetermined position. Next, the registration number input device 10
4, enter the registration number of the seal on disk 10.
8, the address of the disk where the image of the seal is recorded is specified, and the head is set on the specified address. Image A from the camera 114 and image B from the disk 108 are generated by the HD and VD signals given from the generator 120 after the head is set.
The images are taken into the video processing means 100 at the same time. The video processing means 100 determines the positional deviation of the image A in the parallel direction with respect to the image B, and electrically corrects this positional deviation in the parallel direction within the means 100. At the same time, a positional shift in the rotational direction of image A with respect to image B is determined, and the previously determined rotational shift is corrected using image rotation means 110. The signal of the image A whose position has been corrected is superimposed on the signal of the image B, and the signal of the image A whose position has been corrected is superimposed on the signal of the image B.
Images A and B are superimposed and projected on 6. If there are still positional or rotational deviations in images A and B due to stamping or noise,
While watching the monitor television 106, these deviations can be corrected using the joystick 102.

Note that when it is desired to compare images captured by the television cameras 114 and 115, the switch 116 can be moved to the contact 122 side.

Next, each constituent means of the apparatus described above will be explained in detail with reference to FIG. 2 and subsequent figures. FIG. 2a shows the video processing means 100 shown in FIG. 201 and 202 are direction code frequency distribution calculation means, the details of which will be explained with reference to FIG. 3. 203 and 204 are center position calculation means. This will be explained in detail with reference to FIG. Further, 200 is an arithmetic circuit, and here a microcomputer is used. The content of this calculation will be explained in detail with reference to FIG.

Reference numeral 205 denotes a parallel movement means, the details of which will be explained in conjunction with FIG. 7. Furthermore, 206
is a video synthesis circuit, and details of this can be found in Chapter 8.
This will be explained using figures.

Now, 201 is a direction code frequency distribution calculation means for image A obtained by the camera 114,
Reference numeral 202 denotes direction code frequency distribution calculation means for the image B provided from the disk 108. Since both have the same configuration, the operation of 201 will be explained here. In FIG.
The video signal from the image sensor is guided to a shift register 300 that stores information for one horizontal scanning line. This may be a semiconductor shift register or an electromagnetic delay line. A shift-type analog storage device is even more convenient. The output from this shift register 300 is guided to another shift register 301. In this way, in addition to the current video signal, information on the previous scanning line and information on the two previous scanning lines are stored in the shift registers 300 and 30, respectively.
It is obtained as the output of 1. These three outputs are guided to shift registers 302, 303, and 304. These three shift registers have a depth of, for example, 6 bits, and can each store information of 3 picture elements. That is, a in 302,
It is sent to b and c, and is discarded from c. Therefore, a contains the most recent information, b contains information on the previous picture element, and c contains information on the two previous picture elements, and these are updated one after another as the imaging device scans.

Therefore a, b, c, d, e, f, g, h,
A total of nine pieces of information for i are the image values of a 3 x 3 two-dimensional area on the video screen, and since this is updated one after another as it is scanned, it is as if the screen is sequentially scanned through a 3 x 3 two-dimensional window. 3x3 image values are always obtained as if the image was being displayed. This image value a~i
are led to adders 305 and 306 as shown in the figure, their outputs f _x and f _y become f _x = (a + d + g) - (c + f + i) f _y = (a + b + c) - (g + h + i), respectively, in the X direction. A somewhat averaged rate of change in brightness and a somewhat averaged rate of change in brightness in the Y direction are shown. When θ=tan ⁻¹ f _y /f _x is obtained from these two pieces of information f _x and f _y , this can be regarded as the direction of brightness change in the corresponding two-dimensional area.

In order to find this angle θ, arctangent calculation is generally inconvenient. Therefore, as shown in the figure, four adders 307, 308, 309, and 310 are prepared, and f _x , f _y
Enter the following two. That is, each adder calculates pf _x + qf _y , where p and q are coefficients. This value is 0 for the line passing through the origin on _{the f} Therefore, you can choose any position within 360°. Therefore, by appropriately selecting the inputs of the adders 307-310 shown, the four directional lines shown in FIG. 3b can be obtained. Now adder 307
When the binarization circuits 41 to 43 are provided after the circuits 310 to 310, their outputs become 1 when pf x +qf y is positive and 0 when pf _x +qf _y is negative. This means that in FIG. 3b, the shaded area is 1 and the non-shaded area is 0. Follow now 4
If the adders 307-310 are selected as shown in FIG. 3b, the direction is determined from the combination of outputs A, B, C, and D from the binarization circuits 311-314.

