JPH01226086A - Oblique line detecting system - Google Patents

Abstract

PURPOSE:To correctly detect an oblique state by scanning an image in horizontal and vertical directions, finding out image existence information, finding out the side information of an input medium, deciding whether the sides intersect at right angles or not, and when the sides of the lines intersect at right angles, calculating the oblique angle of the input medium. CONSTITUTION:Input medium data 11 is scanned in the horizontal direction and the lengths of a white area (no image part) and a black area (image part) are measured. A continuity recognizing part 3 decides whether a side L1 is continued or not by data read out by a horizontal direction processing part 1. In a lacking part F1, a straight line passing a point A in a start area detected at its continuity and a point B in its end area is defined as a side L1. Then the input medium data 1 is scanned in the vertical direction by a vertical direction processing part 2, its continuity is decided by the recognizing part 3 and a side L2 is obtained in a lacking part F2. Whether the obtained sides L1, L2 intersect at right angles or not is decided by an orthogonalization detecting part 4, and when the sides intersect at right angles, an inclined angle theta is calculated by an inclined angle calculating part 5 about either one of the sides L1, L2.

Description

[Detailed description of the invention] [Table of contents] Overview Industrial field of application Prior art (Figure 6) Means for solving the problems to be solved by the invention (Figure 1) Working examples (Figure 2) ~Figure 5) Effects of the invention [Summary] To enable correct detection of the skew state even when there is a missing part or a bent part on the side, in relation to a skew detection method in image information processing. With the aim of An image recognition device that recognizes an image includes a horizontal direction processing section, a vertical direction processing section, a continuity recognition section, an orthogonality detection section, and a skew angle calculation section, The horizontal processing section performs scanning processing in the horizontal direction to obtain image presence/absence information, the vertical processing section performs scanning processing in the vertical direction to obtain image presence/absence information, and based on these, the continuity recognition section determines the presence/absence information of the input medium. It is characterized in that the side information of the input medium is obtained, the orthogonality detection section determines whether the sides are orthogonal, and when the sides are orthogonal, the oblique angle of the input medium is calculated.

[Industrial application field]

The present invention relates to a skew detection method, and in particular, when a medium to be image-recognized is input to a recognition device in a skewed direction, even if there is a defective part on the side, the skew is accurately detected. Regarding things whose angles can be calculated.

For example, there is an image recognition device that uses a sensor to read an image of a medium such as a banknote in order to determine the authenticity of the banknote. This basically stores the image data of the banknote in advance as XI goons, and compares the read image data with the image data of the basic pattern to determine the authenticity of the banknote. Image data of the basic pattern of a banknote (hereinafter referred to as basic pattern data) is obtained from a banknote that is input orthogonally to a reading unit such as a line sensor that reads the image, but in reality, it is The banknotes are input into the reading section in various directions (angles and inclinations).

Therefore, when actually determining authenticity, it is necessary to perform some form of notching between the read data and the basic notation in order to perform accurate verification.

[Conventional technology]

Accurately match the basic pattern with the object to be verified, such as a bill that is being input obliquely. For the image of a banknote, etc. below, patterns with various oblique input angles are registered in advance as the basic pattern. Just leave it there. In other words, basic pattern data corresponding to various oblique input angles are registered in advance so as to correspond to all input directions of banknotes that are actually input, so that the input data can be immediately responded to. This is fine, but there are problems as described below.

For example, as shown in FIG. Data of medium P, such as, angle θ
Data of medium P3 input at b, medium P input at angle θC. Data for the number of angles required for recognition will be prepared.

Then, the data of the medium such as the banknote to be determined read by the reading unit is compared with the basic pattern data registered in these memories.

[Problem to be solved by the invention]

This method is not particularly difficult technically, as it only requires registering a large number of basic patterns.
In reality, input banknotes are input from various angles, and a problem arises in that an extremely large number of basic patterns are required to be registered for smooth determination. Moreover, it is necessary to calculate the skew angle so as not to be affected even if the input banknote has a bent part or a missing part.

