JPS58207180A - Character detecting and segmenting device - Google Patents

Abstract

PURPOSE:To eliminate the noises touching an entry frame, by obtaining the coordinates of the entry frame and then the coordinates of the continuous blocks within the entry frame and then comparing these coordinates with each other to check whether the entry frame has a contact with each block. CONSTITUTION:The read signal of a mail 1 supplied from a photoelectric converting part 2 is supplied to an entry frame coordinate detecting part 3. The part 3 reads both horizontal and vertical projections of a red signal of the signal supplied from the part 2 to obtain the coordinates of four corners of each postal code number entry frame 1a. A quantitizing part 4 quantitizes only the black signals in the coordinates for each frame supplied from the part 3. A character pattern supplied from the part 4 and equivalent to a character is converted into the coordinates of white and black points with every plural rows and columns through a converting part 5. A block coordinate detecting part 10 obtains the continuous block coordinates from the converted coordinates. Then a contact between the entry frame and the block is detected from the block coordinates and the output of the part 3.

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention provides, for example, a pre-processing for the recognition process of a character reading device that reads a postal code in a postal code entry frame on a mail item, by transmitting a reading signal to the character iP turn of each character in the entry frame. The present invention relates to a character detection and cutting device for detecting and cutting out characters.

[Technical background of the invention and its problems]

In this type of character detection/cutting device, noise a that touches the entry frame as shown in FIG. 1 often appears as an input turn.

In this case, the noise is large, so conventionally the sizes of each block (character element) are relatively compared and small blocks are removed as noise, or the position of each block is examined and the area for each position is Methods that remove noise by comparing it with a reference value have not been able to remove noise adjacent to the entry frame as described above.

[Purpose of the invention]

The present invention has been made in view of the above circumstances, and its purpose is to provide a character detection and cutting device that can remove noise that comes into contact with a writing frame using a simple circuit.

[Summary of the invention]

This invention calculates the coordinates of a writing frame, calculates the coordinates of consecutive blocks within this writing frame, and compares these coordinates to detect whether or not the writing frame and each block contact each other. Depending on the result, a character pattern is output from which blocks that touch are removed.

[Embodiments of the invention]

An embodiment of the present invention will be described below with reference to the drawings.

In FIG. 2, numeral 1 indicates an object to be read, such as a postal item;
A postal code entry frame 1a is provided in advance on this postal item. Reference numeral 2 denotes a photoelectric conversion section, which obtains a turn signal by optically scanning the mail piece 1 and photoelectrically converting it. 3 is an entry frame coordinate detection unit;
This is to obtain the coordinates of the four corners of each postal code entry frame 1a by taking the horizontal and vertical projections of the red signal in the signal from the photoelectric converter 2. 4
5 is a quantization unit, which quantizes only the black signal within the coordinates from the frame coordinate detection unit 3 among the electric signals from the photoelectric conversion unit. 5 is an XI pattern conversion unit;
This is the quantization pattern from the quantization unit 4, that is, the bit pattern for a character, and the coordinates that change from a white point to a black point for each line, that is, the black point coordinates, and the coordinates that change from a black point to a white point, and the white point coordinates. It is something that converts. That is, when a bit/turn for one character as shown in FIG. 3 is supplied, it is converted into an XI pattern as shown in FIG. 4. For example, channel] is blank, so the black point coordinates are r8, Oj', and the white point coordinates are r8. O
J, and channel 5 is the sunspot coordinates! ri r 8.2
'', the white point coordinates are [8, A'', channel] 11, the black point coordinates are ro, 2'', r8.6'', the white point coordinates are r C
1, 3J, r 8, 8J. To the left of the above coordinate data, "8" is added when there is one data, "ro" when there are two, "]" when there are three eight balls, and "0 and "8" respectively. ing.

The 61-1i block determination unit compares the two channels and identifies consecutive blocks with the same number.
In addition, when discontinuous blocks merge, the merge is determined and the correspondence relationship is stored in the corresponding block memory 8. That is, as shown in FIG. 5, the corresponding block is determined by detecting the connection between the white point coordinates and the black point coordinates of the upper and lower two channels. For example, the sunspot image of the upper channel is r A L J , and the white point coordinates are r A
R, J, the black point coordinates of the lower channel are rBLJ, the white point coordinates are rBRJ, and when [AR≧BL and AL≦BRJ hold, it is detected that the two XI patterns of the above two channels are connected. , otherwise, it is detected that the X■ patterns of the two channels are not connected. For example, if XI-7 turns as shown in FIG. Due to the generation of a new block on the left side of the It can be stored in the block number table 7.

