JPH02210591A - Picture processor - Google Patents

Abstract

PURPOSE:To discriminate a specified picture without fail by detecting the positioning state of an original, taking out the picture data of a specified part in the original and discriminating the degree of similarity between the data and a pattern which is registered in advance. CONSTITUTION:Detecting means 210-1 to 210-3 are provided to detect the positioning state of the original and a discriminating means 417 is provided to take out the picture data of the specified part in the original in correspondence to the detection of these detecting means 210-1 to 210-3 and to discriminate the degree of the similarity between the data and pattern which is registered in advance. Thus, for example, the position and angle of a bank note, etc., are detected according to a base part, a red mark is detected and pattern matching is executed. Accordingly, the specified picture can be discriminated without fail so that the forged bank note set in the arbitrary position and angle can be prevented from being used in advance or the like.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention particularly relates to a copying machine provided with a counterfeit prevention function.

[Conventional technology]

2. Description of the Related Art Conventionally, with the increase in image quality and color copying machines, there has been a particular concern about counterfeiting of banknotes. On the other hand, various methods have been devised for recognizing banknotes, etc., such as Jitsuma No. 50-60090.

[Problem that the invention is trying to solve]

However, when this is applied to a copying machine, for example, it is difficult to detect a banknote at any arbitrary position and angle on the document table of the copying machine, and as a result, it is difficult to prevent counterfeiting. There was a problem.

In view of this problem, it is an object of the present invention to provide an image processing device for reliably identifying a specific image.

[Means to solve the problem]

In order to achieve the above-mentioned object, the present invention includes a detection means for detecting the positional state of a document, extracts image data of a specific portion of the document according to the detection by the detection means, and extracts image data of a specific portion of the document, and compares the similarity with a pre-registered pattern. and determines a specific document based on the degree of similarity.

〔Example〕

The present invention will be explained below based on preferred examples.

In the following embodiments, a copying machine is shown as an example of application of the present invention, but it is needless to say that the present invention is not limited to this and can be applied to various other apparatuses.

FIG. 2 shows an external view of the apparatus according to the first embodiment of the present invention.

In FIG. 2, reference numeral 201 is an image scanner section, which reads an original and performs digital signal processing. Further, 202 is a printer section, and an image scanner 20
This part prints out an image corresponding to the original image read in step 1 on paper in full color.

In the image scanner unit 201, 200 is a mirror pressure plate, and an original 204 on an original table glass (hereinafter referred to as a platen) 203 is irradiated with a lamp 205, and a mirror 206,
207 and 208, an image is formed on a three-line sensor (hereinafter referred to as CCD) 21O by a lens 209, and the image is sent to the signal processing unit 211 as full color information red (R), green (G), and blue (B) components. In addition, 205, 20
6 is a speed V, and 207 and 208 are mechanically moved at a speed of 1/2 V in a direction perpendicular to the electrical scanning direction of the line sensor to scan the entire document surface.

The signal processing unit 211 electrically processes the read signal, decomposes it into magenta (M), cyan (C), yellow (Y), and black (Bk) components, and sends them to the printer unit 202. Furthermore, M, C, and Y are used for one document scanning (scanner) in the image scanner unit 201.

One component of Bk is sent to the printer section 202,
One printout is completed by scanning the document four times in total.

M sent from the image scanner unit 201.

The C, Y or Bk image signal is sent to the laser driver 212. The laser driver 212 modulates and drives the semiconductor laser 213 according to the image signal. The laser beam scans the photosensitive drum 217 via a polygon mirror 214, an f-theta lens 215, and a mirror 216.

Reference numeral 218 denotes a rotary developing unit, which is composed of a magenta developing unit 219, a cyan developing unit 220, a yellow developing unit 221, and a black developing unit 222. Develop the electrostatic latent image with toner.

223 is a transfer drum, and paper cassette 224 or 225
The paper fed from the transfer drum 223 is wound around the transfer drum 223, and the image developed on the photosensitive drum 217 is transferred onto the paper.

After the four colors M, C, Y, and Bk are sequentially transferred in this manner, the paper passes through the fixing unit 226 and is discharged.

In the following, this embodiment will be explained in detail using the counterfeit prevention function of a 10,000 yen Bank of Japan note as an example of a manuscript subject to counterfeit prevention, but the present invention is not limited to such a target. Of course.

