JPS5935278A - Recognition system of line figure - Google Patents

Abstract

PURPOSE:To speed up correction processing by correcting coordinate value after line recognition in a pattern recognizing device. CONSTITUTION:Line data read from a drawing is recognized without coordinate conversion to recognize feature points constituting the line. The coordinates of the feature points are converted into coordinate values X' and Y' on the drawing by using the coordinates of Pox and Poy, Pix and Piy, and P2x and P2y of the reference points P0, P1 and P2 on the drawing. Then, the converted coordinate values are corrected so that the coordinate values after the conversion have constant quantizing values and the line expressed by the coordinate values after the conversion has a previously specified direction. Therefore, the rotation position shift correction of a large-sized drawing, grid-formation coordinate correcting processing with specified precision, and line direction coordinate correction processing are carried out at a high speed.

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention provides high-speed processing for drawings in which lines are drawn on a grid printed on drawing paper in only specific directions (for example, horizontal, vertical, ±45 degree directions). Line figure recognition method for accurate recognition Conventional experimental systems that recognize line figures drawn on printed grids and allowed only in specific directions (horizontal, vertical, ±45 degree directions) Although it has been reported (IEICE, Pattern Recognition and Learning Study Group Materials PP97-104 PPL-80-53
) One of the practical problems is the correction of the rotational positional deviation of the drawing paper with respect to the device reference coordinates. Published reports adopt a method in which lines are recognized based on line data after a positional shift correction process (coordinate transformation) is applied to each picture element of the obtained video data. However, when dealing with large screens, the problem is that correction processing for each pixel takes too much time.

SUMMARY OF THE INVENTION In order to solve the above-mentioned problems, it is an object of the present invention to provide a line figure recognition method that requires correction processing only for a small number of picture elements.

In order to achieve this purpose, the line data read from the drawing is recognized without coordinate conversion to obtain the coordinates of the feature points that make up the line, and the coordinates of the feature points on the drawing are Using the coordinates of the reference point (origin, reference point on the X-axis, reference point on the X-axis), add processing to convert the coordinates to coordinate values on the drawing.

After that, the coordinate values are adjusted so that the coordinate values after the transformation have a certain quantum value (relative to the grid points on the drawing), and so that the line represented by the coordinate values after the transformation has a prespecified direction. A process to correct this is now performed.

FIG. 1 shows an embodiment of a recognition device that implements the line figure recognition method according to the present invention, and is an example of a drum-type line figure recognition device. In Figure 1, drawing 1 to be recognized
is tightly wound around the drum, and drum 2 is
Direction Motor 3 rotates in the direction of the arrow. The displacement of the position of the drum 2 due to this rotation is detected by the Y-direction encoder 4, and its pulse signal is sent to the Y-direction up/down counter 5. The image input head 6 is composed of a lens 7, an illumination lamp 8, an optical sensor 9, and a pole nut 10 as shown in FIG. 2, and is moved left and right by a ball screw 13 directly connected to the X-direction motor 11 shown in FIG. A cursor 12 is fixed to and interlocks with the image input head 6 in order to designate any position on the drawing. Also, the image input head 6
is equipped with a linear scale 14 in order to detect displacement in the position in the X direction, and a pulse generator 15 directly connected to the image input head 6 sends pulses to an up/down counter 16 in the X direction every time the head position changes by a predetermined amount. It will be done. The X-direction motor 11 and the Y-direction motor 3 in FIG.
and a Y-direction motor drive circuit 19.

The control command and status of control circuit 17 are reflected by command status register 20. The command status register 20, the Y direction up/down counter 5, and the X direction up/down counter 16 are connected to the selector 22.
and is connected to the data bus 23. The microprocessor 24 can access a program memory 25 , a data memory 26 , a selector 22 , a correction data memory 27 , and an image memory 28 via a data bus 23 .

Microprocessor 24 decodes and executes instructions stored in program memory 25.

On the other hand, the motor control keyboard 21 is connected to the control circuit 17 to give the operator control to move the cursor 12 to any position on the drum 2, and includes an X-direction motor drive circuit 18 and a Y-direction motor drive circuit 19.
control offline. At the same time, a command is issued to read the current cursor position coordinate values from the values of the X-direction up-down counter 16 and the Y-direction up-down counter 5, and to transfer the values to the correction data storage memory 27. Command status register 2
Give to 0.

