JPH11238095A - Mail address reader - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a mail address reader with improved read performance as the result of being capable of more precise character segmentation suited to respective character kinds by judging the character kind of a printed character string or a handwritten character string beforehand. SOLUTION: An address row is detected from input image data from an image memory 7a in an address row detection part 7c, whether the character kind entered there is a handwritten character or a printed character is judged in a character kind judgement/character segmentation part 7b and the characters are recognized corresponding to the judged character kind. Also, in the case that the address entered row is composed of the printed character string, the character kind judgement/character segmentation part 7b equally divides the character string and segments the characters.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a mailing address reading apparatus having a function of detecting whether a character describing a destination such as a destination address or a recipient name is a printed character or a handwritten character.

[0002]

2. Description of the Related Art Conventionally, when characters are cut out from a character line image describing a destination such as a destination address or a recipient name and are recognized, a projection value scanned in the short side direction of the character line is first obtained. Characters that are cut at substantially equal intervals are selected so that both print characters and handwritten characters can be used as cutting point candidates, and characters are recognized.

[0003]

For example, as shown in FIG. 16, even in the case of handwritten characters describing "Tama-ku, Kawasaki-shi", even if the size of each character is slightly irregular but the intervals between them are almost equal, it is difficult to obtain. The character candidates can be cut out at substantially equal intervals according to the projection value given. However, as shown in FIG. 17, when the character intervals are not uniform or when there is a portion where characters are consecutive, the uniform If it is cut by, characters may be cut out by mistake.

On the other hand, in the case where the labeling method is used without using the above-described equal division method even for handwritten characters, individual characters can be cut out relatively accurately even if the size and spacing of the characters are somewhat irregular.

[0005] If it is known beforehand that a character is a printed character or a handwritten character before the character is cut out, an optimal method can be taken from the beginning correspondingly. Never did before.

In view of this, the present invention makes it possible to easily know whether a character is a printed character or a handwritten character before cutting out a character in a character line describing a destination, and to perform a character cutout optimal for each character type. It is an object of the present invention to provide a mail destination reading device configured so as to be able to perform the operation.

[0007]

According to the present invention, there is provided a mail destination reading apparatus, comprising: input means for inputting image data of a mail;
First detection means for detecting a destination description area in the image data input by the input means, and second detection for detecting a destination description line included in the destination description area detected by the first detection means And a determination unit for determining whether the character type described in the destination description line detected by the second detection unit is a handwritten character or a printed character.

With the above arrangement, it is possible to easily determine whether a character to be read before printing a character is a printed character or a handwritten character, so that the character can be cut out by a method most suitable for each character type.

[0009]

DESCRIPTION OF THE PREFERRED EMBODIMENTS A first embodiment of the present invention will be described below with reference to the drawings.

Referring to FIG. 1, the configuration of a destination reading / sorting machine 1 to which the present invention is applied will be described first. The destination reading sorting machine 1 includes a supply unit 2 that accommodates a plurality of randomly arranged mail items P in a standing position with their rear ends aligned and sequentially supplies the mail items P to a predetermined take-out position. I have.

A take-out section 4 is provided at the take-out position of the postal matter P for taking out the postal matter P supplied to the take-out position by the supply section 2 one by one onto the main transport path 3. Each postal matter P is placed on the main transport path 3 for transporting the postal matter P taken out by the take-out section 4.
A scanner unit 5 of an optical character reader (OCR) for optically reading destination information such as a postal code, a destination address, and an address is provided. The reflected light obtained by scanning the mail surface from the scanner unit 5 is photoelectrically converted by the photoelectric conversion unit 6,
The obtained image signal of the mail P is supplied to the discriminating unit 7 in the state of a multi-value signal.

The main transport path 3 on the downstream side of the scanner section 5 extends below a section stacking section 8 described later. On the main transport path 3 and on the right side of the sorting / stacking unit 8 in the drawing, an ink jet printer that prints a bar code corresponding to the destination information read by the discrimination unit 7 on the mail P on the mail P. Printing section 9 is provided. Further downstream of the barcode printing section 9 is provided a barcode reader 10 for reading destination information from the printed barcode and confirming whether or not the barcode has been correctly printed.

The end of the transport path 3 is connected to a sorting / stacking unit 8 for sorting mail items P into predetermined sorting pockets for stacking. Note that the barcode printing unit 9 and the barcode reader 10 of the transport path 3 actually have a distance of about 10 to 15 m for the purpose of drying the ink of the barcode printed on the mail P by the printing unit 9. Is provided.