In this embodiment, as described above, copying is performed in a total of four reading operations (scans), and the image scanner 201 and printer 2 at each scan
The operation of 02 is shown in FIG.

In other words, if an attempt is made to prevent counterfeiting of 10,000 yen bills, the first
At the time of the second scan, the image scanner
In mode 1, the approximate position of the 10,000 yen bill is detected, and the printer section outputs magenta.

During the second scan, the image scanner is in mode 2 and detects the exact position and angle at which it is placed.

During the third scan, the image scanner is in mode 3, and the red stamp 1 of the 10,000 yen bill is calculated from the position and angle detected during the second scan.
Calculate the position of , extract the red stamp and store it in memory,
Determine whether it is a red seal or not. The printer section outputs yellow.

During the fourth scan, the image scanner is in mode 4, and if it is determined that forgery is attempted during the third scan, specific countermeasures are taken to prevent forgery.

FIG. 5 is a timing diagram showing the operation of each part in the image scanner section of this embodiment.

The VSYNC signal is an image valid interval signal in the sub-scanning direction, and in the interval of "1", output signals of (C), (M), (Y), and (Bk) are generated sequentially for image reading (scanning). do. VE is an image valid section signal in the main scanning direction, and the image becomes valid in the "1" section. H8YN
The C signal is a main scanning synchronizing signal, and takes the timing of the main scanning start position in the "1" section. The CLK signal is a pixel synchronization signal, and image data is transferred at the rising timing of 0→1. CLK 16 is a timing signal for every 16 pixels, and at the rising timing of 0 → 1, 16 as described later
Take the timing of the X16 block processed signal.

[Image Scanner Section] FIG. 4 shows the internal blocks of the image scanner 201. 210-1.210-2.210-3 is a COD with red (R), green (G), and blue (B) spectral sensitivity characteristics, respectively, and has a signal quantized to 8 bits from 0 to 255. Output.

Sensor 210-1°2 used in this example
10-2 and 210-3 are placed apart from each other as shown in FIG. 1, so their spatial deviations are corrected in delay elements 401 and 402.

403, 404, and 405 are log converters, each of which is configured with a look-up table ROM, and converts a luminance signal into a density signal. 406 is a well-known masking and UCR circuit, and although a detailed explanation will be omitted, it outputs Y and M signals based on input three primary color signals.

C and Bk signals are sequentially output with a predetermined bit length, for example 8 bits, each time each reading operation is performed.

407 is an OR gate circuit, which performs logical OR with the value held in the register 408. Normally, 00H is written in the register 408, and the output of 406 is output as is to the printer section, but during forgery prevention processing, the CPU 417 sets F F H in the register 408 via the data bus. By doing so, you can output an image filled with toner.

A CPU 417 controls the device in each mode. 408 is a window comparator, and the CPU 41
It is determined whether or not a specific level signal instructed by 7 has been input, and in modes 1 and 2, the background level is detected, and in mode 3, a red stamp is detected. The blow processing circuit 409 performs 16×16 block processing and processes the output of the window comparator 408 for each 16×16 block.

412 is a read/write random access memory (RAM).
), the data is switched in the selector 411, and the address is switched in the selector 413.

On the other hand, 419 is a main scanning counter, which is reset by the H3YNC signal, counts up at the timing of the CLK signal, and generates 13-bit main scanning addresses (hereinafter referred to as X addresses) x12 to XO.

A sub-scanning address counter 420 is reset during the "0" period of the VSYNC signal, counts up at the timing of the H3YNC signal, and generates a 13-bit sub-scanning address (Y address) Y12 to YO.

The CPU 417 selectors 411 and 41 according to each mode.
3.415.416 controls the address decoder 414 and reads and writes data to and from the RAM 412.

418 is a RAM/ROM added to the CPU 417. 410 is a watermark detection circuit that detects a watermark portion of a banknote.

FIG. 6 is a block diagram of window comparator 408 shown in FIG. 601. Registers 602, 603, 604, 605, and 606 are directly connected to the data bus of the CPU 417, and desired values are written by the CPU 417.

607, 608.609.6], 0.611.612 are comparators, and the output becomes 1 only when A>B holds true for inputs A and B.