Now, if we consider the case where the drawing 1 is wound around the drum 2 by hand, a rotational positional deviation d occurs. The rotational positional deviation d is large in proportion to the size of the drawing; for example, assuming that there is an inclination θ of 1° with respect to the drum reference axis in the longitudinal direction of the AO plate, d=21 degrees. In addition to the rotational positional deviation that occurs when the drawing is mounted, there is also distortion due to expansion and contraction of the drawing itself. Therefore, if you decide to use a Mylar sheet as the drawing paper, there will be little local expansion and contraction of the drawing, and if you wrap the drawing tightly around the drum, the grid coordinates printed on the drawing with high precision and the drum reference coordinates will be Since the deviation has the characteristic that it hardly changes no matter where on the drum it is measured in the drawing, it can be corrected by uniform processing.

FIG. 3 shows the flow of coordinate correction processing according to the present invention. First, the operator points the center of the cursor 12 in FIG.
Using the motor control keyboard 21, move to the positions of the reference points PG, PI, and PI (FIG. 4) in Drawing 10, and calculate the coordinate position of each point from the up/down counter 16 in the X direction and the 7 up/down counter 5 in the Y direction. Correction data storage memory 27
to be memorized. Next, the operator calculates the distance L between the three reference points.
1=POP1 and L2=POP2 are read from the drawing, the data is input from the motor control keyboard 21, and transferred to the correction data storage memory 27. From this state, the area surrounded by the reference point (POPIP3P2P in Figure 4)
O) is divided into small screens (for example, a screen divided into 256 x 256 pixels), and the three processes of capturing the image within the divided screen, line recognition within the divided screen, and connecting the divided screens are performed on the entire screen. Repeat until completed. This method of dividing the large screen into small screen units and recognizing lines in arbitrary directions is described in Japanese Patent Application No. 53-59933. It is explained in detail in Japanese Utility Model Application No. 53-62601. After the recognition of all the line paths in the large screen is completed, the rotational positional deviation correction process, the gridded coordinate correction process, and the oriented coordinate correction process of the recognized coordinates for the feature points are performed. These three processes will be explained in detail below.

Rotational position deviation correction processing First, the rotational position deviation correction processing of recognized coordinates will be explained.

Figure 4 shows the drum reference coordinate system (X, Y) and the drawing coordinate system (X'
, Y'). In FIG. 4, the reference points on the drawing are PO9pu and pz, the distance between the three points POP1 is Ll, and the distance POP2 is L2. The purpose of the coordinate conversion process is to convert the drum reference coordinate values (x, y) of a point P in the drawing in FIG. 4 into drawing coordinate values (X/, y/). Assuming that the values in the correction data storage memory 27 in FIG. The difference between the values is expressed as follows:
= PIX-POXy1 = Ply-POyX
2: p2x-POX Y2=P2F-POy, then the rotational position of the recognized coordinates (9) is replaced by the following equation. Misalignment correction processing is performed.

Gridded coordinate correction processing Rotation position shift The coordinate values obtained by coordinate correction include rounding errors and errors due to line width, and there is a possibility that the lines are recognized to take positions other than the grid. be. In gridded coordinate correction processing, the intersections of grids are called grids, and processing is performed to quantize coordinate values to the nearest grid position. That is, in FIG. 5, the grid positions of the interval g are (Ql).

G2. G3. G4), the square area centered on the grid and the width g-i (i is the dead band width, 0 and i<g) is defined as follows:
D13 Surrounded by D14.

Z2. Point D21D2 in the area centered on H grid G2
2 D23 Surrounded by D24.

(10) Point 1) atD3 in the area centered on z3 nigerid G3
2]) surrounded by sa D34.

Point D411) 4 in the area centered on Z4 Nigerid G4
2]) Surrounded by 43D44.

This area is called a grid sensitive zone, and processing is performed to move all points belonging to the grid sensitive zone to respective grid positions. This process is carried out by the calculation (where [ ] is a Gauss symbol), where the coordinate values of a point within the grid image are (x, y), and the coordinate values after the grid position shift are (y', y'). It will be done. Next, there is a point D12 in the rectangular area sandwiched between Za:Zl and Z2 in Fig. 5, which is a point that does not belong to the grid sensitive zone.
It is surrounded by D21 D24 D13.