The section stacking section 8 has a plurality of section pockets 8a divided into a plurality of steps in the horizontal direction and a plurality of rows in the vertical direction, and extends between the steps above each section pocket 8a. A stepped path section 11 is provided. Each of the step pass sections 11 is provided with a plurality of division gates (not shown) for selectively communicating the step pass section 11 with each of the division pockets 8a.

Further, at a position near the end of the transport path 3 connected to the sorting / stacking section 8 and corresponding to each step path section 11 of the sorting / stacking section 8, the transport path 3 is selected to each step path section 11. There are provided a plurality of switching gates (not shown) for making the connection.

The destination information read by the determination unit 7 using the OCR 13 under the control of the control unit 12 is finally supplied to the control unit 12 and stored therein. A sorting destination of the mail is determined from the read destination information, and a switching gate and a sorting gate (not shown) provided in each of the sorting pockets 8a are selectively switched according to the determined sorting destination. Are divided into predetermined division pockets 8a.

The destination reading / sorting machine 1 configured as described above reads the destination of the postal matter and, based on this, sorts the
A control unit 12 including a CPU for controlling the sorting operation is provided, and on the left side of the sorting machine 1, a cursor mode or a specific company mode setting for the operator in response to an operation input by the operator, operation guidance, and the like are displayed. An operation / display panel 14 is provided. Although not shown here, the control unit 12 is provided with a memory for storing the reading result, an address dictionary and a recipient dictionary for reading the destination of the mail.

FIG. 2 is a functional block diagram of an apparatus according to an embodiment of the present invention for explaining the destination reading operation of the mail by the photoelectric conversion unit 6, the discrimination unit 7, and the control unit 12 shown in FIG. In FIG. 2, image data on the surface of a mail is taken in from a scanner / photoelectric conversion unit, that is, an image input unit 6, and sent to an image memory 7a in a discrimination unit 7. Although FIG. 1 shows the scanner 5 and the photoelectric conversion unit 6 as separate units, FIG. 2 shows the scanner / photoelectric conversion unit as the image input unit 6.

In the following description, each function execution unit operates under the control of the control unit 12, although not otherwise specified.

The image data obtained from the image input unit 6 is temporarily stored in an image memory 7a in the discriminating unit 7, read out, and supplied to a character type discriminating / character extracting unit 7b according to the present invention. The character type discriminating / character extracting unit 7b detects whether the destination character described in the destination description area is a printed type or a handwritten character from the image data of the mail, and based on the detection result. The individual characters are cut out, and the detailed configuration and operation will be described later.

In order to detect whether a character describing a destination is printed or handwritten, first, a destination description area is detected from an image of a postal matter, and a character line describing a destination included in the detected destination description area, that is, a destination is described. The row is detected by the destination row detection unit 7c.

For example, in the mail image shown in FIG.
Six blocks of “Kawasaki-shi”, “Yuki-ku”, “Yanagimachi”, “70”, “Co., Ltd.”, and “Tokyo” are extracted from the proximity state of characters and the like. These extracted blocks are integrated from, for example, the relative positional relationship with the postal code description frame F above the postal matter, and determined as destination lines A1 to A3 (portions surrounded by broken lines).

When the destination line is detected in this way, a label is obtained by the next label detection unit 7d for each character of the detected destination line, and a circumscribed rectangle of the individual label is detected. For example, assuming that a destination character line "Tama-ku, Kawasaki-shi" as shown in FIG. 4 is detected by the destination line detection unit 7c, a circumscribed rectangle L of a label is detected for each of these characters.

The label obtained in this way is stored in the first label memory 7e together with the detected circumscribed rectangle, and sent to the image preprocessor 7f. The image preprocessing unit 7f removes, for example, a label Ls having a smaller size than the other labels L shown in FIG. 4 as noise,
A process of integrating three labels L arranged in the short side direction like the character “river” is performed. In this way, noise removal,
The label data that has been subjected to the label integration processing is stored in the second label memory 7g and is selectively supplied to the character type determination / character cutout unit 7b according to the image data stored in the image memory 7a and the purpose of use.

Hereinafter, the operation of the character type judging / character extracting section 7b will be described in detail with reference to FIGS.

First, in step S1 in FIG. 5, the character type is determined by the histogram, that is, whether the destination character line is described in print or a handwritten character is determined. The determination of the character type based on the histogram in step S1 is performed according to the flowchart of FIG. 6, and thus will be described in detail with reference to FIG.

In the first step S11 in FIG. 6, the destination line of the image data read from the image memory 7a in FIG. 2 is specified from the coordinate data acquired by the detection unit 7c, and the black pixel portion is determined from the obtained line image data. A contour line that is a continuation of contour points is obtained.