This invention has been made in view of these points, and can reduce the number of basic patterns to be registered, and can also handle media such as banknotes that are input from various angles. The purpose of this invention is to provide an image recognition correction method.

[Means to solve the problem]

FIG. 1 is a diagram showing the principle of the invention. In the figure, 10 is an image memory that stores medium data from a reading unit such as a line sensor (not shown);
As shown in 1), this memory stores data on the medium, that is, the object to be verified, which is input to the reading section at an angle e. Reference numeral 15 denotes a memory for storing basic pattern data, and as shown at 16, the basic pattern data of the original image to be compared is stored in a form input orthogonally to the reading unit. There is.

To compare input data with basic pattern data,
First, find the input angle e to the reading section of the medium from the input medium data 1), and calculate from this θ the data string 1)-1)1-2, -1)-n in the direction parallel to the edge of the medium.
seek. This is converted into data strings 16-1, 16-2, - of the basic pattern data 16 stored in the memory 15.
- The authenticity of the input medium is determined by comparing it with 16-n.

In this case, even if there are missing portions F ls, F z or bent portions on the sides, it is necessary to accurately determine the input angle θ.

Therefore, in the present invention, as shown in FIG. 1(C), the horizontal processing section l, the vertical processing section 2, the continuity recognition section 31,
An orthogonality detection section 4 is provided.

As shown in FIG. 1(D), the horizontal processing unit 1 scans the input medium data 1) in the horizontal direction and counts the lengths of the white area (portion without image) and black area (portion with image). do.

The continuity recognition unit 3 determines whether or not the side L1 is continuous based on the data read by the horizontal direction processing unit l. That is, when horizontal scanning is performed at the Y coordinate n, the X coordinate at which the input medium data 1) is first detected is W(n+), and n,
Let W(nz) be the same X coordinate when horizontally scanning at the Y coordinate n2 which is separated by a unit length (for example, 1 bit), then find out how many bit lengths W(n+) and W(nz) differ from each other. Judge and check continuity. In the normal case, since θ is about 6 degrees at most, W (n +) - W (n2), and in the case of 6 degrees, it is the same even if 10 bits apart in the Y coordinate, 1)
This is the extent to which the X coordinate changes by 1 bit apart.

However, in the region of the missing part F, n, which is 1 bit apart,
Since there is a difference of more than 1 bit between w(n3) and w(n4) of do.

Similarly, the input medium data 1) is scanned in the vertical direction by the vertical direction processing unit 2, and based on this, the continuity recognition unit 3 similarly determines continuity and removes the missing portion F2 to obtain the side L2. Can be done.

The orthogonality detection unit 4 determines whether or not Ll and L2 obtained in this way are orthogonal, and if they are orthogonal, the oblique angle θ is adjusted by It can be calculated by the line angle calculating section 5.

[Effect]

In this way, the oblique angle e can be adjusted without being affected by missing or bent parts.
can be found.

Once this θ is determined, data is read out from the input medium data 1) stored in the image memory 10 in a direction parallel to the edge of the medium to obtain data sequences 1)-1, 1)-2, -41-n. . This data string 1)-1.1)-2, -1)-n is
Since the data is in the same direction as the data obtained from the medium input perpendicularly to the reading section, it can be directly compared with the only basic pattern data stored in the memory 15.

For this reason, the basic pattern data that should be registered is
Data input perpendicular to the reading unit may be used.
Not only can the memory required to store basic pattern data be significantly reduced, but also image recognition can be performed accurately.

〔Example〕

An embodiment of the present invention will be described based on FIGS. 2 to 5.

Figure 2 (A) is a configuration diagram of an embodiment of the present invention, Figure 2 (B) is a diagram explaining sensor data, Figure 2 (C) is a diagram explaining line unit data, Figure 3 is an example of a reading section, and Figure 4 is An explanatory diagram of the operation of the present invention, FIG. 5 is an explanatory diagram of input medium data.