Z7 2 Z41 Point (1) in the rectangular area sandwiched by Zl
1) Surrounded by D42 D311) at D4B.

For points belonging to , only the X coordinate is moved to the grid position. That is, the points belonging to Za are moved to the line connecting Gl and G2 in FIG. The points belonging to Z7 are moved onto the line connecting Q 4 and Q B in FIG. 5. In addition, in FIG. 5, the point that does not belong to the grid sensitive zone is a rectangular area sandwiched between Za:Z21 and Za, and is surrounded by points D24, D23, D32, and D31.

Za: Point D14 in the rectangular area sandwiched between Zl and Z4
It is surrounded by D13 D42 D41.

For points belonging to , processing is performed to move only the X coordinate to the grid position. That is, the points belonging to G2.26 in FIG. Move to the line connecting Qai. The points belonging to Za are moved onto the line connecting Gl- and G4 in FIG.

Oriented Coordinate Correction Processing In drawings where the direction of the line is only allowed in a specific direction (horizontal, vertical, ±45 degree direction), the coordinates do not simply exist on the grid, but all the lines (12) are in a specific direction. It is necessary to have The lines after gridded coordinate processing do not necessarily have a specific direction.

In the oriented coordinate correction process, two processes are performed in stages: oriented coordinate correction process for points outside the grid sensitivity zone and oriented coordinate correction process for points on the grid. The general flow of processing is as follows:
As shown in FIG. 9, a counter for counting the number of coordinates that are subject to oriented coordinate correction processing for points outside the grid sensitivity zone is used as a counter for counting the number of coordinates that are subject to oriented coordinate correction processing for points on the grid. A counter f for counting the number of coordinates that have become
Assuming CF10, the individual processes are repeated until CFO=0 and CFI O=O, and are further repeated until there is no mutual influence between the two processes, that is, until CFO+CF10=0. Programs for these processes are stored in the program memory 25 in FIG.
4 is executed.

The orientation coordinate correction process for points outside the grid sensitivity zone will be explained. By the grid coordinate correction process, points that have been moved to the grid position are regarded as points that are at a more reliable position, and points that are not on the grid are brought into the grid (13). For example, in Figures 6 and 7, we focus on a straight line made up of two points, and if the angle between the direction of the straight line and the specified direction is less than a certain value,
If one point is on the grid or on a line connecting the grids in that direction, A1 is on the grid, or B is a point that is not on the grid in that direction, then point B is moved to point 0 on the grid. Perform processing.

FIGS. 10 and 11 show a detailed flow of the orientation grid processing for points outside the grid sensitive zone. Further, FIG. 12 shows the contents of a registration table of lines to be processed, which is stored in the data memory shown in FIG. 126.

In Fig. 12, LNO is the registration number of the line, Nl is the number of constituent points of one line, Pl (i=1 to Nl) is the coordinate of the point, and Ll is the index showing the coordinate position of the LNO-th line. . For example, it shows that Ll corresponds to the position of point P32 on the LNO+2nd line in FIG. 12. Let P(Ll) be the L1-th point on the line, P (L1+1) be the L1+1-th point, and P(LX-1) be the Ll-1-th point. Checks whether the line composed of the two points P(LL) and (14) P(L1+1) has one of the specified directions (horizontal, vertical, ±45 degree direction). . If it has either direction, processing is moved to the next point. line PCLL
) If P(LL+1) does not have the specified direction, check whether the point P(Ll) is within the dead zone. If point P(Ll) does not exist within the dead zone, processing is moved to the next point. Point P (LL+1)
The number of candidates that the point can move onto the grid is CN1, and the point P(
The number of candidates that the point Ll-1) can move onto the grid is CN
2 and initialize CN1=O and CN2=0. Also, CN3 is the counter for the direction in which point P (Ll) can move onto the grid, Ml is the number of directions in which point P (L1+1) can be moved onto the grid, CN4 is the counter for the direction in which point P (L1+1) can be moved onto the grid, and CN4 is the number of directions in which point P (L1) can be moved onto the grid. M2, point P
Let CN5 be the counter for the direction in which (Ll-1) can move onto the grid, and M3 be the number of directions in which it can move. First, point P
(Ll) is moved in one of the movable directions (15). Check whether the point P(L1+1) exists on the grid, and if it exists, m P (Ll ) P (L1+ 1 )
Check whether the direction force is in the specified direction. If the specified direction is determined, this direction is set as a candidate that can be moved to the grid, and the counter CNI is incremented by 1.

