JPH07334605A - Tilt detecting method - Google Patents

Abstract

PURPOSE:To fast detect a skew (tilt) angle by means of a rectangular pattern. CONSTITUTION:The scanning of a bar code pattern is carried on when the detection of tilt of a bar code is started (step 1), and it is decided whether a subject rectangular area is being scanned or not (step 2). If not scanned, the step 2 returns to the step 1. If scanned, the scanning distance (run length) is measured in the rectangular area (step 3). Then, it is decided whether the scanning of the rectangular are has been completed or not (step 4). If not completed, the step 4 returns to the step 1. If completed, a histogram is acquired (step 5). Then an estimated skew angle theta is calculated from the maximum value of the histogram (step 6).

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a tilt detecting method,
More specifically, skew (slope) even when there is a large amount of information
The present invention relates to a tilt detection method for detecting a high speed. For example, it is also applied to the detection of the inclination of a barcode and the detection of the inclination of a document in a copying machine.

[0002]

2. Description of the Related Art In general, skew detection is only performed practically simply, but in a document image, for example, in Japanese Patent Laid-Open No. 2-116987, a straight line connecting the connected components of the image is assumed and The azimuth is detected as the skew angle. That is, the above publication describes a method for detecting the inclination of a character line in a character recognition device. It extracts a character line from character image data, cuts out each character from the extracted character line, and extracts a reference line. By a means, for a set of parameters representing an arbitrary straight line at a specific point of each character of a character line, a histogram is taken on the parameter space, and a straight line defined by the set of parameters giving the highest frequency on the histogram is used as a reference line, Since the inclination of the reference line is obtained, it is not necessary to use an advanced image input means for collecting image data including a character line, and the inclination of the character line can be calculated from a relatively small amount of image data. It is a thing.

In Japanese Patent Laid-Open No. 5-174182, a circumscribing rectangle processing is performed on a document image, a circumscribing rectangle of one character in a character string is selected, and a reference point is extracted from the selected circumscribing section rectangle. Then, a vector is obtained by connecting another reference point within a certain distance from the reference point of interest, and the average direction of the obtained vector is set as the reference direction of the document, and the reference direction and the reference direction at the time of input are set. The angle difference is used as the document inclination angle.

Further, Japanese Patent Publication No. 60-57108 relates to a mark sheet reading system in which a facsimile transmitter is used as a mark sheet reading device, and a reference mark, a scanning position designation mark, a tilt angle detection mark and a mark column are provided on the mark sheet. Of the facsimile signals obtained by scanning the mark sheet, the tilt angle of the mark sheet is first detected, and then a plurality of scanning lines for scanning the mark column are selected according to the tilt angle to perform mark processing. Is to do. That is, in order to detect the inclination of the form with a special mark,
There are multiple specific marks on a specific part of the form, the reading direction of the form is predetermined, and the calculation method of the inclination is not strictly specified, but roughly the appearance order of the marks is used. To do.

Further, there is one in which inclination detection is applied to a bar code. There are various types of known barcodes, and their applications are diverse. Typical examples are black-and-white patterns arranged at equal intervals, black-and-white pair patterns arranged, and black-and-white patterns arranged at different pitches. Different from ordinary barcodes as encoding the character information itself 2
There is a dimensional data code. The information that can be encoded by this two-dimensional data code can handle not only alphanumeric characters, kana characters, and kanji but also all character types such as symbols. And can be read from any direction of 360 degrees. For example, since it can be read from all directions such as the price tag of a product in the distribution industry, it can be handled without problems even if the label is cluttered.

Regarding the types and characteristics of the two-dimensional bar code, the following points are mentioned. Symbol Technology: Developed to encode large amounts of information, but designed for dedicated scanners. The skew is dealt with by tracking the arrangement of the codes in the memory after reading. DATA Code: Specify the recording range using the edges of black solid and black and white mottle. The maximum amount of information that can be recorded is 500 bytes.

Softstrip: Designed for dedicated scanners. The recording density is high because recording is done in units of black and white pairs. Vericode: Used to encrypt the code. Philip dot code: A bit code is recorded in an 8 × 8 area. There are special applications such as print boards. Carla code: A bit code is recorded in a 2 × 2 square.

