JPH0115911B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION (1) Technical Field of the Invention The present invention relates to a line segment extraction method, particularly to a method for extracting only line segments from a drawing in which line segments and characters/symbols coexist.

(2) Prior art and problems Generally, symbols, characters, and line segments are written by hand on a document with grid axes, such as graph paper, and the information in this handwritten drawing is input into a data processing device such as a computer. However, methods are being used to obtain clean drawings.

However, in the conventional method, the work of inputting the necessary information from a drawing containing a mixture of characters, symbols, and line segments into a data processing device was entirely dependent on humans.

Therefore, input methods that require manual input, such as punch input, require a great deal of time and are simple and cumbersome for the operator, resulting in a problem of lower data processing efficiency.

(3) Purpose of the Invention The purpose of the present invention is to set a lattice point code that compresses information in the vicinity of each lattice point in a drawing containing a mixture of symbols, characters, and line segments, and to use this compressed code as a reference. Conventionally, by first extracting only the symbol, then separating this symbol from other characters and a group of line segments, and finally extracting only the line segments, this method conventionally took a lot of time and was troublesome for the operator. The aim is to eliminate manual input work and thereby improve data processing efficiency.

(4) Structure of the Invention According to the present invention, a symbol drawn based on a predetermined grid, a line segment between the symbols drawn along the grid axis, and a character denoting the name of the symbol. In the line segment extraction method for automatically inputting drawings containing a mixture of characters into a computer, an input image of a drawing containing a mixture of line figures (symbols, line segments) and characters drawn along predetermined grid axes is used. Processing step; A step of compressing and expressing the information possessed by the input image as a lattice point label code that expresses the graphic structure in the vicinity of the lattice points in lattice units; Extracting an initial lattice point label code and a plurality of initial lattice point label codes after the compressed expression step; This is the process of determining the grid point label code after the verification process and the extraction of the verification label code.
The first-order third verification process is performed using the four areas in the window as the reference to extract the first-order three-verification label code, and then the second four areas in the window are used as the reference and the second
The following third verification process is performed to extract the third verification label code, and finally, the third verification process is performed using the third four areas in the window as a reference, and the third verification label code is extracted.
includes a process of determining a third verification label code according to a specific decision logic after extracting the verification label code; after normalizing the input image on the grid axes by a second verification process; When performing projection processing to extract a label code, the gap direction is extracted by determining whether or not the projection results in each direction are continuous based on the projection results; The process of combining discontinuities in grid point label codes as ambiguous arms, and applying processing to change and modify them as reliable arms in subsequent processing to deal with breaks between one grid distance; a single line segment is extracted twice as a lattice point label code on two parallel lattice axes due to positional deviation, the state of the positional deviation of the line segment is estimated from the lattice point label code, and the assumed lattice A step of generating a reliable grid point label code only in the axis; and a process of first detecting an ambiguous grid point label code and sequentially searching for grid point label codes in directions along the four-way code;
When a reliable grid point label code sandwiching an ambiguous grid point label code appears, the ambiguity is interpreted as being due to local variations in the line segment, and the process is converted into a reliable grid point label code. ; the object to be recognized is a drawing in which line figures and characters are mixed, the image information is encoded in units of small rectangles, only the line segments are extracted, and the line segments are converted into vectors. A separate extraction method is provided.

(5) Embodiments of the Invention The present invention will be explained below by way of embodiments with reference to _the accompanying drawings _. (Figure 2)
For example, if a line segment V (Fig. 3) exists, the grid point
Centering on G ₅ and the range where the arrow shown in Figure 2 exists, determine whether or not an image exists, and if so, its certainty as line information,
In which direction of the grid point _G5 , top, bottom, left, or right, the graphic data exists, and the deviation from this to the position of this image are stored as information about the grid point _G5 . Then, based on the data of each of these lattice points, only the line segments of the handwritten drawing are extracted as vectors, which can be automatically input into a data processing device such as a computer.

