JPH0696273A - Recognition field retrieving method in business form reader - Google Patents

Abstract

(57) [Abstract] [Purpose] It is an object of the present invention to provide a recognition field search method for a form reading device, which is capable of coping with flexure during conveyance of a form. [Structure] Image data obtained by reading a form S by an image sensor 24 is temporarily stored in an image memory 40, and a character at a position of a recognition field I is detected from the image data stored in the image memory 40 to obtain character data. In the form reading device that reads the contents described in the form by converting the form into the form, the reading direction of the image data from the image memory 40 is set to be perpendicular to the vertical direction of the form S. As the vertical direction, a line connecting two positions vertically separated by a predetermined dot from the left vertical edge or the right vertical edge of the form S corresponding to the vertical recognition field position is obtained.
A line perpendicular to the position connecting the two positions is obtained, and the image data is read based on the line.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a form reading device, and more particularly to a recognition field searching method in the form reading device.

[0002]

2. Description of the Related Art FIG. 3 is a side view showing an outline of a mechanical system of a form reading device, and FIG. 4 is a block diagram of an electric system of the form reading device.

The form S placed on the hopper 10 is sent to the conveying means by the pick roller 11. The scanning means 20 is provided in the middle of the transportation path, and the light from the light source 21 is condensed in a slit shape and is irradiated onto the transportation path at a right angle.

Based on such a structure, when the form S is sent into the conveyance path and the light from the light source 21 is applied to the form S, the reflected light is reflected by the folding mirror 22 and the lens 2.
3 is input to the image sensor 24. here,
The characters written on the form S are photoelectrically converted to obtain analog image data.

The analog image data obtained by the image sensor 24 is converted into digital data by an A / D converter 32 via an amplifier 31, and then binarized by a binarizing means 33 at a predetermined slice level. It is stored in the image memory 40.

In this way, the character extraction means 41 extracts the characteristics of each character from the image data stored in the image memory 40, and the recognition processing means 42 receives the reference data obtained from the recognition dictionary 43 and performs the recognition processing. I am supposed to do it.

FIG. 5 shows the procedure from the acquisition of image data to the character recognition in the above configuration.
Is an explanatory diagram thereof. The host CPU 50 has a position in the character recognition range (hereinafter referred to as a recognition field) on the form as position information in advance, and the position information is transmitted to the recognition processing means 42 prior to reading the form S (FIG. 5). , F111). Next, the form is read, and at this time, when the form is skewed or meandering and unreadable, rereading is performed (FIG. 5, F112). Then, the image data obtained by this reading is stored in the image memory 40 (FIG. 5, F113).

When reading image data from the image memory 40, first, a portion other than the form S in the image memory 40 (for example, represented by "1") and a form portion (for example, at least 3 mm from the edge portion is " 0 ") is read to determine the leading edge M ₀ (see FIG. 6), and the scanning start point A and scanning end point B at the leading edge M ₀ are determined (FIG. 5, F114). When the leading edge M _{0 of the} form is obtained in this way, reading is started in parallel with it.

As described above, the host CPU 50 previously holds the recognition field Ia ... Position on the form S as the position information, and based on the position information of the recognition field Ia .. (the vertical middle position of the normal recognition field). Characters in the recognition field Ia ... Are extracted (FIG. 5, F11).
5). That is, as shown in FIG. 6, the vertical direction m ₁ of the form S is
If there is a 3-character recognition field Ic at the position of,
Searching from the vertical direction m ₁ as a starting point to a position where the first character is written in parallel with the leading edge M ₀ , each character is within a frame of a size determined by the definition format transmitted from the host CPU 50. Search image data. If the position where the character is written cannot be retrieved at this time, image data in a frame larger than the above-defined frame (for example, 1.5 times in length) is searched (FIG. 5). , F1
15a, F115b).

After that, the center position of each character is determined, the feature is extracted, and the character is recognized. Further, this work is performed in the order of the characters belonging to one recognition field from left to right and from the top to the bottom of the recognition field belonging to one form (F117) (F118).

[0011]

According to the above-mentioned conventional method, the image data is read out from the image memory 40 in parallel with the leading edge M ₀ of the form. On the other hand, as the sheet used as the form S, not only a thick sheet with little bending but also a considerably thin sheet is used. Therefore, FIG.
As shown in FIG. 6, there is no problem when the paper is normally conveyed by the conveying means. However, as shown in FIG. Even if it is fed in the direction where there is no bending in the direction of, the bending may occur from side to side for some reason.

However, as described above, the host CPU 5
Since 0 has the position of the recognition field Ia with characters as position information, when the form begins to bend halfway, the recognition field supposed to be at the position l ₁ from the left end of the distance m ₁ from the leading edge M _0. There is a case where Ia ... Is absent as shown in FIG. 6 (b). This phenomenon becomes more prominent as the scanning distance from the scanning start point (distance from the leading edge) m ₁ becomes longer, as shown by the scanning line n in FIG. 6B. In this case, the characters written in the recognition field IC of the form S are recognized as unreadable characters or erroneous characters.