10, 9 is a block number table editing section, which stores the contents of the block number table 7 as the contents of the block memory 7, assuming that one character's worth of pro/ri buds is stored in the pro/ri number table 7. , and the same block number is given to the same block. Even if you get XT/k turn for a character/g turn as shown in 1d1 Figure 1, first
i? For the turn, a block number ``2'' is stored in the block number table 7, and then a new block is generated on the left, so a block number ``4'' is stored, and then a new block is generated on the right 1dll. Then, when blocks r2J and r4J merge, block number "8" is stored for them, and blocks r2j and r6J
When the blocks merge, the block number "10" is stored for that block.

Further, when blocks r2Jr4J merge, the block determination unit 6 stores the block "2" as "8" and "4" as "8" in the corresponding block memory 8, and blocks "2" and "4" as "8".
6" merge, the block determination unit 6 determines that the block r
2 J CarlOJ.

"6" is stored in the corresponding block memory 8 as "10". As a result, the block number for one character of
When stored in the corresponding block memory 8, the block number table editing unit 9 stores the corresponding program stored in the corresponding block memory 8.
→"8".

"4" → r 8 J, r2J → rlOJ, r6J → 71
By editing the contents of the block number table 6 according to the contents of 0J), all block numbers are set to ``10''.
Using the XI pattern of the f-turn conversion unit 5 and the block number of the block number table 7, obtain the horizontal projection and vertical projection of the It is something to detect.

No. 1 is a block area calculation unit, which corresponds to the same block number in the block number table 7.
This is a challenge to calculate the area of each block by summing up the difference between the number of black points and the black point coordinates for nine turns. Reference numeral 12 denotes an entry frame and a block contact detection section, which detects the coordinates of the four corners of the entry frame 1a supplied from the entry frame coordinate detection section 3 and the coordinates of the four corners of each block from the block coordinate detection section 10. Second, it detects whether each block is in contact with the entry frame 1a, and if so, it is detected where on the top, bottom, left, or right of the entry frame it is in contact. Reference numeral 13 denotes a noise block determination unit, which calculates the dimensions and block area of each block according to the above-mentioned entry frame and the contact content of each block from the block contact detection unit 12, and the coordinates from the coordinate detection unit 10 of the block. The area from the section 11 is compared with the size reference value and area reference value from the reference value table 14, and blocks having a value smaller than the reference value are determined to be noise. The reference value table 14 stores reference values for the dimensions and area of the parts that touch the top, bottom, left and right of the entry frame. Reference numeral 15 denotes a noise block removal section, which outputs a pattern signal obtained by removing noise blocks from the quantization pattern from the quantization section 4 according to the judgment result from the noise block judgment section 13.

Note that each of the above-mentioned parts is controlled by a control part (not shown).

Next, the operation in such a configuration will be explained. For example, right now, light 1. The reading signal of the postal item 1 from the conversion section 2 is supplied to the frame coordinate detection section 3.

Then, the entry frame coordinate detection unit 3 obtains the coordinates of the four corners of each postal code entry frame 1a by taking the horizontal projection and vertical projection of the commission signal in the signal from the photoelectric conversion unit 2. As a result, the quantization unit 4 quantizes only the black signal within the coordinates of each frame supplied from the frame coordinate detection unit 3, thereby obtaining a bit pattern as shown in FIG. 3, for example, and an XI pattern. It is output to the converter 5.

Then, the X■ pattern converter 5 converts the character pattern into white point coordinates and black point coordinates for each line and outputs XI/4' turn data as shown in FIG. Accordingly, the block determination unit 6 compares the X work a + turn of each channel with the XI pattern of the next channel, and stores successive blocks in the block number table 7 with clear block numbers. In both cases, the correspondence relationship is determined by checking the connection of each block and stored in the corresponding block memory 8. Then, when the storage of block numbers for -characters of XI/g turns is completed, the block number table editing section 9 rewrites the contents of the block number table 7 according to the contents of the corresponding block memory 8. In this case, there are two blocks or there is no corresponding block, so
The stored contents of block number table 2 are not rewritten.

The output of the XIa# turn conversion section 5 and the contents of the block number table are supplied to the block coordinate detection section IO and the block area calculation section NO. As a result, the block coordinate detection unit 10 obtains the horizontal projection and vertical projection of the XI/arm turn corresponding to the same block number supplied from the block number table 7.
The coordinates of the four corners of each block are detected, and the detection results are output to the noise block determination section 13.