The present invention first compresses information on an input image in units of rectangular areas (FIG. 5) with grid axes as a reference, and extracts grid point label codes (FIG. 4). The lattice point label code in Fig. 4 is 20 bits and is the final compressed information, and as is already known, the determination of this lattice point label code involves the extraction of the initial lattice point label code (LBL), the first This is performed after the verification process, extraction of the second verification label code (LB1), second verification process, extraction of the second verification label code (LB2), third verification process, and extraction of the third verification label code (LB3).

The present invention improves the third verification process by using an improved window SW' in which D ₃ ', L ₃ ', and R ₃ ' parts are added as shown in FIG. 5 instead of the conventional simple window. As a result, the third verification label code (LB3) is extracted, and the gap direction information of the 4th to 7th bits is added to the lattice point label code in addition to the conventional information.

(A) Improvement of the third verification process The extraction method is first
Based on the four areas D ₃ , L ₃ , U ₃ , and R ₃ , LB3' is extracted using the same method as the conventional method for extracting the "third verification label code (LB3)". An example of this is shown in FIG. Next, area D ₃ ′ in window SW′
Based on the four regions L ₃ ′, U ₃ ′, and R ₃ ′, LB3″ is extracted in a similar manner. An example of this is shown in FIG.
Finally, from LB3′ and LB3″ obtained in this way, the “third verification label code (LB3)” is generated using the following logic.
Determine.

Decision logic of LB3 First, assign LB3' to LB3. Next, R″=1&R=0→R=1, F _R =1, F _F =1, G
_R = 1, G _F = 1 R'' = 1 & R = 0 → R = 1, F _R = 1, F _F = 1, G
_R = 1, G _F = 1 U'' = 1 & R = 0 → U = 1, F _U = 1, F _F = 1, G _U = 1, G
_F = 1 R'' = 1 & R = 0 → R = 1, F _R = 1, F _F = 1, G
_R = 1, G _F = 1 U'' = 1 & R = 0 → U = 1, F _U = 1, F _F = 1, G _U = 1, G
_F = 1 L'' = 1 & L = 0 → L = 1, F _L = 1, F _F = 1, G _L = 1, G
_F = 1 R'' = 1 & R = 0 → R = 1, F _R = 1, F _F = 1, G
_R = 1, G _F = 1 U'' = 1 & R = 0 → U = 1, F _U = 1, F _F = 1, G _U = 1, G
_F = 1 L'' = 1 & L = 0 → L = 1, F _L = 1, F _F = 1, G _L = 1, G
_F = 1 D″ = 1 & D = 0 → D = 1, F _D = 1, F _F = 1, G _D = 1, G
Executes the process of _F = 1, corrects LB3, and
Generate information for G _F , G _D , G _L , G _U , and G _R in the figure.

By the above processing, the breakage of the line pattern as shown in FIG. 5 can be sufficiently absorbed. (Figure 8) (B) Addition of gap direction control extraction process After normalizing the input image on the grid axis by the second verification process, execute the projection process to extract the second verification label code. However, the gap direction is extracted by determining whether or not the projection results in each direction are continuous based on the projection results. Processing examples are shown in FIGS. 9 and 10.

Gap direction information (G _F , G _D , G _L , G _U ,
The method of integrating the grid point label codes except for G _R ) is completely the same as the conventional method. Regarding the gap direction, use the 4-way code (Fig. 4, 17~
20 bits) and put it in G _D ~ G _R (4 to 7 bits in Figure 4). Furthermore, if any of G _D to G _R is 1, set G _F to 1 and set F _F
Set to 1.

Through the above processing, the lattice point label code shown in FIG. 4 is extracted. All of the following processes of the present invention are performed based on this code information.

FIG. 11 is a flowchart of the process according to the present invention.

Processing: This is exactly the same as the “pair processing” proposed in the conventional method.

Line break correction; "Line break absorption" when extracting grid point label codes is effective for breaks up to at most 1/3 of the inter-lattice distance, but it cannot cope with breaks in the 1-lattice distance device.

Therefore, this process was introduced to absorb the discontinuity of one interstitial distance.