613 is an AND gate, which outputs 1 only when the outputs of all the comparators become 1, and otherwise outputs O. Now, registers 601, 602, 603,
604. 605. 606 and RH and R1-, respectively.

When the values GH, GL, BHI, BL are written, RL < R < RH, GL < G < GH, and BL < B < Bu all hold true for the input signals R, G, and H. The output of the window comparator 408 becomes "L" only when the signal is on, and becomes "0" at other times.

FIG. 7 is a diagram showing internal details of the block processing circuit 409 shown in the fourth diagram.

701, 702.703. ..., 714.715 is 15 D flip-flops arranged in series (hereinafter referred to as D flip-flops).
FF), which sequentially delays the input signal with the pixel clock CLK signal, and is cleared to VE="0", that is, "0" in the non-image section.

738 is a 5-bit up/down counter, 737 is O
The R gate 740 is an AND gate whose operation is based on the following table.

That is, the output of counter 738 is VSYNC or VE.
is cleared to 0 in the “0” interval, and X [-X L-I
6, it is held, and X t = 1 or X t-+
When 5 = 0, it is counted up and becomes X [20 k x
When t15=1, it is counted down.

That is, as a result, the sum of 16 data (=number of 1) of X + -X t-+s is output.

Furthermore, the output passes through DFF739 and is sent to a first-in first-out memory (FI
FO) 721, 722, 723. ..., 734, 73
5, 16 lines of data are simultaneously input to the adder 741, and the total sum is output. As a result, 16×
The total number of 1s in 16 windows SUM is 0 to 25
6 is output.

742 is a digital comparator, and adder 741
The output SUM is compared with a comparison value TW predetermined by the CPU 417, and the result is output as "0" or "1".

Therefore, by setting an appropriate value in TW in advance, 1
Noise removal can be performed in units of 6×16 blocks.

FIG. 8 is a diagram illustrating the watermark detection circuit 410.

819.820. . , 821 has 18 FIFOs, each giving a delay of 1 line, and 19 lines being processed simultaneously.

801.802,803.・,804,805,80
6゜807, ・,80B, ・,809,810,
811. -...812,..., 813.814.8
15. ..., 816 are DFFs arranged in series for each of the 19 lines, and...
817 and 818 are nine DFFs further arranged in series after the DFF 812, and all DFFs are driven by CLK16. AND gate 823, 824, 825
After that, DFF804,808°..., 812.8
16 (vertical 19) block) and DFF809°81
0, 811.812.817.818 (horizontal 19 blocks) are all "1", 1 is output, and Po5ition at that time is latched and sent to the CPU.

FIG. 9 is a block diagram of address decoder 414.

901, 902.903.904.905.906.9
07°908 and 909 are registers connected to the data bus of the CPU, into which desired values are written by the CPU.

910, 911, 912, and 913 all have the same structure, so only 910 will be explained.

914 and 915 are subtracters, which output A-B for inputs A and H, and the MSB of the output is a sign bit;
If it becomes negative, it is output as MSB=1.

916.917 is a comparator, input A, B
On the other hand, when A<B, "1" is output. However, if the MSB of A is "1", "0" is output regardless of the input of B.

918, 919, and 920 are AND gates, and as a result, the following is output only when the value BXY is written to the register 909.

Similarly, in 911 Only when X2 = X address - RX2 Y2 = Y address - RY2 E2=1 The following output is output.

Similarly, in 912 Only when X3 = X address - RX3 Y3 = Y address - RY3 E3=1 The following output is output.

Similarly, in 913 E4 = 1 The following output is output.

914, 915.916.917.918 are each O
It is an R gate, and conditions (1), (2), (3)
, (4) is RXI so that two or more do not hold at the same time.
, RYI,...RX4°When RY4 is set, the 914 outputs one of XI, X2...X4, and the 917 outputs "1" when either condition is met. ”
is output, and one of Yl, Y2...Y4 is output from 918.

Also, 915. Carbon is emitted from 916.

Only when X4 = X address - RX4 Y4 = Y address - RY4 [Processing flow]
The four scans up to the first scan, the four modes from mode 1 to mode 4 of the image scanner, and the output contents of the printer section are shown.