[0008]

In recent years, with the speeding up of information processing such as document management, a specific figure, for example, a rectangle is used to detect the inclination of a sheet-like object, but the figure is a form. It does not have to exist in a specific part of the
It is required to realize a method for obtaining a tilt angle at high speed by utilizing the contour of a form without explicitly printing a rectangle. Further, as described above, various methods are known for tilt detection, and their applications are diverse. However, with the advent of digital copiers, their usage has changed. Table 1 below shows the differences between the conventional method and the case of using a copying machine or a camera.

[0009]

[Table 1]

A major difference between the conventional method and the application in a copying machine, which can be seen from Table 1, is that the bar code applied to a document has a large amount of information and a short allowable time for recognition. In addition, since the scanner format is different and the amount of information is large, the skew (tilt) of the code pattern must be corrected in an instant.

The present invention has been made in view of the above circumstances, and an object thereof is to provide a tilt detecting method for realizing pattern cutting and skew (tilt) angle detection at high speed.

[0012]

In order to solve the above-mentioned problems, the present invention (1) finds a frequency distribution of the lengths of scanning lines passing through the inside of the rectangle for a rectangular pattern, To calculate the skew angle of
(2) Obtain the frequency distribution of the lengths of scanning lines that pass through the inside of a rectangular pattern that surrounds the bar code that extends two-dimensionally or that accompanies the bar code, and calculate the skew of the rectangular line. Calculating the angle, (3) the rectangular pattern is a frame-shaped line or a boundary line, and (4) the patterns forming the edges of the rectangular are regular with a constant period. It is a pattern, or (5) determining means for determining whether or not the target rectangular area is being scanned, and the scanning distance inside the rectangle when the rectangular area is being scanned by the determining means. Counting means for counting, histogram creating means for obtaining a histogram of frequency distribution of scanning distances counted by the counting means, and histogram creating means Skew angle detecting means for obtaining a skew angle expected from the maximum value of the strogram, and (6) in (5), the size of the rectangular region is set to a vertical Hv and a horizontal Hh, and a rectangle is set to an arbitrary rectangle. When scanning is performed at an angle θ, and when scanning is performed at the angle θ, when the length of a perpendicular line from the output side corner to the scanning line passing through the input side corner of the rectangle is L, the scanning distance M = Hh / Cosθ, and the area that contributes to the frequency peak value is S = Hvcosθ
−Hhsinθ, and from the relation of L = Hhsinθ, the area contributed by the upper and lower corners of the rectangle is obtained from the relational expression that the area other than the peak of the frequency distribution is 2Hhsinθ, and any one of the above relations The skew angle θ is obtained from the equation, and (7) in (6), the average value of the skew angles obtained from the relational expression is used as the most plausible skew angle.

[0013]

The inclination detecting method according to the present invention scans a rectangular area which is a rectangular pattern, obtains a histogram from the frequency distribution of the scanning distance (run length) inside the rectangle, and predicts from this histogram distribution. Calculate the skew angle. This makes it possible to detect the skew angle at high speed.

[0014]

Embodiments will be described below with reference to the drawings. FIG. 1 is a diagram for explaining an embodiment of a tilt detecting method according to the present invention. The size of the rectangular area defined by a specific mark is Hv in the vertical direction and Hh in the horizontal direction. When this rectangle is scanned at an arbitrary angle θ, the distribution of the scanning line length (run length) M inside the rectangle becomes a histogram as shown in FIG. That is, it has a sharp peak and a gentle distribution curve, except when θ is in the diagonal direction.

Therefore, the run length value having the peak value (or the maximum value) of the distribution corresponds to the scanning section (scanning distance) M having a constant width in FIG. 1 and has the value of Hh / cos θ. Therefore, the skew angle θ is obtained.