This process detects the state shown in Fig. 12 based on the grid point label code (Fig. 4), and in the case of Fig. 12 A, the grid point label codes of a and b are
In the case of Figure B, the lattice point label codes of c and d are corrected. Figure 12 is a pictorial reconstruction of the four-way code (D, L, U, R) of the grid point label code, and the parts indicated by broken lines are
Expresses don't care. When the state shown in FIG. 12A is detected, the lattice point label codes of a and b are modified as follows.

a; Set 20bit (R), 16bit ( _FR ), 0bit ( _FF ), 7bit (GR ₎ , and 3bit ( _GF ) of the grid point label code to 1.

b; Set 18bit (L), 14bit ( _FL ), 0bit ( _FF ), 5bit (GL ₎ , and 3bit ( _GF ) of the grid point label code to 1.

Similarly, when the state shown in FIG. 12 (b) is detected, the grid point label codes of c and d are modified as follows.

c; Set 17bit (D), 13bit (F _D ), 0bit (F _F ), 4bit (G _D ), and 3bit (G _F ) of the grid point label code to 1.

b; Set 19bit (U), 15bit (F _U ), 0bit (F _F ), 6bit (G _U ), and 3bit (G _F ) of the grid point label code to 1.

In this way, the discontinuities in the grid point label code due to breaks in the line pattern are combined as ambiguous arms, and by performing processing to change and modify them as reliable arms in subsequent processing, one grid distance Can handle minor cuts.

Misalignment correction: This is exactly the same as the "misalignment correction" proposed in the conventional method.

Ambiguity correction: This is exactly the same as the “ambiguity correction” proposed in the conventional method.

Classification of symbols: Symbols are extracted using the method proposed in the conventional method.

Separation of symbol areas: Same as the method proposed in the conventional method. However, the symbol area flag is set to 1 bit of the grid point label code in Figure 4.

Misalignment correction: When processing handwritten drawings, it is necessary to fully consider misalignment of line segments. Due to this positional shift, what is originally a single line segment may be extracted twice as a lattice point label code on two parallel lattice axes.

This process estimates the state of positional deviation of such line segments from the lattice point label code and generates reliable lattice point label codes only on the lattice axes considered to be intended by the designer. Here, a reliable lattice point label code is one in which the gap direction (4 to 7 bits), shift direction (8 to 11 bits), and ambiguous direction (13 to 16 bits) are all 0, and is called an ambiguous lattice point label code. indicates that any bit in the gap direction, deviation direction, or ambiguous direction is 1.

FIG. 13 is an example of this processing. First, a pair of adjacent grids having "misalignment directions" in opposing directions are captured, and this is interpreted as an expression of misalignment by a single line segment. Next, by sequentially searching for adjacent lattice point label codes based on the "four-way code", an axis on which a lattice to which a certain lattice point label code is attached is present is detected. Finally, while raising the axis to a reliable grid point label code,
Remove the other grid point label code.

Ambiguity correction: This process captures the "ambiguities" of grid point label codes from a broader perspective and improves the accuracy of grid point label codes. Eliminate "ambiguities" caused by characters drawn close to each other.

An ambiguous lattice point label code having a "gap direction" and an "ambiguous direction" is assigned to a lattice point that has an incomplete form as a line segment. Naturally, the same ambiguous grid point label code is given to grid points where characters exist, but here we wonder whether the "ambiguity" of this code information is due to local variations in line segments, or whether the characters Give an interpretation of whether it is caused by that,
Make corrections based on that interpretation.

FIG. 14 shows an example of this process. Specifically, an ambiguous lattice point label code is first detected, and lattice point label codes in directions along the four-way code are sequentially searched. In the process of this search, if a reliable grid point label code that is sandwiched between ambiguous grid point label codes appears, the "ambiguity" is interpreted as being due to local variations in the line segment, and the reliable grid point Convert to label code.

As a result of this processing, reliable lattice point label codes are interpreted as constituent elements of line segments, and lattice point label codes that retain "ambiguity" until the end are interpreted as character parts.

Thereafter, the line segments are traced between symbols based on the flag (S) indicating that they are symbol areas and the "4-way code", and are expressed by the network structure and vectors.

FIGS. 15 and 16 show an embodiment of the present invention. The overall configuration is the same as the conventional method.