The present invention has been proposed in view of the above conventional circumstances, and it is an object of the present invention to provide a recognition field search method capable of coping with a flexure of a form during conveyance.

[0014]

The present invention employs the following means in order to achieve the above object. That is, for example, as shown in FIG. 1, the image data obtained by reading the form S by the image sensor 24 is temporarily stored in the image memory 40.
In the form reading device for reading the contents described in the form by detecting the character at the position of the recognition field I from the image data stored in the image memory 40 and converting it into the character data, The reading direction of the image data from 40 is perpendicular to the vertical direction of the form S.

As the vertical direction, a line connecting two positions vertically separated by a predetermined dot from the left vertical edge or the right vertical edge of the form S corresponding to the vertical recognition field position is obtained, and the two positions are connected. A line perpendicular to the position is obtained with respect to the line, and the image data is read based on the line.

[0016]

[Function] From the contour line of the form S on the image memory 40, the form S
The left (or right) edge of the can be determined. When the form S is distorted, there is some error between the recognition field position information held by the host CPU 50 and the actual recognition field position detected from the line perpendicular to the left end (right end) edge. However, the error does not increase as the distance from the left edge (right edge) increases as in the conventional case, and thus the error does not become unreadable.

[0017]

1 is a conceptual diagram showing an embodiment of the present invention, and FIG. 2 is a flow chart showing the procedure of the present invention.

First of all, prior to the work, the form processing unit 42 previously input to the host CPU 50 is loaded into the recognition processing means 42 (FIG. 2, F111). Then
The conveying means and the reading means 20 are driven to read the image data, and the image data is stored in the image memory 40 (FIG. 2, F112 → F113). The above is the same as the conventional example shown in FIG.

Then, the recognition field position designated by the form format obtained from the host CPU 50 is detected, that is, the left edge position of the form S corresponding to the position of the recognition field I ₁ in FIG. 1A is detected ( F
1). Then, a position is obtained at a position that is X bits higher (X is several bits, for example) from the above position, and a position that is X bits lower than the above position is obtained (F2 → F3).

Then, with respect to the line connecting the above-mentioned positions, scanning is performed along a line perpendicular to the position (F4). As a result, similarly to the case shown in FIG. 5, it is possible to search for a written character in the recognition field I ₁ and position the detected character (F115 → F116). Here, when a character cannot be detected, an allowable range (for example, a size range of 1.5 times the size of the frame defined by the format definition) is searched to detect the character. Here, when the character cannot be detected, the unreadable process is performed, and further the character thus detected is sequentially recognized (F117, F11).
8).

When the scanning for character retrieval is performed as described above, the scanning at a right angle to the left edge can be performed, and even if the form is slightly bent during conveyance, the host CPU 50 The recognized recognition field position and the actual recognition field position do not largely shift, and the probability of rejection is reduced.

In the above embodiment, the left edge of the form is at a right angle.
Only the case of scanning was explained, but Fig. 1
As shown in (b), the recognition field is based on the right edge.
Do I _3,I_FourOf course, you can search for characters in
Is.

[0023]

As described above, according to the present invention, the image data is read out from the image memory at right angles to the vertical direction of the form. Therefore, the registration position of the recognition field due to the positional deviation caused by the bending of the form during conveyance of the form can be obtained. There is an effect that a large error does not occur between the actual position and the need for rereading work.

[Brief description of drawings]

FIG. 1 is a conceptual diagram of an embodiment of the present invention.

FIG. 2 is a flow chart showing a procedure of the present invention.

FIG. 3 is a conceptual diagram of a mechanical system configuration of a form reading device.

FIG. 4 is a block diagram showing an electrical system configuration of a form reading device.

FIG. 5 is a flowchart of a conventional example.

FIG. 6 is a conceptual diagram illustrating a conventional defect.

[Explanation of symbols]

 A form I recognition field 24 image sensor 40 image memory

Claims (2)

[Claims] 1. The image data obtained by reading the form (S) by the image sensor (24) is once stored in the image memory (40), and the recognition field (I) is stored from the image data stored in the image memory (40). ), The character at the position is detected and converted into character data to read the contents described in the form.In the form reading device, the reading direction of the image data from the image memory (40) is set to the vertical direction of the form (S). A method for retrieving a recognition field in a form reading device, which is perpendicular to the direction. 2. A line in a direction perpendicular to a line connecting two positions vertically separated by a predetermined dot from the left vertical edge or the right vertical edge of the form (S) corresponding to the vertical recognition field position. The recognition field search method in a form reading device according to claim 1, wherein the image data is read based on the line.