FIG. 1O shows a processing flow controlled by the CPU. In FIG. 1O, at 1001, the CPU sets mode 1 for the first scan. In mode 1,
The printer outputs magenta and also detects the approximate position of the watermark in the bill. For this reason, the CPU uses the window comparator 408 to
Values that allow the background of the bill to be detected are written into registers 601 to 606 shown in FIG.

In this state, the first scan is started at 1002.

Figure 1 shows a 10,000 yen bill placed on the manuscript table.
In the first scan, that is, in mode I, the output of the window comparator becomes "1" in the shaded area in FIG. 1, that is, in the area surrounding the watermark.

Furthermore, in the watermark part, this continues for more than a certain number of pixels vertically/horizontally, so the watermark detection circuit 410 in the part corresponding to (Xc, Yc) in FIG.
The address for Yc) is latched and sent to the CPU.

The CPU can roughly know the value of (Xc, Yc), which is the center of the card.

Next, in FIG. 10, setting for mode 2 is performed at 1003. That is, in mode 2, selector 411
is set to A, selectors 413 and 415° 416
is set to the A side, and in the address decoder 414, RX2=RX3=RX4=RY2=RY3=RY4=0
Write the value in advance.

At this time, in the second scan of 1004, R
In AM412, by setting T to a value corresponding to about 128 mm, it is possible to obtain a bitmap in which only the diagonally shaded area is "1" within the broken line area. The value of 128 mm is set as a value such that the banknote is included within the broken line shown in FIG. 1, regardless of the type of banknote or the angle at which the banknote is placed.

Furthermore, the CPU 417 detects the exact position and angle of the bill from this bitmap at 1005. At that time, CP
U selects selector 411 to C, 413 to B,
Read the contents of RAM 412 freely. The principle is shown in FIG. 10, and the method is to detect the four corners around the bill. That is, when placed as shown in FIG. 11, the four corners are detected using the fact that among the portions of the bitmap that are 1, the following also occurs.

Although a detailed explanation will be omitted, by selectively using either condition (5) or (6), the four corners can be detected no matter what angle they are placed.

Therefore, in Figure 1 The angle θ is also calculated in the equation.

Furthermore, the position of the red seal is determined from the obtained angle and position. Since the banknotes are placed facing up, facing down, or on the back/front, (Xs+ in Figure 1).

Ys+), (XS2.YS2), (Xsa,
Either YS3) or (XS4. YS4) is the position of the red seal. These are simple calculations (Xo, Y
o) and θ.

Next, in step 1006, mode 3 for the 3rd scan is set. That is, the selector 411 is B
, the selector 413 is set to A, and the selectors 415 and 416 are set to B.

In the address decoder 414, the position of the red stamp is RX1=Xs+RY+=Y
s+RX 2: X S2 RY 2
= Y S2RX 3 = X s3 R
Y s = Y S3RX 4 = X S4
It is set so that RY 4 = Y 54 (one pixel unit), and the number of pixels corresponding to about 30 mm, which sufficiently covers the red stamp in FIG. 1, is set in BXY.

Next, values are set in each register in the window comparator 408 shown in FIG. 4 so that only the red part of the red seal outputs l.

Next, at IQ07, a third scan is performed, and the output of the window comparator 408 is written into the RAM 412 at four locations considered to be the positions of the red stamps in FIG.

Further, in 1008, a red stamp is extracted by an algorithm described later, and judged in 1009. If there is no possibility of forgery, that is, if the red stamp is not detected,
A 4th scan is performed at 1010, and the image is developed with four color toners of Y, M, C, and Bk in the normal operation at 101.
2, the image is fixed and output.

On the other hand, if it is determined in 1009 that there is a possibility of forgery, that is, if a red stamp is detected, counterfeit prevention etc. are taken in 1011. Specifically, FIG. 4 408
By setting F F I-1 in the register (usually OOH is set), F F is sent to the printer section.
H is sent and Bk toner adheres to the entire surface, making it impossible to copy.

[Pattern Matching] Next, pattern matching of the red stamp 1008 will be described in detail.

Note that the red stamps on the front and back of banknotes are different, so in order to distinguish one type of banknote, two steps are required when performing pattern matching.
Two red stamp patterns are registered in advance.