(6) Effects of the Invention According to the present invention, it is possible to absorb line breaks and positional deviations peculiar to handwritten drawings, and remove the influence of characters drawn adjacent to line segments, so that drawings containing mixed characters and drawings can be eliminated. It is very effective for extracting line segments from inside.

[Brief explanation of drawings]

Fig. 1 is a diagram showing an example of a handwritten drawing, Fig. 2 is a schematic diagram of lattice points, and Fig. 3 is an explanatory diagram of lattice point data.
FIG. 4 is an explanatory diagram of the lattice point label code, FIGS. 5 to 8 are explanatory diagrams of the third verification process and third verification label code extraction, FIGS. 9 and 10 are explanatory diagrams of gap direction extraction, and FIGS. Fig. 11 is an explanatory diagram of the processing flow of the present invention, Fig. 12 is an explanatory diagram of the line break correction process, Fig. 13 is an explanatory diagram of the misalignment correction process, Fig. 14 is an explanatory diagram of the ambiguity correction process, and Fig. 15 is an explanatory diagram of the processing flow of the present invention. and FIG. 16 are configuration diagrams showing one embodiment of the present invention. 1... Image input device, 2... Image memory, 3...
... Verification circuit, 4... Reference point detection circuit, 5... Grid point table, 6... Grid conversion circuit (horizontal), 7...
...Lattice point label code generation circuit (horizontal), 8...
Lattice conversion circuit (vertical), 9... Lattice point label code generation circuit (vertical), 10... Address control section,
11...Control unit, 12...Verification window setting circuit, 13...LBL table, 14...LB1 table, 15...LB2 table, 16...SX1・
SY1 table, 17...SX2/SY2 table, 1
8... Address conversion circuit, 19... LB3 generation circuit, 20... LB3 table, 21... Grid point label code determination circuit, 22... Grid point label code table, 23... Pair processing circuit, 24... Line Pattern breakage correction circuit, 25...first deviation correction circuit,
26...First ambiguity correction circuit, 27...Second deviation correction circuit, 28...Second ambiguity correction circuit, 3
0... Dictionary memory for symbol classification, 31... Symbol area detection circuit, 32... Symbol area setting circuit, 33...
...Symbol classification table, 34...Similarity calculation circuit.

Claims (1)

[Claims] 1. A drawing in which symbols drawn based on a predetermined grid, line segments between the symbols drawn along the grid axis, and characters denoting the names of the symbols coexist. , in a line segment extraction method for automatic input processing into a computer, processing an input image of a drawing containing a mixture of line figures (symbols, line segments) and characters drawn along predetermined grid axes; A step of compressing and expressing the information possessed by the input image as a lattice point label code that expresses the graphic structure in the vicinity of the lattice points in lattice units; After the compressed expression step, extraction of the initial lattice point label code and verification processing and verification of multiple figures are performed. This is a step of determining grid point label codes after label code extraction, in which a first-order third verification process is performed using the first four areas in the window as a reference to obtain a first-order third verification label code. Then, the second four areas in the window are used as the reference to extract the second-order third area.
Verification processing is performed to extract the third verification label code, and finally, the third verification processing is performed using the third four areas in the window as a reference to extract the third verification label code. includes a process of determining a third verification label code according to decision logic; after normalizing the input image on the grid axis by a second verification process, a projection process to extract the second verification label code; When performing this process, the gap direction is extracted by determining whether or not the projection results in each direction are continuous based on the projection results; The process of combining them as ambiguous arms and changing and modifying them as reliable arms in subsequent processing to deal with breaks in one grid distance; When a lattice point label code on two lattice axes is extracted twice, the state of positional deviation of the line segment is estimated from the lattice point label code, and a lattice point label code that is reliable only for the assumed lattice axis is generated. First, an ambiguous lattice point label code is detected, and in the process of sequentially searching for lattice point label codes in directions along the four-way code, When a grid point label code appears, the ambiguity is interpreted as being due to local variations in the line segment, and is converted into a reliable grid point label code; and the recognition target is a line figure. The line segment extraction method is characterized in that the image information is encoded in units of small rectangles, only the line segments are extracted, and the line segments are converted into vectors.