That is, as shown in Figure 3c, direction 0 is a region where A is 1 and B is 0, that is, when A and B are 1, and direction 5 is a region where B is 0 and C is 1, That is, when C becomes 1. Therefore, by providing 315 to 322 that perform the eight logical operations shown in FIG.
Up to 7 signals can be obtained.

As a result, a certain picture element group is transferred to the shift register 30.
2,303,304, the logic circuit 315
One of ~322 becomes 1, and the rest become 0.
Therefore, as scanning progresses, the direction of brightness change is detected one after another for each picture element.
When this result is entered into counters 323 to 330,
The total sum in direction 0 is obtained in counter 324, and the sum in direction 1 is obtained in counter 324, respectively. Similarly, counters 325 to 330 provide the sum totals for directions 2 to 7, respectively. In addition, here, the upper direction of the explanation is 8.
However, for practical purposes, it is desirable to use 64 or more directions. Also, this method can easily expand the number of directions.

Next, the center position detection means 203 shown in FIG. 2 will be explained. Since 203 and 205 have the same configuration, only the detection means 203 will be described here. The details are shown in FIG. 4, but here, the coordinates Y ₁ , Y ₂ , X ₁ , and X ₂ of the upper, lower, left, and right edges when the input image is projected horizontally and vertically are determined. The actual center coordinates are calculated in the arithmetic circuit 200. Fourth
The relationship between the timing pulses used in the circuit shown in the figure is shown in FIG. 5a, and a diagram for explaining the x and y coordinate values in FIG. 5a is shown in FIG. 5b. It can be seen from FIG. 5 that the x-coordinate is reset at each end of the horizontal blacking signal, and the y-coordinate is reset at each end of the vertical blacking signal.

Coordinate registers 431 and 43 shown in FIG.
3 are set to _x'nax and _y'nax , which are the maximum values of the x-coordinate value and y- _coordinate value, respectively, at the beginning of one screen in synchronization with the vertical blanking signal V- _BLK by the T1 and T2 signals. Ru. Similarly, coordinate register 4
32 and 434 are set to "0", which is the minimum value of the x-coordinate value and y-coordinate value, respectively. Further, the comparator 441 compares x, which is the current x-coordinate value, with _X1 , which is the previous x-coordinate value recorded in the coordinate register 431, and when x _< Give signal “1”. Therefore, when a binary signal "1" is given as a video signal from the binarization circuit 401, this signal is added to the register 431 as a write pulse. At the time when the write pulse is given to the register 431, the x coordinate value from the signal generator 120 is stored in the coordinate register 4.
31. In this way, the coordinate register 431 is rewritten until just before x≧X ₁ , and the final value is sent to the arithmetic circuit 200. Further, comparators 442, 443 and 444 respectively compare x>X ₂ , y<Y ₁ and y>Y ₂ , and coordinate registers 432 , 433 and 434 are rewritten in the same manner as register 431. .
The final value of each register is then sent to the arithmetic circuit 200 via the data bus 23.

Next, the operation of the arithmetic circuit 200 will be explained. Here, calculations for determining the amount of rotation of the image from the output value of the direction code frequency distribution calculation means and calculations for determining the amount of deviation between the center coordinates of images A and B from the output value of the center position detection means are performed in a time-sharing manner. camera 11
The amount of rotation and translation of the image obtained in step 4 are calculated using the output values of 201 and 203, respectively. Further, the amount of rotation and the amount of translation of the image obtained from the disk 108 are 202 and 202, respectively.
Calculate using the output value of 04. Note that the images recorded on the disk 108 are normalized (aligned) in advance.
In this case, the above-mentioned means 202 and 204 are unnecessary.