If it is not in the specified direction, move the point P(L1+1) in one of the movable directions and move the line P(Ll)P(L1
+1) Check whether the direction is the specified direction. If the specified direction is determined, this direction is set as a candidate that can be transferred to the grid, and the counter CNI is incremented by 1. If the direction is not the specified direction, if the counter CN4 for the number of directions in which the point P (L1 + 1) can move is less than or equal to M2, the point P (L1
+1) in the next movable direction and repeat the above process until CN4 becomes M2 or more. Next, point P(Ll−
Similar processing is performed for 1). That is, point P(L
l-1) in one of the movable directions, and check whether the direction of line P(Ll)P(LX-13) forms the specified direction.If it forms the specified direction, move this direction. (16) Set it as a candidate that can be moved to the grid, and increase the counter CN2 by 1. If it is not in the specified direction, point P(Ll-
If the counter CN5 for the number of movable directions in 1) is less than or equal to M3, move the point P (Ll-1) to the next movable direction, and
The above process is repeated until N5 becomes equal to or greater than M3. Above 3
Points P(Ll), P(L1+1).

The line P(Ll) P(L1+1) composed of P(Ll-4)
) and the line P (Ll-1) P (Ll ) are uniquely determined, that is, when CNI and CN2 both have a value of 1, move the point P (Ll) onto the grid that is a candidate for migration. Perform migration processing and increase CFO by 1.

If it is not uniquely determined, the process of moving to another point on the line specified by LNO is continued until the value of Ll reaches the number of constituent points N. Further, the above series of processing is continued until the LNOO value reaches the end value of all registered lines.

The orientation coordinate correction process for points on the grid will be described. The processing policy focuses on partial lines consisting of two points on each line, preferentially fixes the line whose direction has the specified direction, and fixes the specified direction (17). The line with the direction is rotated in the direction that is considered to have the least amount of rotation. FIG. 13a shows the contents of a table prepared in advance for specifying the direction of rotation. Further, FIG. 13 shows the contents of a table for specifying the grid position in which direction a point should take in order to realize a specified rotation direction.

The numbers ``■'' to ``■'' in FIG. 13a correspond to the numbers ``■'' to ``■'' in FIG. 13. Regarding the direction of rotation of the line and the direction of transition of the grid position, there are the following cases.

1. Lines belonging to area ]) o ]) s rotate the line in the direction of 8084, points belonging to area 1) o move to the grid in the direction of ■, points belonging to area D8 move to the grid in the direction of ■ Move to.

2. The line belonging to the area DID9 is ll1if, S I
By rotating in the direction Ss, points belonging to the region D1 are moved to the grid in the direction of ■, and points belonging to the region D9 are moved to the grid in the direction of ■.

3. The line belonging to the area D2DIG rotates the line in the direction of 8185, and the point belonging to the area D2 rotates the line in the direction of (18
) The points belonging to the area D10 are moved to the grid in the direction of ■.

4. Lines belonging to area 1) al) 1t are lines 8286
The points belonging to the region D3 are moved to the grid in the direction of ■, and the points belonging to the region D11 are moved to the grid in the direction of ■.

5. The line belonging to the area D4D12 rotates the line in the direction 8286, the point belonging to the area D4 is moved to the grid in the direction of ■, and the point belonging to the area D12 is moved to the grid in the direction of ■.

6. Area]) sl) The line belonging to ta rotates the line in the direction of 53S7, the point belonging to area D5 moves to the grid in the direction of ■, and the point belonging to area D13 moves to the grid in the direction of ■. .

7, area]) 6 The line belonging to D14 is 838
7, the points belonging to the area D6 are moved to the grid in the direction of ■, and the points belonging to the area D14 are moved to the grid in the direction of ■.

8. The line belonging to the area D 7 DIS is the line 54SO
Rotate in the direction of , and the points belonging to area D7 are rotated in the direction of (
19) Move to the grid, and move the points belonging to the area ])ts to the grid in the direction of ■.

On the other hand, FIGS. 14 and 15 show a detailed flow of the orientation coordinate correction process for points on the grid. If CFIO is a counter indicating the location to be processed, CFIO is first cleared. If LNO is a counter indicating the position where the line to be processed is registered, then the number N of constituent points of the specified line is obtained by LNO=LNO+1. Then, let Ll be a counter indicating the position of interest on the line, P(Ll) be the L1th point on the line, P (L1+11 be the L1+1st point,
Let P(Ll-1) be the Ll-1th point.