As shown in FIG. 7, the contour points of the black pixel portion are the contour points of the contour point black pixel C constituting the contour line of the portion bent at an obtuse angle as shown in FIG. , The black pixel Cp at the forefront is defined as a “projected contour point”. (B) is a contour point where the foremost black pixel Cp of a portion bent at an acute angle of 90 degrees or less in a line segment constituting a character is projected. However, if it is determined that the contour point is in the middle of a line segment even at the inflection point of the contour line as shown in (c), the contour point black pixel C of the inflection point is not regarded as a “projecting contour point”. . In this case, the direction of a line segment connecting the leading edge black pixel Cp shown in FIG.
When the direction of the line segment connecting the black pixel Cp and the right outline point intersects at a predetermined angle or less, the foremost black pixel Cp is defined as a protruding outline point.

In the case of FIG. 7 (b), they intersect at a more acute angle than in FIG. 7 (a), and are recognized as protruding contour points. However, in the case of FIG. 7 (c), the direction of the line segment on the left side and the direction of the line segment on the right side indicate an angle that is equal or more than a predetermined angle and is nearly parallel. It is determined to be in the middle of a minute.

For example, if the destination line obtained by the destination line detecting unit 7c is "Tama-ku, Kawasaki-city" as shown in FIG. 8A, the long side of this destination line is determined in step S12 of FIG. As shown in FIG. 8A, a contour point Cp protruding from a neighboring contour point is detected from among the contour points C. Here, the outline points are shown by white circles. The protruding contour point Cp is as shown in FIG. 7 (a) or (b).

Of the detected contour points Cp, the contour points spaced apart to some extent are connected by a line segment in the next step S13, and the angle formed by the line segment with the long side direction is taken as a histogram. .

FIG. 8B shows an example of a histogram H obtained in this manner. When the histogram H is viewed, the number of line segments is concentrated on a specific line segment slope Hz, and the peak Hp is reduced. It turns out that it forms. Step S
In 14, it is determined that when the peak value Hp exceeds a certain threshold Th, the upper and lower ends of the character string are arranged substantially linearly in the direction of the angle Hz.

As described above, in the case of a horizontally written character as shown in FIG. 8A, when it is determined that the contour points at the upper and lower ends of the character string are arranged in substantially the same direction, this character is determined in step S15. The column is determined to be a print type string. Similarly, although not shown, when the character string is written vertically, the character string is also a print type character string when it is determined that the contour points are arranged in substantially the same direction at the left and right ends of the character string. Needless to say, it can be determined.

On the other hand, if these conditions are not satisfied, it is determined that the input character string is a handwritten character string. This will be described later with reference to FIG.

As a result of the character type determination based on the histogram, when it is determined that the character is a print type, the process proceeds from step S1 in FIG. 5 to step S3 via step S2, where the character type determination based on the label is performed. The character type determination based on this label will be described in detail with reference to the flowchart in FIG.

In FIG. 9, in the first step S31, the label data read from the second label memory 7g in FIG. 2 is received. As shown in FIG. 10, the label data is data representing the size (w × h) of the circumscribed rectangle of each label and the coordinates (x, y) of the corner position of the circumscribed rectangle.

In step S31, it is checked whether or not the heights of the circumscribed rectangles of the received labels are substantially equal. That is, it is checked whether or not the height h, which is one of the circumscribed rectangle information of the labels of the characters, is substantially equal to the character string “Tama-ku, Kawasaki-shi” shown in FIG. If the result of this check is YES, processing transfers to step S32, where it is determined that the character string in the destination line is a print character string.

If it is determined in step S31 that the heights h are not the same, the process proceeds to step S33, and the remaining information of the circumscribed rectangle information of each label, that is, the size (w)
It is checked whether any one of the width w for determining × h) and the label center distance D is substantially aligned.
When it is detected that any one of the width w and the label center-of-gravity interval D is substantially equal, the process proceeds to step S34 to perform a process of determining a print character string. If none of the above conditions is met, the process moves to the next step.

If it is determined in step S32 or S34 that the character string is a print character string, the flow shifts from step S4 in FIG. 5 to step S5 to output the character type determination result from the character type determination / character cutout unit 7b in FIG. And the result of extracting the characters is output.

If it is determined in step S4 in FIG. 5 that the height and width of the label are not the same, characters are cut out from the print character string by the label in the next step S6. Hereinafter, the operation of character cutout using this label will be described in detail with reference to the flowchart of FIG.