In addition, the block area calculation unit 11 also calculates the
(i) Calculate the area of each block obtained by summing up the number of bits of 4 turns, that is, "white point coordinates - black point coordinates" for each XI pattern, and output this calculation result to the noise block determination section 13. For example, the horizontal projection of the 4 turns of XI corresponding to block number "2" shown in FIG.
The 9th bit starts from the 9th bit, and the coordinates of its four corners are 1c.
r ] r :) j r 9 y 5j r ]
, 20 Jr9, 20J. Also, the block number “
3” corresponding to X IA? The horizontal projection of the turn is 11
From the 14th bit to the 14th bit, the vertical projection is from the 0th bit to the 3rd bit, and the coordinates of the four corners are ro, 1]J
r4. ]]jrO, ]4Jr4, 14J. Furthermore, the XI Noreen bit corresponding to block number “2”,
:':l'''In other words, in channel 5, the white point coordinate r
AJ-black point coordinate rlJ=9J is performed every XI-turn, and the cumulative total becomes "50" bits. Also,
The total number of bits of the XI pattern corresponding to the professional number "3" is 1-8" bits.

On the other hand, the coordinates of the four corners of each program and the coordinates of the coordinate detection unit 10 are supplied to the entry frame and block contact detection unit 12. Then, this detection 'h12 detects contact between the four corners of the two blocks supplied and the sitting/drifting of the writing frame 1a of the writing frame coordinate detecting section 3. In case of rice,
In the coordinates of the four corners of block "2" are ro, IIJrO,
14J, it is not detected that this block "2" is in contact with the left side of the entry frame. Through this detection, the noise block determination section 13 detects the block "2" from the entry frame and the block contact detection section 12. '' is in contact, the dimensions are determined from the coordinates of the four corners of block B from the block coordinate detection unit 10, and compared with the dimension standard values in the standard value table 14, and the block area is still determined. The area for block "2" from the calculation unit 11 is compared with the area reference value in the reference value table 14. As a result, the noise block determination unit 13 determines that the block “2
It is determined that the size and area of " are below the reference value, and this determination result is output to the noise block removal section.

As a result, the noise block removal unit 15 outputs a character pattern from which block "2" has been removed, as shown in FIG.

〔Effect of the invention〕

As detailed above, according to the present invention, with a simple configuration,
It is possible to provide a character detection and cutting device that can remove noise that comes into contact with a writing frame.

[Brief explanation of drawings]

Fig. 1 is a diagram for explaining characters containing noise and 1,000 turns in contact with the entry frame, Figs. 2 to 8 show an embodiment of the present invention, and Fig. 2 shows character detection cutting A schematic block diagram showing the configuration of the device and its surroundings, Figure 3 is a diagram showing an example of bit soturn, Figure 4 is a diagram showing the XI pattern for each channel, and Figure 5 is a diagram showing the connections between each program. Figure 6 is a diagram showing an example of storing a block number table, Figure 7 is a diagram showing an example of a character pattern to explain the block number table editing section, etc., and Figure 8 is a diagram showing noise block removal. It is a figure which shows an example of the character/mother turn output from a part. 1...Postal item (object to be read), 1a...Entry frame, 3.
... Entry frame coordinate detection unit, 4... Quantization unit, 5...X
I-pattern conversion section, 6... Block determination section, 7...
Block number table, 10...Block coordinate detection unit, 12...Entry frame, professional contact detection unit, 13...
- Noise Pro, ri discrimination section, 15... noise block removal section. Applicant's agent Patent attorney Takeshi Suzue Intent 4
Figure 6 Figure 5 Figure 7rj4 Figure 8

Claims (1)

[Claims] By scanning the object to be read and converting it into an electrical signal,
In a character detection/cutting device used in a character reading device that reads character information written in a writing frame on an object to be read, the coordinates of the writing frame are determined by a signal obtained by scanning the object to be read. a quantization section that quantizes the scanning signal within the entry frame obtained by this means; and a character pattern for one character from the quantization section in a plurality of lines. Conversion means for converting into a pattern of black point coordinates and white point coordinates for each column, and block coordinate detection for determining the coordinates of consecutive blocks based on the black point coordinates and white point coordinates for each row and column obtained by this conversion means. detect whether or not the entry frame and each block come into contact based on the block coordinates obtained by the block coordinate detection section and the coordinates obtained by the entry frame coordinate detection means, and according to the detection result. 1. A character detection and cutting device, comprising means for outputting a character pattern from which blocks that touch a writing frame are removed.