First, let's focus on a line composed of two points P(Ll) and P(L1+1). Two points P (Ll) and P (L1+1)
Check if both exist on the grid. If even one of them does not exist on the grid, move to the next processing position. 2 points P (Ll).

If P(L1+1) both exist on the grid, it is checked whether the lines P(L1) and P(L1+1) have the specified direction. has the specified direction (
20) If there is one, it will not be processed and will move on to the next processing position.

If line P(Ll)P(L1+1) does not have a specified direction, line P(LL-1) P(Ll) and line P(
Check the direction of L1+1)P(L1+2). If any of the lines has a specified direction, a fixed flag is set that means that the point P (Ll) or the point P (L1+1) is not moved to another grid position. Next, the flags that have just been set are checked, and if both point P (Ll) and point P (L1+1) have fixed flags, they are not to be processed, and the process moves to the next processing position. If only point P(Ll) has a fixed flag, check the rotation direction of the line as shown in Figure 13, and select point P(Ll).
Candidate P'(L1+1) of the grid position to be taken by
). The uniqueness determination process shown in FIG. 16 is used to determine whether the current candidate P'(L1+1') can be uniquely determined. If it can be uniquely determined, the point P(L1+1) is '(L1+1) position. Counter CF that counts the positions to be processed at the same time
Increase 10 by 1. If only point P (Ll+1) has a fixed flag (21), check the rotation direction of the line as shown in Figure 13,
Candidate P'( of the grid position that point P (Ll) should take)
LX) is obtained. A uniqueness determination process as shown in FIG. 17 is performed to determine whether the current candidate P'(Ll) can be uniquely determined. If it can be uniquely determined, the point P(Ll) is changed to P' (Ll) The counter CF1'Oei for counting the number of positions to be processed simultaneously is increased. If the fixed flag is not attached to both point P (Ll) and point P (L1+1), check the rotation direction of the line as shown in Fig. 13 and set the point P (LL) and point P (L1+
1) Candidate grid position P'(Ll) that should be taken by
Obtain P'(L1+1). P'(Ll) is now a candidate
and P'(L1+1) can be determined uniquely by the uniqueness determination process as shown in FIG. 17, and if both points can be uniquely determined, the point P
(Ll) is moved to the P'(Ll) position, point P (L1+1) is moved to the P'(LL+1) position, and a counter CPIO that counts the positions to be processed is increased by 2. If it is not uniquely determined, the process of moving to other points on the line specified by LNO is continued until the LlO value reaches the number of constituent points N (22). Further, the above series of processing is continued until the LNO value reaches the end value of all registered lines.

Figure 16 shows P' (L1+1) in Figure 14.
FIG. 3 is a diagram showing a detailed processing flow of uniqueness determination processing. The flag to indicate the determination of uniqueness is FLOKl, and when F'LOKI=1, there is uniqueness, and FLOK1=O
There is no uniqueness when . Initially, FLOKI = 0, and the line P' (Ll +1) P (L 1+2)
Checks whether or not follows the specified direction. If it is in the specified direction, it is assumed that it is unique and PLOKI is set to 1.

If they do not form the specified direction, find the next point P(L1+3) and check whether the lines P(L1+2) P(L1+3) form the specified direction. If it is in the specified direction, it is considered unique and F'LOKI = 1. If the specified direction is not formed, line P(L1+23 P(L1+3)
Check the direction of , and find the transition position of the point where the line is in the specified direction.
'(L1+2). Next, line P (L1+1)
Check the direction of P'(L1+2) and set the transition position (23) of the point where the line is in the specified direction as P''(L1+1).Here, the transition position P'(Ll-1-1) found earlier and If they are the same, it is determined that there is uniqueness and FLOK1=1.

However, if the point P(LL+2) or P(L1+3) does not exist, this process is interrupted at that point.

FIG. 17 is a diagram showing a detailed process flow of the uniqueness determination process of P'(Ll) in FIG. 15.