In FIG. 2, the same reference numerals as in FIG. 1 indicate the same parts, 6 is a continuity detection section for detecting continuity, 7 is a counter, 8 is a flag section, 9-1 is a starting point holding section, 9- 2 is an end point holding section, 20 is a control section, 21 is a processor, 22 is a sensor section, 23 is a sensor data reading section, 24 is a RAM
(Random Access Memory), 25 is an external interface section.

The horizontal processing section 1 horizontally scans the image memory 1O as shown in FIG. 2(B) to create line unit data as shown in FIG. 2(C). Line unit data includes data number division d3, first dot-free number division d2, and first dot-free number division d2.
It is composed of a dot count division d1, a second dot count division d4, etc., and other dot count divisions d3, d3, etc. as necessary.
d6... is prepared. Here, data number classification d,
indicates the sum of the white area, that is, the no-dot section, and the black area, that is, the dot section that exists in the horizontal scanning line, and the first no-dot number section dt indicates the dot length of the first white area of the horizontal scanning line. , the first dotted section d indicates the dot length of the first black area, and the second dotless section d4 indicates the dot length of the second white area.

Therefore, in FIG. 2(B), since the horizontal scanning line H1 is an all-white area, the first
Line unit data is created in which the number of no dots is the number in the horizontal direction of the image memory 10, and for H2, dl=3, the first number of no dots is WI, the first number of dots is B, and the second number of no dots is B. Line unit data of W2 is created, and for H3, d1=5, the first number of no dots is Wl', the first number of dots is B,' (not shown), and the second number of no dots is W2'. , the second number of existing dots is W, and line unit data is created using the dot number divisions d3 and d6. Each of these line unit data is held in the RAM 24. Note that the image memory 10 also uses a part of the area of the RAM 24.

Counter 7 is used to count the number of these dots.
is used and reset at the end of the W and B sections and the horizontal scan line, and the previous count value will be written into each dot number section.

The vertical processing unit 2 performs scanning processing on the image memory 10 in the vertical direction, and similarly creates line unit data. Therefore, in FIG. 2(B), the vertical scanning line ■
Line unit data of data number d = 5 is created for 1, and data number a I = 3 for 2 is created.

Based on the respective line unit data created by the horizontal direction processing section 1 and the vertical direction processing section 2, the continuity recognition section 3 performs
It identifies continuity on each side L1 to L4,
It includes a continuity detecting section 6, a flag section 8, a starting end holding section 9-1, a terminal end holding section 9-2, and the like.

Note that the continuity recognition section 3 and the orthogonality detection section 4 are characteristic features of the present invention, and their operations will be described in detail later.

The control unit 20 performs comprehensive processing in the present invention, and also holds the size NXM of the image memory 10.

The sensor section 22 is for reading the medium when it is conveyed. The sensor section 22 is configured as shown in FIG. 3, for example. A table 35 on which the sensor unit 22 medium is placed and conveyed;
It consists of sensor arrays 3136 arranged on both sides.

In this case, the stand 35 is made of a transparent material,
Both sides of the medium can be read simultaneously. The sensor array consists of a light emitting element section 33.37 such as an LED and a light receiving element section 34.38 such as a photosensor. Light emitting element section 33.
The data on the medium can be read by reflecting the light emitted from the medium 37 on the front and back surfaces of the medium and receiving the reflected light at the light receiving element portions 34 and 38.

Alternatively, the structure may be such that the light from the upper light emitting element is guided to the medium and the transmitted light is received by the lower light receiving element. That is, as shown by the two-dot chain line in the figure, light from the upper light emitting element that has passed through point A1 of the data sample line of the medium is received by the lower light receiving element. This method using transmitted light is most suitable for detecting the presence or absence of a medium, and can also be used to determine the input angle of the medium of the present invention.

Note that the arrow in the figure indicates the traveling direction of the medium on the medium platform 35.

The medium is read by such a sensor section 22, processed into binary information by the sensor data reading section 23, and then RA
The images are sequentially stored in the image memory 10, which is a part of the M24.

The external interface 25 performs input/output control with external devices, and connects with other input/output devices, a host device, and the like.

Next, the operations of the continuity recognition section 3 and the orthogonality detection section 4 will be explained.

First, the operation of the continuity recognition unit 3 for finding a straight line around , will be explained with reference to another country according to the flow shown in FIG. 4(A).

The continuity detection unit 6 transmits the data to the control unit 20 from the data shown in FIG. 2 (C
) is read out in sequence, and the data number classification d is checked. When this first shows 3, it means that input medium data 1) has been detected, so
The first dot-free number h (i) is read and held. Then, read the first number of non-dots h (i+l) of the line unit data following one dot, and read the difference h (i) - h
Calculate (i+1). At this time, in the image memory 10,
When the input medium data 1) is stored in the state shown in FIG. 2(B), this difference is 1 or more in the portion corresponding to the side L2. At this time, since no flag is set in the flag section 8 and the count value of the counter 7 is zero, the continuity detection section 6
This is reset in the processing routine, and the count value Cn of the counter 7 during horizontal direction processing is set to 0. Then, the line unit data of the next coordinate position is read out by the control section 20.

By the way, when the line unit data is read out,
The above h(i)-h(i+1)=0 holds true. The continuity detection unit 6 holds the position i of the line unit data and its h (i) at this time, and increments the counter 7 by 1. Further, the next line unit data is read out by the control unit 20, and similar processing is performed. In this case, if the inclination angle θ of Ll shown in FIG. 1(D) is set to 66, then
The above relationship holds true when the distance from n to n2' in FIG. Hold it at 6. When the above formula no longer holds true at n3, the continuity detection unit 6 recognizes that the flag is not set in the flag unit 8 and that the count value C of the counter 7 is CAT, and the continuity detection unit 6 recognizes that the flag C of the counter 7 is CAT. The point P indicated by and h (i) is set in the starting point holding section 9-1 as the starting point indicating the linearity of the side L1, and a flag is set in the flag section 8.

Then, counter 7 is reset.

However, when the line unit data of n4 is read out, the above equation is established again. Then, similar processing is performed, but n,
This is not true. Since the flag is set at this time, the continuity detection unit 6 now detects i and h in n3.
The point P shown in (i) is set in the end point holding section 9-2, and the counter 7 is reset.

When line unit data of nh is read out, the above formula holds true and similar processing is performed, but when line unit data of n7 is read out one after another, the above formula does not hold and the flag is set. Count value CH of thing counter 7
is 1 and CM≧T does not hold, so the counter 7 is reset.

After this missing portion F passes, a new Pg obtained from the above relationship will be obtained again, so the continuity detecting section 6 rewrites the end point holding section 9-2 with this new P2.

Such processing is performed sequentially, and when processing for line unit data up to the size N of the image memory IO is completed, Pl written in the starting point holding section 9-1 and ending point holding section 9-
P entered in 2! The line that connects the , or the area around it, is the line that indicates the state of .

Pl and P2 obtained in this way are sent to the orthogonality detection section 4.

Next, the continuity detection unit 6 performs a process of determining the straight line of the side Lt. In this case, based on the line unit data created by the vertical direction processing section 2, processing as shown in the flowchart of FIG. 4(B) is performed. At this time, the counter 7 counts the continuous value C■ for the data in the vertical direction, and the same processing as in FIG. 4(A) above is performed except that it is performed in the area up to the size M of the image memory 10. Therefore, detailed explanation thereof will be omitted. The starting point Pl' and ending point P2' thus obtained are also sent to the orthogonality detecting section 4, and the orthogonality of L2 is checked.

The orthogonality detection unit 4 performs the following calculation, for example, to detect orthogonality between two straight lines. That is, as shown in FIG. 4(D),
A straight line passing through point P+ (x+, y+), Pg (Xi, y2), and point PI' (X3, y3), P2'
(X4%) If the straight line LN passing through 'a) is orthogonal, the product of their slopes is -1, as is well known.