Please refer to the flowchart in FIG. It is checked whether a vertical blanking signal (V-BLK) is input to the arithmetic circuit 200 (step 602).
While there is no V-BLK signal, each means 201 to 204 is in operation, so no calculation is performed within the circuit 200 (see FIG. 5). V-BLK to circuit 200
When the signal is input, the count values are read from the direction code counters 323-330 shown in FIG. 3a (step 604). From this count value, image A
Find the maximum frequency direction of the direction code frequency distribution.
On the other hand, the maximum frequency direction of the direction code frequency distribution of image B is determined from the count value of the counter in the means 202 (step 606). The amount of rotational shift of image A with respect to image B is calculated from the direction of maximum frequency of both images A and B (step 608). This amount of rotational deviation is
As understood from FIG. 3c, each counter is installed corresponding to the rotation angle, so it can be easily determined. For example, if the most frequently occurring direction code of image A appears on counter 1 and that of image B appears on counter 2, it can be seen that the amount of rotational deviation between the two images is 30 degrees. Once the amount of deviation is known, a rotation command is given to the image rotation means via the data bus 11 (step 610). Next, the calculation circuit 200 calculates the amount of parallel movement. Read the final value from the coordinate register in FIG. 4 (step 612),
Perform predetermined calculations. (Step 614). This calculation is also executed for the center position detecting means 204 (FIG. 2) to determine the amount of positional deviation between the center coordinates of images A and B (step 616). This shift amount is specified to the parallel shift circuit 205 (FIG. 2) via the data bus 20, and the position shift in the parallel direction is interpolated.

Next, the parallel movement circuit 205 will be explained with reference to FIG.

The switch 730 is a switch for switching between automatic verification mode and manual verification mode, and the illustrated state shows automatic verification mode. Therefore, in this state, the command signal from the arithmetic circuit 200 is sent to the selector 7.
selector 710 for movement in the y direction and x
It is given to the selector 711 for direction movement. Closing switch 730 switches to manual verification mode and command signals from joystick 102 are similarly applied. Here, the case of automatic verification mode will be explained.

The video signal from the camera 114 sequentially passes through a large number of one-line (one horizontal scanning line) delay elements 700 to 703 connected in series. Output lines from each element are output to a selector 710, respectively. Therefore, by selecting one of the outputs based on a command from the arithmetic circuit 200, the y-direction coordinate position of the image A can be moved relative to the image B using an electrical method. For example, when the output of the delay element 702 is selected, the y-coordinate position of the image A is moved by three horizontal scanning lines downward in the drawing in FIG. 5b. Selector 7
The video signals selected by 10 are then sequentially connected to one picture element (unit of x coordinate value).
It is sent to delay elements 720-723. Each element 72
From 0 to 723, each output line is the selector 71
It's in number 1. Therefore, by selecting one of the output lines with the selector 711 in response to a command signal from the arithmetic circuit 200, the image A is
The x-coordinate position of can be moved. The video signal whose y-coordinate position and x-coordinate position have been electrically corrected is supplied to a video synthesis circuit 206 via a line 25.

FIG. 8 shows an example of a video synthesis circuit.
800 is an operational amplifier. The video signal of the image A whose coordinate position has been corrected and the video signal of the image B from the disk 108 are both input to the + input terminal of the operational amplifier 800, where a composite video signal is applied to the monitor television 106 via the line 16. It will be done.

Now, return to Figure 2. The circuit shown in FIG. 2b can be added instead of the video synthesis circuit 206 or in parallel with the circuit 206. Here, 21
0 and 211 are binary coder circuits, 212 is an exclusive OR circuit, and 215 is a control means for taking necessary measures according to the verification result. The video signals representing image A and image B are each binarized,
It is added to exclusive OR circuit 212. When the two inputs do not match, the output of the EOR 212 is "1", so the counter 213 counts the number of times that the two video signals do not match. This count value C is compared with a predetermined threshold value ε, and when C<ε, an output signal “1” indicating the coincidence of both images is set as C≧ε
In this case, an output signal "0" indicating a mismatch between the two images is given to the control circuit 215.

Finally, the image rotation means 110 will be explained with reference to FIG. In the figure, 902 is a pulse motor, which rotates by an angle corresponding to the number of pulses generated by the drive circuit 900 in response to a rotation command given from the image processing means 100. This rotation is transmitted to the prism holding cylinder 904 via gears 906 and 907, and the Wallaston prism 905 housed within the holding cylinder rotates. The image of the document 112 is rotated by the rotation of the prism 905, and is input to the television camera 114 (first
(see figure). In the manual verification mode, the rotation command from the joystick 102 is directly transmitted to the drive circuit 9.
00 and rotates the image in the same manner as described above.