Further, since the area S (the gray portion in FIG. 2) that contributes to the peak value of the run length frequency distribution should be Hvcosθ'-Hh sinθ '(corresponding to the number of scans), the skew angle is calculated from this. θ'is obtained. The area S in FIG.
Corresponds to the area S in. Further, the distance L is the length of a perpendicular line from the output side corner with respect to the scanning line passing through the input side corner of the rectangle. Similarly, the area N of the gentle distribution in FIG. 2 corresponds to the upper and lower triangular portions (N ₁ + N ₂ ) in FIG. 1, and the total number of scanning lines is 2L = 2Hh sin θ, so this is 2Hh sin θ ″ Then, the skew angle θ ″ can be obtained.

In FIG. 1, as the angle θ becomes larger, L becomes larger and M becomes larger. Along with this, in FIG. 2, the maximum value of the histogram becomes smaller, M shifts to the right, and the gentle area of N expands to the right. On the contrary, in FIG. 1, when the angle θ becomes smaller, L becomes smaller and M becomes smaller. Along with this, the maximum value of the histogram in FIG.
Shifts to the left, and the gentle area of N shrinks to the left. That is, if the angle θ is zero, L is also zero and M is the minimum. Along with this, in FIG. 2, the maximum value of the histogram becomes the largest, and the smooth area of N disappears.

Of course, since the skew angles θ, θ ′, θ ″ should be the same value, the average value of these is the most probable skew angle (θ = (θ + θ ′ + θ ′) /
3). However, it should be noted here that the positive and negative signs of the angle θ are indeterminate. That is,
I don't know which direction the calculated angle is.
This measurement accuracy depends on the accuracy of the frequency distribution and the accuracy of the run length shown in FIG. Therefore, it is desirable that the size of the target rectangle be large, and that the number be large.

FIG. 3 is a diagram showing a case where the inclination detecting method according to the present invention is applied to the inclination (skew) detection of a bar code. ABCD in the figure indicates the four outer corners of the rectangle,
EFGH indicates the inner four corners of the rectangle, and the rectangle E
A barcode pattern is shown inside the FGH. The target code here is a pattern spread in a two-dimensional plane and is usually called a two-dimensional bar code. This bar code is composed of a pattern in units of rectangles.

The central portion of the rectangle EFGH in FIG.
It is a so-called two-dimensionally spread bar code, and the upper and lower diagonal lines are figures showing the existing range, and the left and right images represent the start or stop of the mark. In a normal reading optical system, a light spot as shown by an arrow in FIG. 3 scans, and in the process, a start and stop symbol is detected and a signal train existing therebetween is detected. Generally, a plurality of scans are continuously performed in parallel, so that a two-dimensional code can be restored by rearranging a signal sequence on a memory. Further, even when the azimuth of the mark is significantly inclined, the conventional optical system attempts to scan in various directions of 8 directions or more, so that the condition that the scanning line passes from end to end is satisfied in most cases. .

However, when reading the mark with a scanner of a copying machine or a facsimile, the scanning is performed only once. Therefore, after the image is input, it is necessary to detect and correct the inclination (skew) of the mark on the memory. The present invention relates to the detection of skew detected by such a single scan, and particularly aims at detection in real time.

Although FIG. 3 shows a rectangular pattern surrounding the barcode, this rectangular pattern may be provided in the vicinity without surrounding the barcode. Further, although FIG. 3 shows the case where the rectangular pattern is a regular pattern with a constant period, it is clear from FIG. 1 that it may be a simple frame-shaped line. When applied when detecting the inclination of a sheet original to be copied by a copying machine or the like, the sheet original has a rectangular shape, and the skew angle can be detected by using the boundary line (edge). Further, the frame-shaped line may be a frame-shaped line having a black width. Also,
In FIG. 3, two types of measurements are possible, one for the rectangle ABCD and one for the rectangle EFGH.

FIG. 4 is a flow chart for explaining the tilt detecting method according to the present invention. Hereinafter, each step will be described in order. First, when the barcode inclination detection is started, scanning of a rectangular pattern surrounding the barcode is advanced (step 1). Next, it is determined whether or not the target rectangular area is being scanned (step 2). That is, it is determined whether the barcode is inside the rectangle. If No, return to step 1 above, and if Yes,
Next, the scanning distance (run length) inside the rectangle is counted (step 3). Next, it is judged whether or not the scanning of the target rectangular area is completed (step 4).
Returning to the step 1, if Yes, then a histogram is obtained (step 5). Further, the skew angle θ expected from the maximum value of the histogram is calculated (step 6).