In FIG. 12, in the first step S61, the heights h of the obtained labels are compared, and if the size is significantly different from one of the adjacent two adjacent labels, those labels are combined. For example, as shown in FIG. 13A, of the label of the character string "Kawasaki City", "kawa"
Are significantly smaller than the height of the connected label of "Sakiichi". In such a case, the label of the character string "Kawasaki" is combined. As a result, the height hc of the combined label is several times the width w.

In this case, the process proceeds to step S62, and when the height hc of the label obtained by the combination is several times the width w, a value obtained by dividing the height hc by a value substantially equal to the width w (an integer). Is rounded to the basic division number n. In the case of FIG. 13, the basic division number n is n = 3 as shown in FIG. When the division labels are divisible, the obtained divided labels are labels having substantially the same height and width.

When the basic division number n is obtained in step S62, if there is room for the number of extraction candidates, the division number is varied by ± 1 with respect to the basic division number n in the next step S63 (n, n + 1). , N-1), and processing is performed in which individual division labels obtained by division with the respective division numbers are used as character candidates. In the case of FIG. 13, since the basic division number is n = 3, the variable division numbers are 3, 4, and 2. In FIG. 13, the character string "Tama-ku" is similarly divided by the division numbers 3, 4, and 2.

Further, the width / height may be changed only to a direction closer to 1: 1 without changing both the plus and the minus. In the case of a handwritten character string, each label is a character candidate. However, a small-sized label may be combined with an adjacent label. This handwritten character will be described later in detail.

As described above, when characters are cut out by labels in step S6 of FIG. 5, it is checked in step S65 that the heights of the circumscribed rectangles of the divided labels are substantially the same. The process proceeds to S66 to perform character type determination and character clipping result output processing.

If the heights of the divided labels are not uniform, the process proceeds from step S65 to step S67, where it is checked whether the widths of the circumscribed rectangles of the labels or the character intervals are substantially equal. If YES, the process also proceeds to step S66, where character type determination and character clipping result output processing are performed.

When it is determined in steps S65 and S67 that none of the heights, widths, and character spacings of the divided labels are aligned, the input character string is once determined to be a print character string, and the process has proceeded. With this as an error, in the following step S7, the input character string is used as a handwritten character string, and characters are cut out by labels.

The character is cut out by the label in step S7 according to the flowchart shown in FIG. In the process shown in FIG. 14, the integrated label read from the second label memory 7g in FIG. 2 is used.

In the first step S71, each read label is used as a character candidate. In the next step S72, the height and width of the label are compared, and when it is detected that the height of the label is several times the width, a projection scanned in the short side direction within the label is taken.

For example, when the handwritten character string image "Kawasaki-shi" as shown in FIG. 15A is read from the image memory 7a, a projection scanned in the short side direction (x) is obtained, and the projection is performed in step S73. Small value parts J1, J2, J3
Is detected, and a point at which the projection value of this portion is equal to or less than a threshold value TH and a continuous portion are set as cutting candidates.

In step S74, it is determined whether or not the continuous portion is long. If it is long, the continuous portion is made a single line in step S75, and both ends are cut. However, J3 is also a single line, but is not cut because it is the end of the label.

Each of these portions J1 to J3 is a single line, but if a continuous portion is short, one point (for example, the middle) in the middle of the portion is set as a cut portion in step S76. Since the continuous portion is short at the portion J2 in FIG. 15A, it is cut in the middle.

When the result of determination that the character is a handwritten character and the result of cutting out the character are output in step S8, these result data are sent to the character candidate creating unit 7h of FIG. Is created. Here, the label obtained by the above-described method is used as a character candidate, and in addition, as shown in FIG. 15B, the label is combined with an adjacent label, and the height of the combined label is equal to or less than a threshold value. Let the object be a character candidate. However, in the case of the above-mentioned cut single line, it is not necessary to independently set it as a character candidate.

The above-described character cutout by label and the character type discrimination / output operation of the character type when the character string is determined to be a handwritten character string in steps S7 and S8 described above are performed after the character string is determined as a handwritten character in step S2 in FIG. The processing steps S9 and S10 can be executed with exactly the same contents.

As described above, the characters are separated into individual characters by the labeling method, the projection method, or the like, and the character candidates created by the character candidate creating section 7h for the individual cut-out characters are used in the subsequent character recognition. The character recognition is performed by the unit 7i, and further, the town name, the recipient name, and the like are recognized by the destination determination unit 7j based on the recognized characters.

When reading the postal address, the address and the recipient name described in the destination area are recognized and read, and at the same time, the postal code is read and the result of the destination reading is confirmed and supplemented. I have.

