JPH036553B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to an optical character reading device that is capable of reading forms in which character lines are close to each other.

In conventional optical character reading devices, there is a certain limit on the line pitch to prevent character patterns from entering the line to be read from lines above and below the line to be read due to skew of the form, and it is not possible to read character lines that are close to each other. A problem when reading forms with close character lines is that the character patterns of the lines above and below the line to be read invade, making it impossible to separate the lines. That is, it is impossible to separate horizontal and vertical oblique shadows.

SUMMARY OF THE INVENTION An object of the present invention is to solve these conventional drawbacks and to provide an optical character reading device that can read forms in which character lines are close to each other.

The present invention will be explained in detail below using the drawings. FIG. 1 is a block diagram showing one embodiment of the present invention. Reference numeral 1 indicates a control section that controls the entire device. What is indicated by the reference numeral 2 is a bus. Reference numeral 3 denotes a photoelectric conversion unit that converts a character pattern recorded on a form into an electrical signal, and performs quantization and preprocessing. Skew detection unit 4
detects the skew of the form from the photoelectric conversion signal output from the photoelectric conversion unit 3. The image memory 7 is
By storing the photoelectric conversion signal, a form image for about one line is stored. The image memory control unit 6
Address control of this image memory 7 is mainly performed. The mask unit 5 generates a projection mask whose position is relatively variable according to the Y-direction address of the image memory 7.

What is indicated by the reference numeral 8 is a gate circuit. code 9
10 is a vertical projection unit that takes a vertical projection of the character pattern on the X axis using the photoelectric conversion signal stored in the image memory 7 and stores it.
The part shown by is the horizontal projection part. Character recognition section 11
recognizes character patterns and makes reject decisions using known recognition methods such as feature extraction, structural analysis, and pattern matching. 12 is a buffer for storing recognition results for one line or one form, and 13 is an interface section.

Hereinafter, one embodiment of the present invention will be described in detail using FIGS. 1 to 3. The photoelectric conversion unit 3 outputs a photoelectric conversion signal obtained by scanning the form. This photoelectric conversion signal is given to the skew detection section 4, the image memory 7, and the gate circuit 8. Next, the skew detection operation of the skew detection section 4 will be explained using FIG. What is indicated by the reference numeral 20 is a form and is conveyed in the direction of arrow C. A certain value n (0<n
≦0.5) to calculate l・n and l−l・n,
The value is stored in a register within the skew detection section 4. The skew detection section 4 inputs the photoelectric conversion signal output from the photoelectric conversion section 3 and detects the width of the white level of this photoelectric conversion signal. When the width of the white level reaches the lengths l.n and l-l.n, the skew detection section 4 generates a status indicating that a blank portion at the leading edge of the form 20 has been detected. The control unit 1 counts the time difference between the two statuses or the drive signal of the form conveyance system, and
The distance L that the form 20 is transported while both statuses occur is calculated. Point A and point B on the form 20 are points located at distances l and n from the left and right ends, respectively. When the form 20 is conveyed and reaches the photoelectric conversion unit 3, if there is no skew, both points A and B are detected at the same time, and both of the above-mentioned statuses occur simultaneously. When skew occurs, the inclination D of the form is D=L÷(l
-2ln)≒1÷d. Furthermore, the direction of skew can be detected by determining which side, left or right, the white level appears first in one scan of the photoelectric conversion signal. These statuses and calculated data are all stored in a register within the skew detection section 4.

Now, suppose that the form 20 is conveyed while skewing to the left as shown in FIG. The photoelectric conversion unit 3 takes the midpoint M of the distance between points A and B as the virtual starting end of the form 20, and forms an area surrounded by a dotted line centered on a point a distance F away from the midpoint M, with a vertical field of view J. E
scan. The image memory 7 is an image memory control unit 6
Under the control of this scan, a photoelectric conversion signal representing the form image obtained from this scan is stored. Image memory 7
The memory content of is set to be larger than one line of text to cover skew in the form. In the example shown in FIG. 3, the form image is stored in the image memory 7 as a photoelectric conversion signal over a range of distance G with the character pattern intruding from the lower character line. If this continues, even if the vertical diagonal shadows are removed from the contents of the image memory 7, the vertical diagonal shadows over a distance G will be connected due to the effect of the character pattern entering from below, making it impossible to detect and cut out one character.