As described above, in the present invention, for example, regarding a two-dimensionally spread bar code, when a rectangular pattern surrounding the bar code or accompanying the bar code exists, the bar code passes through the inside of the rectangle. The skew angle of the rectangle is calculated from the frequency distribution of the lengths of the scanning lines.
Further, the pattern forming the edge of the rectangle is a regular pattern with a constant period, and further, the skew angle is obtained using three or more independent measurement values, and the statistical calculation result of the obtained value is This is the most likely skew angle.

As described above, the inclination detecting method according to the present invention uses a specific figure (rectangle), but this figure does not necessarily have to exist in a specific portion of the form, and the direction of reading the form is arbitrary. . That is, the form may be rotated by 90 degrees and read. Further, as described above, there are three methods for calculating the inclination angle of the form, and the angles obtained by these are independent of each other. Therefore, the detection error can be suppressed to be small by averaging each of the three angles. Furthermore, the inclination angle can be obtained using the contour of the form without explicitly printing the rectangle by utilizing the fact that the form itself is rectangular. Or a rectangle for other purposes that exists in the form (eg a table)
It is also possible to use. However, since the tilt angle is not known until the reading of the rectangle is completed, an image memory of a sufficient size is necessary to store the image in the meantime.

[0026]

As is apparent from the above description, according to the present invention, a rectangular area that is a pattern is scanned, a histogram is obtained from the frequency distribution of the scanning distance (run length) inside the rectangle, and this histogram is obtained. The skew angle expected from the distribution of is calculated. This makes it possible to detect the skew angle at high speed.

[Brief description of drawings]

FIG. 1 is a diagram for explaining an embodiment of a tilt detection method according to the present invention.

FIG. 2 is a diagram showing a length distribution of scanning lines inside a rectangular region in the present invention.

FIG. 3 is a diagram showing a case where the inclination detection method according to the present invention is applied to inclination detection of a barcode.

FIG. 4 is a flow chart for explaining a tilt detection method according to the present invention.

Claims (7)

[Claims] 1. A tilt detecting method comprising: calculating a frequency distribution of lengths of scanning lines passing through the inside of a rectangular pattern and calculating a skew angle of the rectangular pattern. 2. A frequency distribution of the lengths of scanning lines which pass through the inside of a rectangular pattern which surrounds a two-dimensionally expanded bar code or which accompanies the rectangular code is calculated to obtain the rectangular pattern. An inclination detection method, characterized in that a skew angle is calculated. 3. The tilt detecting method according to claim 1, wherein the rectangular pattern is a frame-shaped line or a boundary line. 4. The pattern forming the edges of the rectangle is
The inclination detecting method according to claim 1 or 2, wherein the inclination detecting pattern is a regular pattern having a constant period. 5. A judging means for judging whether or not a rectangular area of interest is being scanned, and a counting means for counting a scanning distance inside the rectangle when the rectangular area is being scanned by the judging means. A histogram creating means for obtaining a histogram of the frequency distribution of the scanning distances counted by the counting means, and a skew angle detecting means for obtaining a skew angle expected from the maximum value of the histogram obtained by the histogram creating means. An inclination detection method characterized by: 6. The size of the rectangular area is Hv in the vertical direction and H in the horizontal direction.
h, the rectangle is scanned at an arbitrary angle θ, and when scanning is performed at the angle θ, the length of a perpendicular line from the output side corner to the scanning line passing through the input side corner of the rectangle is set to L. At this time, the scanning distance M = Hh / cosθ is obtained, and the area contributing to the peak value of the frequency is obtained by S = Hvcosθ−Hhsinθ, L = H
From the relationship of hsinθ, the area contributed by the upper and lower corners of the rectangle is obtained from the relational expression that the area of the portion other than the peak of the frequency distribution is 2Hhsinθ, and the skew angle θ is obtained from any of the relational expressions. The tilt detection method according to claim 5, wherein 7. The tilt detecting method according to claim 6, wherein an average value of the skew angles obtained from the relational expression is used as the most probable skew angle.