That is, X, -X.

x3-x.

Then, when L, and, are orthogonal) (, x K
z=1.

Therefore, when the orthogonality detection section 4 receives the data of Pl and P2 for the area outputted from the continuity detection section 6 and Pl' and PZ' for the side L2, the orthogonality detection section 4 performs the above calculation using these data. The product of the slopes is calculated and if the product is approximately -1, it is determined that the sides L1 and L2 are orthogonal. However, when it is determined that they are not in an orthogonal state, the continuity recognition unit 3 outputs the start point and end point data of around, and determines the orthogonality between around and L3, as will be described later. When orthogonality is not determined as a result of this, the continuity recognition unit 3:
Output the start point and end point data of side L4, and now side L4 and L
2 or L4, find the orthogonality.

As shown in Figure 4 (C), if orthogonality cannot be obtained even with all of these methods, the result is NG, and the input medium is not rectangular or square, or the input medium is not in a normal condition because there are many damages on the sides.・It will be determined that

Note that in order to find the start and end points of side L3, the same control can be performed using the number of dots in the last section of the line unit data created by the vertical direction processing section 2, as shown in Fig. 2 (B
) In the example of (2), the second dot-free number may be used. Similarly, to find the start and end points of the side L4, the same control can be performed using the number of dots in the last section of the line unit data created by the horizontal direction processing section 1, as shown in FIG.
In B), the second dot-free number may be used at the H2 location, and the third dot-free number may be used at the H1 location. In this way, in any case, the number of dots in the last classification can be used, so the control j1)1 is easy.

In this way, for example, as shown in FIG. 5(A), even if there is a missing portion C or a bent portion on the side L1, a straight line obtained from the point AB can be obtained.

When the orthogonal sides, for example Ll and L2, are determined in this way, the skew angle calculation unit 5 can calculate the skew angle θ′.

Based on this, data string 1)-1.1 in Figure 5 (A)
)-2- can be obtained from the input medium data 1), so these can be compared with the data strings 16-1, 16-2- of the basic pattern data 16, as shown in FIG. 1(B). This allows accurate recognition.

Note that the above explanation was based on an example where θ is 6″ at maximum, but
Of course, the present invention is not limited to this, and if θ changes, T will also change accordingly.

〔Effect of the invention〕

According to the present invention, even if there is a missing part or a bent part in a part of the input medium, it is possible to obtain the correct oblique angle without being affected by this, so that accurate image recognition can be performed using one basic data. Can be done.

[Brief explanation of the drawing]

Fig. 1 is an explanatory diagram of the principle of the present invention, Fig. 2 is an embodiment of the present invention, Fig. 3 is an example of a reading section, Fig. 4 is an explanatory diagram of the operation of the present invention, and Fig. 5 is an illustration of input medium data. Explanatory diagram: FIG. 6 is an explanatory diagram of a conventional example. 1--Horizontal direction processing unit 2-Vertical direction processing unit 3--Continuity recognition unit 4-Orthogonality detection unit 5-Oblique angle calculation unit

Claims (1)

[Claims] (1) The image information of the medium to be recognized is registered in advance as a basic pattern, the image information of the input medium read by the sensor is temporarily stored in memory, and the image of the input medium is imaged by comparing it with the basic pattern. In an image recognition device that recognizes 5), wherein the image information on the input medium is scanned in the horizontal direction by the horizontal direction processing section (1) to obtain image presence/absence information, and the image information on the input medium is scanned in the vertical direction by the vertical direction processing section (2). Then, the presence/absence information of the image is also obtained, and from this information, the continuity recognition section (3) obtains side information of the input medium, and the orthogonality detection section (4) determines whether or not the sides are orthogonal.
A skew correction method characterized in that a skew angle of an input medium is calculated when the input media are perpendicular to each other.