As explained above, according to the present invention, it is possible to perform fully automatic matching of two images, automatically determine whether the two images match or do not match, and take appropriate measures depending on the respective conditions. . Also, two images can be manually compared if necessary. The present invention is particularly effective in comparing a pre-registered seal with a seal stamp stamped on a document.
By using the present invention, it is possible to significantly improve the efficiency of seal verification operations.

In the above-mentioned embodiment, image rotation is performed optically, but if it is desired to avoid the presence of a functional operating section, a two-dimensional local memory (see Japanese Patent Publication No. 51-12492) may be used using a shift register. ), and by controlling the shift timing of this register, it is also possible to electrically rotate the image.

Regarding the setting of the registration number, in the above example, we talked about setting it manually, but in cases such as checks, the number on the check can be set using OCR or MICR.
It is also possible to read the information using an existing reader using an optical or magnetic means called . . . and automatically input it to the video disc.

These alternative methods include means to correct rotational misalignment;
It is not just a means of entering a registration number. Means for storing images, means for reproducing recorded images, means for inputting images, means for determining the amount of positional deviation, means for automatically repairing positional deviations, means for displaying the matching results, which are the constituent elements of the present invention; Various methods of image processing technology should be applied as a means for manually correcting positional deviations depending on the purpose, and the detailed embodiments are merely limited to implementation methods for the sake of easy-to-understand explanation. do not have.

Furthermore, in order to effectively realize the functions according to the purpose, it is naturally possible to delete some of the above-mentioned means.

[Brief explanation of the drawing]

FIG. 1 is an overall configuration diagram of an image matching device according to the present invention, FIG. 2a is a detailed configuration diagram of the image processing means shown in FIG. 1, and FIG. FIG. 3a is a detailed configuration diagram of the direction code frequency distribution calculation means shown in FIG. 2a, and FIGS. 3b and c explain the direction code of a. FIG. 4 is a detailed configuration diagram of the center position detection means shown in FIG. 2a, and FIG.
is a timing chart of the signal provided by the timing pulse generator shown in FIG. 1, FIG. 5b
is an explanatory diagram of coordinate values used in the fourth illustration means,
FIG. 6 is a flowchart of the calculation executed in the calculation circuit shown in FIG. 2a, and FIG.
8 is a diagram showing an example of the image synthesis circuit shown in FIG. 2a, and FIG. 9 is a detailed configuration diagram of the parallel moving means shown in FIG. 1. It is a figure showing an example. 100... Video processing means, 102... Joystick, 104... Registration number input device, 106... Monitor television, 108... Video disk, 110... Image rotation means, 114, 115... Television camera, 11
6, 118... Changeover switch, 120... Timing signal generator, 201, 202... Direction code frequency distribution calculation means, 203, 204... Center position detection means, 200... Arithmetic circuit (microcomputer), 205... Parallel movement means.

Claims (1)

[Scope of Claims] 1. A first means for storing video signals of a plurality of images, and an address of a video signal of a specific image stored in the first means, and a method for storing the video signal of the specific image. a second means for extracting, a third means for inputting a video signal of an image to be compared with the specific image, and a rotation direction and a rotation direction between both images from the video signal of the input image and the video signal of the specific image; a fourth means for detecting a displacement in the parallel direction; a fifth means for optically correcting the displacement in the rotational direction based on the detection result of the displacement in the rotational direction; and a fifth means for optically correcting the displacement in the rotational direction; a sixth means for correcting the positional deviation in the parallel direction based on the detection result; a seventh means for displaying the specific image and the corrected image in an overlapping manner on the same screen; In response to the operator's rotation command and parallel movement command, a command signal to correct the positional deviation in the rotational direction is sent to the fifth means, and a command signal to correct the positional deviation in the parallel direction is sent to the sixth means. An image matching device characterized in that it has an eighth means for. 2. A first means for storing video signals of a plurality of images, and a second means for specifying the address of the video signal of a specific image stored in the first means and retrieving the video signal of the specific image. and a third means for inputting a video signal of an image to be compared with the specific image, and a positional shift in the rotational direction and parallel direction between the two images from the video signal of the input image and the video signal of the specific image. a fifth means for correcting a positional deviation between the two images based on the output of the detecting means; and a fifth means for detecting a mismatch area between the specific image and the corrected image. , and a sixth means for determining whether or not both images match based on the detection result.