When a specific part of a specific manuscript is written to the RAM 412,
Next, the CPU 417 refers to the contents of the RAM 412 and performs a pattern matching operation. A flowchart of pattern matching is shown in FIG. Four types of binary data are stored in the RAM 412 as candidates for specific portions. These are named Area 1 to Area 4.

First, processing 2102 and subsequent steps are performed for area 1. In 2102, window processing is performed to remove noise.

Assume that the binary image of area l is 2201. Here, a small square represents one pixel, a blank pixel is a white pixel,
Assume that the shaded area is a black pixel. This is shown as 22022
A window of ×2 pixels is scanned, the number of black pixels within the window is counted, and a portion where the count value exceeds 2 is newly defined as a black pixel. By doing this, the processing result is 2203
As shown in the figure, a pattern is obtained which has been reduced both vertically and horizontally, and from which noise has been removed. The number of black pixels in the window at position 2202 is 1, so 2204 is the white pixel.
It has been moved to the position of

Next, the center of gravity position of the pattern 2203 is calculated.

This can be calculated by a well-known method by projecting the pattern 2203 in the vertical and horizontal directions.

Next, the degree of similarity is calculated by standard pattern matching. First, in step 2105, a standard pattern previously registered as a dictionary is read into the CPU from the ROM 418 shown in FIG. The standard pattern refers to the red stamp pattern of the current banknote, but the patterns extracted up to 2103 may have been rotated depending on the angle at which the banknote is placed on the platen, and this can be considered as a single standard pattern. Comparing with the pattern does not give satisfactory results.

This state is shown in FIG. Therefore, as a standard pattern, you can create multiple patterns in which the red seal pattern is rotated every few positions, store them in advance in the ROM, and then select an appropriate pattern from among these and read it into the CPU. . As the plurality of patterns, for example, a total of 24 patterns are used, in which the red stamp is rotated every 15 degrees from 0 degrees to 360 degrees, and the selection method is based on the angle θ at which the banknote is placed, which has already been detected. , -8≦θ≦8 degrees, the rotation angle is the standard pattern of φ degrees, when 8≦θ≦24 degrees, the rotation angle is 15 degrees, and when 24≦θ<40 degrees, the rotation angle is 30 degrees. The pattern of...
If this is done, similarity calculation can be executed with sufficient accuracy.

Here, there is clearly an uncertainty of 180 degrees between the rotation angle θ of the banknote and the rotation angle of the red seal, so in reality, if the standard pattern with a rotation angle φ degree is selected, the rotation angle θ is 180 degrees. It is also necessary to select a standard pattern; similarly, when the angle is 15 degrees, 195 degrees must be selected, and when the angle is 30 degrees, 210 degrees must be selected at the same time, so the similarity calculation must be performed twice.

Next, in 2106, the degree of similarity is calculated. Various methods can be considered for calculating the degree of similarity. For example, the following methods are possible. As shown in FIG. 15, let the pattern extracted up to the above be (a) or the standard pattern with a predetermined rotation angle selected by the above method be (b), and respectively represent B(1+])+P(++j). represent (B
(i, D, P (i, D takes the value of 1 when it is a black pixel, and φ when it is a white pixel) Also, the barycenter coordinates of B (i, D obtained in 2104 in FIG. 12 are expressed as (i BC, j Bc
), the barycentric coordinates of P (i, D obtained in the same way as (i
pc + jpc), the similarity COR between the two is given by the following equation.

COR-ΣΣP (i-4pc, j jpc) OB (
i -1nc, j jBc) --(1)j ■ represents the exclusive OR of P and B, and equation (1) is pattern B (i, D and pattern P (when the centers of gravity of i, D are aligned) The larger the COR, the greater the similarity between the two.

In this embodiment, in order to improve the reliability of the similarity and to minimize the occurrence of erroneous recognition expressions, the similarity COR is calculated using equation (2), which is a modification of equation (1).

COR= 2x (IX:P(i-ipc, j-
jpc) ・B (i-4nc, j-jsc)) J (ΣΣ舊工] Akira Roar B (i-iBc, j-jnc)
) - (2) J where ・ is logical product, P 1-ipc, j-jpc
represents the determination of P, and when both P and B are black pixels, COR is added by 2, and P=0. When n=1, COR
By subtracting 1 from

When the degree of similarity COR is calculated as described above, in 2107, the COR is compared with a threshold Th obtained in advance.