The flag for indicating uniqueness determination is FLOK2, and F
When LOK2=1, there is uniqueness, and when FLOK2=0, there is no uniqueness. First, set FLOK2=0, and line P'
Check whether (Lx) P (r, 1-1) forms the specified direction. If it is in the specified direction, it is considered unique and FLOK2=1. If the specified direction is difficult to follow, find the previous point P (LL-2) and draw the line P (Ll
-1) Check whether PCLl-2) forms the specified direction. If it is in the specified direction, it is considered unique, and FLO
Let K2=1. If it does not form the specified direction, line P (
Ll -1) Check the direction of P (Ll-2), and set the transition position of the point where the line becomes the designated direction as P'(Ll-1).

Next, examine the direction of the line P (Ll) P' (LL-1), and set the transition position of the point where the line becomes the finger (24) constant direction as P" (Ll). Here, the transition position found earlier is Check the identity with P'(Ll), and if they are the same, it is assumed that there is uniqueness and FLOK2 = 1.However, if point P(Ll-1) or point P(Ll-2) does not exist, its Interrupt this process at this point.

As described above, according to the present invention, by adding correction processing to the coordinate values after line recognition, rotational position shift correction of large drawings, gridded coordinate correction processing with specified accuracy, and oriented coordinates of lines are performed. Correction processing can be performed at high speed.

[Brief explanation of the drawing]

Fig. 1 is a block diagram of an embodiment of a device realizing the line figure recognition method according to the present invention, Fig. 2 is a specific block diagram of a part of Fig. 1, and Fig. 3 is the overall flow of coordinate correction processing. A flowchart showing
Figure 4 is a diagram showing the principle of positional deviation rotation correction to the drawing coordinate system, Figure 5 is a diagram showing the principle of gridded coordinate correction processing, and Figures 6 to 8 are diagrams showing the principle of line orientation coordinate correction processing. FIG. 9 is a diagram showing the flow of line orientation coordinate correction processing, and FIG.
C25) Figure 11 is a diagram showing the flow of directional grid processing for points inside the dead zone in Figure 9, Figure 12 is a diagram showing the contents of line registration, and Figure 13 is a diagram showing the direction of the line and how it should be moved. FIGS. 14 to 17 are diagrams showing the direction of the grid, and diagrams showing the flow of oriented coordinate correction processing for points on the grid. 1...Drawing, 2...Drum, 6...Image input head, (26) J p -^ To C= Figure 1 12 1G "f, /4 Figure 7Q IJ" (Otsushi) 4 (b) ■ ■ ■ ■・※・■ ■ ■ ■ (The collar of 7゛jiyh is the same) ■ Gu-o! Figure 16" fJ17

Claims (1)

[Claims] 1. In a pattern recognition device installed in a scanning device that recognizes a line figure written on a drawing, the drawing is scanned and the line figure data represented by the coordinates on the scanning device is obtained. Obtain the coordinate data of the feature points of the line figure expressed in coordinates on the scanning device based on the line figure data, input the coordinates of a plurality of reference points on the drawing on the scanning device, and Using the coordinate values,
A first correction is performed to convert the coordinate data into coordinates on the drawing, and then a second correction is performed to position the coordinate data at grid positions with a specific interval, and a second correction is performed to convert the coordinate data into coordinates on the drawing. Add a third correction to make it have direction,
A line figure recognition method characterized by outputting corrected coordinate data as a recognition result. 2. In the line recognition method described in claim 1, the distance between the reference points is input in addition to the coordinate values of the reference points, and the line recognition method is performed based on the coordinate values of the reference points and the distance. A line figure recognition method characterized in that, at the time of the correction in step 1, correction for expansion and contraction in the dimensions of the drawing is also performed. 3. In the line figure recognition method described in claim 1, the second correction process sets a minute area centered on the grid center position, and points existing within the area are moved to the grid position; The point belonging to the area sandwiched between the minute areas is X
A line figure recognition method characterized in that only coordinates or X coordinates are transferred onto lines connecting grids. 4. The third correction process in which the line after recognition in the line figure recognition method described in claim 1 has a specific direction focuses on each line composed of two points of the line figure. In this case, if both points are not on the grid and they are placed so that the grid position is sandwiched in the center, both points are moved to the grid position, and if one point is on the grid and the other point is not on the grid. If both points are on the grid and the line adjacent to one point is in the specified direction, that point will not be migrated preferentially and the other points will be shifted. A line figure recognition method characterized in that the line figure is shifted only when the direction can be uniquely determined.