The seven-digit zip code read in this way is compared with the acquired destination address, and the destination finally determined is determined. The obtained destination information is output to the reading result output unit 14.

[0058]

As described in detail above, according to the present invention,
By determining the character type of a printed character string or a handwritten character string in advance, a more precise character cutout suitable for each character type can be performed. As a result, it is possible to provide a mail destination reading device with improved reading performance. it can.

[Brief description of the drawings]

FIG. 1 is a configuration diagram showing an overall configuration of an embodiment of the present invention.

FIG. 2 is a functional block diagram of the embodiment of FIG.

FIG. 3 is a diagram showing a destination line from which a destination of a mail is read.

FIG. 4 is a diagram showing a result of labeling of a destination line character.

FIG. 5 is a flowchart illustrating a character type determination / character cutout operation according to an embodiment of the present invention.

FIG. 6 is a flowchart showing details of character type determination based on the histogram of FIG. 5;

FIG. 7 is a view showing detection of a protruding contour point for acquiring the histogram of FIG. 6;

FIG. 8 is a view for explaining the operation of FIG. 6;

FIG. 9 is a view showing a method of character type determination based on the labels in FIG. 5;

FIG. 10 is a diagram showing coordinate data of a label and data indicating the size of the label.

FIG. 11 is a diagram showing a result of labeling of a character string having substantially the same height, width, and center of gravity interval.

FIG. 12 is a flowchart illustrating an operation of extracting characters from a print character string.

FIG. 13 is a diagram illustrating a method of creating a divided label by combining character string labels that are not print characters.

FIG. 14 is a flowchart illustrating character extraction of a handwritten character string using a label.

FIG. 15 is a view for explaining a method of cutting and integrating the result of projection of a handwritten character string.

FIG. 16 is a view showing a result of projection of a handwritten character.

FIG. 17 is a diagram showing a result of projection of a handwritten character.

[Explanation of symbols]

 6 image input unit, 7 discriminating unit, 7a image memory, 7b character type determination / character cutout unit, 7d label detecting unit, 7e first label memory, 7f image preprocessing unit, 7g second label Memory, 12 ... Control unit.

Claims (10)

[Claims] 1. An input means for inputting image data of a postal matter, a first detection means for detecting a destination description area in the image data input by the input means, Second detection means for detecting a destination description line included in the destination description area, and whether the character type described in the destination description line detected by the second detection means is a handwritten character or a printed character A mail destination reading device, comprising: determining means for determining 2. The mail destination reading apparatus according to claim 1, further comprising a cutout unit that cuts out a character according to the character type determined by the determination unit. 3. The character extracting device according to claim 2, wherein, when the destination description line is a print character string, the character extracting means performs a character extracting operation by equally dividing the character string.
The mail address reading device according to 1. 4. The cut-out means, when the destination description line is a handwritten character string, obtains a projection value in a line direction of the destination description line and a direction orthogonal thereto, and based on the obtained projection value. 3. A mailing address reading apparatus according to claim 2, further comprising means for cutting out characters. 5. The apparatus according to claim 4, wherein the means for extracting a character based on the projection value sets a point where the projection value is equal to or less than a predetermined threshold value and a continuous part as a cutting point candidate. Mail address reader. 6. The cut-out means, when the destination line is a handwritten character string, detects a single line connecting characters in the character string, and the single line is longer than a predetermined value. Means for combining the single line with one of the two characters to create a character candidate, and cutting the middle of the single line to create two character candidates when the single line is shorter than a predetermined value. 3. The mail destination reading device according to claim 2, further comprising means. 7. The determining means obtains a histogram of inclination of a line segment connecting the detected contour points on the long side of the destination line, and determines the character type from the contents of the histogram. 3. The mail destination reading apparatus according to claim 2, further comprising: 8. The labeling means for labeling the destination line, a height of a circumscribed rectangle of each label obtained by the labeling means, and
3. The mailing destination according to claim 2, further comprising: means for determining a print character when any of the widths and the character intervals are substantially equal, and determining a handwritten character string when none of the widths and the character intervals correspond to each other. Reader. 9. When the height of a circumscribed rectangle of a label obtained by the labeling means is several times the width, the cutout means sets a value obtained by dividing the height by the width of the label as a basic character division number. 9. The mail destination reading apparatus according to claim 8, further comprising means for determining and cutting out characters. 10. The mailing destination according to claim 9, wherein each label obtained by changing the number of divisions by ± 1 with respect to the basic character division number and dividing by each division number is used as a character candidate. Reader.