If COR>Th, it is determined that a red stamp exists, and the verification operation ends as a banknote exists (2108).

If COR<Th, it is determined that there is no red stamp in the current processing area, and if n is smaller than 4, n is added by 1, and 2101 to 21 is applied to the next area n.
Repeat the verification operation in step 09. If n=4 in 2109, it means that there is no red stamp pattern in any of the first to fourth areas, and the verification operation ends as there is no banknote (2110).

[Other Examples] This specification has described the prevention of counterfeiting of 10,000 yen Bank of Japan notes, but its application range is wide, including relatively wide backgrounds and characteristic patterns such as red stamps and patterns. If so, it can easily be applied to securities, foreign securities, newly revised securities, etc.

Moreover, it is not limited to electrophotographic copying machines, but is also important as a simple image scanner or an additional function as a standalone printer.

Furthermore, in this embodiment, the size and position of the document placed on the document table were determined before performing pattern matching with a pre-registered pattern. Even in an apparatus in which the document is sequentially fed to a reader section, the document may be reciprocated relative to the league section so that the same document can be scanned several times.

As explained above, according to this embodiment, by detecting the position and angle of a banknote, etc. from the background part, detecting a red stamp, and performing pattern matching, a banknote, etc. placed at an arbitrary position and angle can be detected. counterfeiting can be prevented.

[Effects of the Invention] As explained above, according to the present invention, it is possible to reliably identify a specific image.

[Brief explanation of the drawing]

Fig. 1 is a diagram showing a banknote placed on a document table, Fig. 2 is an overview of the device of this embodiment, Fig. 3 is a diagram showing the relationship between four scans and each mode, and Fig. 4 is a diagram showing the relationship between four scans and each mode. A block diagram of the image scanner, FIG. 5 is a timing diagram, FIG. 6 is a block diagram of the window comparator shown in FIG. 4, FIG. 7 is a diagram of the block processing section shown in FIG. 4, and FIG. 4 is a diagram of the watermark detection unit shown in FIG. 4, FIG. 9 is a diagram of the address decoder shown in FIG. 4, FIG. 10 is a flowchart showing the operation of the device, FIG. Figures 1 to 15 are diagrams for explaining pattern matching. 210-1 to 210-3 ・・・・・・・・・CCD
408 ・・・・・・・・・・・・・・・・・・
・・・・・・・・・・・・・・・ Wind comparator 412・・・・・・・・・・・・・・・・・・
・・・・・・・・・・・・・・・・・・RAM417・
・・・・・・・・・・・・・・・・・・・・・・・・
・・・・・・・・・・・・CPUul) Additional fertilizer 3 pattern B(2,)) (ρ2 Rotepa9-7 p (i, l) (b)

Claims (6)

[Claims] (1) It has a detection means for detecting the positional state of the document, and a determination means for extracting image data of a specific part of the document according to the detection by the detection means and determining the degree of similarity with a pre-registered pattern. and an image processing apparatus that determines a specific document based on the degree of similarity. (2) The image processing apparatus according to claim 1, wherein the specific document is a banknote, and the specific portion is a portion including a red stamp. (3) The discrimination means detects the approximate position where the banknote is placed by detecting the watermark part of the banknote, and determines the exact position and angle at which the banknote is placed by detecting the four corners of the banknote. The image processing apparatus according to claim 2, wherein the image processing apparatus is a calculating means. (4) The image processing device performs one mechanical scan multiple times.
Claim 1: The apparatus for completing one image reading operation, wherein the similarity is determined before the start of the final mechanical scanning to determine that a specific document is placed.
) image processing device. (5) The image according to claim (2), wherein the registered pattern is a pattern selected from at least one of a plurality of registered patterns with reference to information on the angle at which the banknote is placed. Processing equipment. (6) The degree of similarity is determined by comparing the specific partial pattern to B(i,
j), and the standard pattern stored in advance as P(i
, j), and a, b, c, and d are predetermined constants, and is calculated by the following formula:
The image processing device described. Similarity = ▲Mathematical formulas, chemical formulas, tables, etc.▼ (However, . represents logical product and - represents inversion).