Further, at the same time, the photoelectric conversion signal is given to the gate circuit 8. The gate circuit 8 uses the mask signal output from the mask section 5 as a gate signal, and when this gate signal is applied, the photoelectric conversion signal passes through and is output. The mask part 5 is
A mask signal is generated from the skew direction and inclination D obtained from the skew detection section 4 in the following manner. The mask signal is positioned with a constant width y centered on the virtual center line N of the row to be read, and every time the photoelectric conversion unit 3 scans d bits in the x direction, it is shifted one bit either up or down in the skew direction. It is something that moves. Since the form 20 in FIG. 3 is conveyed with an inclination to the left, the mask signal is located at a width of y downward from the temporary sweeping center line N at the left end of the form 20, and at the virtual center line at the right end. It is located above N. If the vertical projection unit 9 performs vertical projection using the photoelectric conversion signal obtained via the gate circuit 8 in this manner, the projection can be performed without being affected by character intrusion from the character line below. Next, based on this vertical projection, the image memory control section 6 cuts out a photoelectrically converted character pattern from the image memory 7 one character at a time. This extracted character pattern is given to the character recognition section 11 via the gate circuit 8 which is in an open state. The character recognition unit 11 recognizes the extracted character pattern and outputs the answer to the buffer 12. The answer stored in this buffer is output to the outside via the interface section 13. Further, the cut out character pattern is subjected to horizontal projection in a horizontal projection section 10, where it is checked whether the characters protrude or intrude.

In the above description, an embodiment has been described in which vertical projection is performed in parallel with storing the photoelectric conversion signal in the image memory 7, but the following method may also be used. After storing the photoelectric conversion signal in the image memory 7, the character pattern read out from the image memory 7 may be vertically projected using the character pattern masked by the gate circuit 8 using a mask signal. Furthermore, you may do as follows. After storing the photoelectric conversion signal in the single image memory 7, the skew direction and inclination degree D obtained by the skew detection section 4 are
Based on this, the image memory control section 6 performs address control to eliminate the intrusion of the character pattern from the image memory 7 and read out the remaining portion. At this time, the character pattern read out from the image memory 7 is passed through a gate circuit 8 without a mask, and a vertical projection section 9 takes a vertical projection.

As is clear from the above detailed description, the present invention has the following effects. It is now possible to read forms with text lines that are close together. When vertical projection is taken, the influence of form skew can be removed, so character patterns can be detected and extracted using vertical projection.

[Brief explanation of drawings]

FIG. 1 is a block diagram showing one embodiment of the present invention;
FIG. 2 is a diagram for explaining a form, and FIG. 3 is a diagram showing a form being conveyed in a skewed state. 3... Photoelectric conversion section, 4... Skew detection section, 5
...Mask section, 6... Image memory control section, 7...
Image memory, 8...gate circuit, 9...vertical projection section, 11...horizontal projection section.

Claims (1)

[Scope of Claims] 1. A photoelectric conversion unit that scans a form, photoelectrically converts the form image, and outputs a photoelectric conversion signal, an image memory that stores photoelectric conversion signals for one or more lines, and a direction and degree of inclination of the form. a skew detection unit that detects
a mask unit that generates a mask signal using the inclination direction and degree of inclination of the form; a gate circuit that uses the mask signal to perform mask processing on the photoelectric conversion signal read out from the image memory; An optical character reading device comprising: a vertical projection unit that takes a vertical projection on a line to be read; and a recognition unit that uses the vertical projection to recognize a character pattern read from the image memory. 2. A gate circuit that stores the photoelectric conversion signal in the image memory and performs mask processing on the photoelectric conversion signal using the mask signal in parallel. 1
Optical character reading device as described in Section 1. 3. A photoelectric conversion unit that scans a form, photoelectrically converts the form image, and outputs a photoelectric conversion signal, a skew detection unit that detects the direction and degree of inclination of the form from the photoelectric conversion signal, and a skew detection unit that detects the direction and degree of inclination of the form from the photoelectric conversion signal. a gate circuit that uses the mask signal to perform mask processing on the photoelectric conversion signal output from the photoelectric conversion section; and a gate circuit that stores the photoelectric conversion signal extracted by the mask processing. a vertical projection unit that performs a vertical projection on a photoelectric conversion signal read from the image memory, and a recognition unit that uses the vertical projection to recognize a character pattern read from the image memory. An optical character reading device featuring: