JPS62208181A - Graphic extracting system - Google Patents

Abstract

PURPOSE:To extract the whole closed area of linear graphic data or its vector data and to improve the extracting accuracy by detecting feature angles of a graphic in a closed area to be extracted and processing only a restricted range based on the detected angles. CONSTITUTION:An image digitized by an image input part 3 in a system is stored in an image memory part 4 and rester-scanned by an image scanning part 5 to extract a border point existing between an liner line and the background and the extracted border point is applied to a control part 16. The image obtained from the memory part 4 is processed by clockwise and counter- clockwise rotation part border tracing parts 6, 7 and a counterclockwise rotation closed borded tracking part 9 rspectively and the sequence of points on the border between the graphic part and the background part is stored in a point sequence memory 14. Under control by the control part 16, the feature angles of the graphic in the closed area are extracted by an angular calculating part 8, a feature calculating part 10, an end part detecting part 11, and so on. An image included only within the restricted range is processed, so that the image of the whole closed area can be highly accurately extracted and outputted from an image output part 15.

Description

DETAILED DESCRIPTION OF THE INVENTION (Technical field to which the invention pertains) The present invention relates to a method for automatically extracting figures forming closed regions from line drawings such as maps and drawings.

(Prior art) Conventionally, in a method of inputting a map or drawing as an image into a computer and extracting a figure forming a closed area (for example, a building on the map) from the image by boundary tracing, conditions for the starting point of boundary tracing are used. Since a condition that holds true for many boundary points is adopted, there is a problem that 12,010 outputs are obtained and calculation time is required.

Furthermore, it was not possible to extract figures that were cut off by other figures such as letters or symbols, or by noise, resulting in figures forming incompletely closed regions.

(Object of the Invention) In order to solve these drawbacks, the present invention aims to limit the area from which figures forming a closed area are extracted by boundary tracking, and to extract figures forming an incompletely closed area. do.

In order to achieve such an object, the present invention first extracts a characteristic corner of a figure to be extracted, and then extracts a figure forming a closed region by boundary tracking using the corner as a starting point for tracking. be.

In addition, if the shape features of the extracted figure do not match the features of the desired figure, boundary tracing is first performed from the corner until the end of the line is detected to extract a partial boundary, and then a partial boundary is extracted from the corner. Extract partial boundaries by performing boundary tracing in the opposite direction of tracing until the end of the line is detected, and then if only one end of the line is detected, the end of the line is not detected. The partial boundary is traced again, a boundary point within a certain distance from the edge of the detected line is detected and set as the edge of the line, and when the edges of the line are extracted for both circumferences, the This method extracts the boundary of an incompletely closed region that passes through a corner and reaches the end of the other line.

(Description of structure and operation of the invention) FIG. 1 is a diagram for explaining the system principle of the present invention, where 1 is a line diagram and 2 is a background.

In the present invention, extraction of figures constituting one closed region is realized by executing the following nine steps in FIG.

[Step 11 Boundary point on the line figure (ih
Extract e jh).

[Step 21: Track five boundary points clockwise from the boundary point (ik, jk). The last tracked boundary point clockwise is b++j
+).

Next, trace s-1 boundary points from (ihe j++) counterclockwise. Let the boundary point last tracked counterclockwise be (inn Jn).

These boundary tracking methods are already well known (translated by Nagao, Digital Image Processing, Kindai Kagakusha, pp. 3).
53-359).

The difference between the clockwise boundary tracing section and the counterclockwise boundary tracing section is
The tracking point detection method shown in FIGS. 2 and 3 is used.

FIG. 2 is a diagram illustrating clockwise boundary tracking.

As shown in Figure 2, in one clockwise tracking, the 8 neighborhoods of the boundary point of interest are scanned clockwise from the position of the previously tracked boundary point, and the pixels that change from the background part to the figure part are scanned next. Let this be the boundary point of interest.

FIG. 3 is a diagram illustrating counterclockwise boundary tracking.

On the other hand, as shown in Fig. 3, in counterclockwise tracking, the 8 neighborhoods of the boundary point of interest are scanned counterclockwise from the position of the previously tracked boundary point, changing from the background part to the figure part. Let a pixel be a first-order boundary point of interest.

[Step 31 Partial boundary B, which is a set of boundary points obtained in step 2
= (11+ N +...'r L-ty j++-+
L (tbr jh),...p (”nt jn)
) forms a characteristic angle.

That is, first, two boundary points (IIIJI) and Dk-
11jk-t) can be approximated to a straight line.

This means that there are two boundary points (i++jt) and (!+a−+
The leg length of the perpendicular to the straight line passing through the boundary point (lr+ Jj to the two boundary points b++j□) between +jh-+) and (jb-+ t jk-+) is smaller than the given threshold. This can be determined by inspecting.

Then, the two boundary points (lh * jk) and (lny j
Check whether the boundary between n) can be approximated to a straight line. And if both can be approximated to a straight line, the boundary point (L+ j+)
Vector 1 connecting the boundary point (lk-1+ jh-+) from
and from the boundary point (ihe jk) to the boundary point (inn ja
) Calculate the angle formed by Buck 1 and Buck 2. Otherwise, the process ends.

For the angle, calculate υ using the formula, and add (1+++
, 'l++), then 0 is the sum of 180° added to 0.

However, the old "In lh+ V1=, :+n, :1
++12=1. -, lI+V2=, :1+-+
l ++.

If the angle 0 is not within a predetermined range, the process ends.

[Step 41: Extract the closed field B1 including the partial boundary 13. Closedness field B1
In the case shown in Fig. 1(d)-1, and in the case shown in Fig. 1(d)-1,
There are cases like 2.

[Step 51: Based on the characteristics of the closedness field B1 that can be calculated from the boundary point sequence, such as 2-area, boundary angle, circularity, etc., determine whether the figures forming the closedness field B1 are figures forming a closed region to be extracted. .

When the feature satisfies the given condition, the figure is determined to be a figure to be extracted, and the process is terminated.

[Step 6] Extract partial boundaries forming an incompletely closed region as shown in FIG. 1(d)-2.

First, the point sequence from boundary point (1 + jJ to B1) is traced counterclockwise, and tracing is terminated when the end of the line is detected or the number of traced boundary points exceeds a threshold value.

The obtained partial boundary is designated as 82 as shown in FIG. 1(e).

Any method may be used to determine the end of the line.

For example, two boundary points (lr-tt, jr-+L (lr*+t Jr, +) that are located before and after the boundary point (lr+ Jr) to be determined in the tracked boundary point sequence.
, and from the boundary point (lr-++ jt-j) to the boundary point (
The boundary point (l
r
When the distance between r−++ jr−+) and (lr*It jr−+) is less than or equal to the threshold, the boundary point to be determined (ir
+ Jr) as the end point of the line.

Next, the point sequence of B1 is traced clockwise from the boundary point (lke ++a), and the tracing is terminated when the end of the line is detected or the number of traced boundary points exceeds a threshold value. The obtained partial boundary is designated as 83 as shown in FIG. 1(e).

[Step 7] In step 6, if only one edge of the line is detected as shown in Figure 1(f), trace the sequence of points on the partial boundary where no edge of the line has been detected, and The boundary point whose distance from the edge of is within the threshold value is defined as the edge of the line.

[Step 81 If two ends of the line are detected in steps 6 and 7, interpolate between them with a straight line to
)-2, create a closedness field B4.

In other cases, the process ends.

[Step 9] Based on the characteristics of the closedness field B4 that can be calculated from the boundary point sequence, it is determined whether B4 is a figure forming a closed region to be extracted.

FIG. 4 is a diagram showing the configuration of an embodiment according to the present invention, in which 3 is an image input section, 4 is an image memory section, 5 is an image scanning section, 6 is a clockwise partial boundary tracking section, and 7 is a counterclockwise section. 8 is an angle calculation unit.

9 is a counterclockwise closed field tracking unit, 10 is a feature calculation unit, 11
12 is a narrow part detection part, 13 is a linear interpolation part,
14 is a point sequence memory section, 15 is an image output section, and 16 is a control section.

The image input section (3) is a section that samples and quantizes images using a facsimile, drum scanner, TV camera, etc., converts them into digital images, and inputs the images.

The image memory section (4) is a section that stores digital images provided from the image input section (3), and is composed of a random access memory and the like.

The image scanning section (5) scans the digital image stored in the one-image memory section (4), executes procedure 1, and outputs the result to the control section (16).

The clockwise partial boundary tracing unit (6) and the counterclockwise partial boundary tracing unit (7) trace the boundary between the figure part and the background part of the image stored in the image memory unit (4) from a given tracing start point. It is converted into a point sequence and stored in the point sequence memory section (14).

Tracking ends when the number of tracked boundary points reaches both thresholds.

The angle calculation unit (8) calculates the angle as described in step 3 from the four boundary points obtained by counterclockwise tracking.

Counterclockwise closed boundary tracking unit (9) is one image memory unit (4)
The boundary between the figure part and the background part of the image in
4).

The tracking ends when a closed boundary is extracted.

The feature calculation unit (lO) calculates features such as area, boundary length, circularity, etc. from the point sequence of the closed boundary.

The end detecting section (11) reads out the point sequence from the point sequence memory section (14), traces the point sequence, and detects the end of the line shown in step 6.

The outputs of the end detecting section (11) are whether or not the end of the line was detected in B2, whether the end of the line was detected in B3, the dot sequence of B2, and the dot sequence of B3.

The narrow portion detection section (12) reads out a point sequence from the point sequence memory section (14), tracks the point sequence, and detects that the distance from the edge of the already extracted line in the end detection section (11) is less than or equal to a threshold value. Tracking ends when a certain boundary point is detected.

The linear interpolation unit (13) reads out a point sequence that is not a loop from the point sequence memory unit (14), interpolates both ends of it with a straight line to obtain a looped point sequence, and stores it in the single point sequence memory unit (14). Store.

The point sequence memory section (14) is a section that stores a sequence of addresses of boundary points defined by pairs of X and X coordinates.

The image output unit (15) is a part that displays the dot sequence stored in the dot sequence memory unit (14), and is a printer, a display, or the like.

The control unit (16) controls the entire system so that a figure forming a closed area having a certain characteristic is extracted from the entire line figure data stored in the image memory unit (4) according to steps 1 to 9. It is a part. The control unit (16) controls and operates each part as follows.

First, a line figure is input to the image input section (3) and stored in the image memory section (4).Next, the line figure is input to the image scanning section (5).
The digital image stored in the image memory (4) is scanned by rask and step 1 is executed, and the boundary point (ik, jk
) is obtained, the rask scan is interrupted.

Next, the obtained boundary point bht Jb) is converted into the clockwise portion H
It is given to the 1-field tracking unit (6) and the counterclockwise partial boundary tracking unit (7) to execute step 2, and the obtained point sequence is stored in the point sequence memory unit (
14).

Next, the point sequence obtained in step 2 is stored in the point sequence memory section (14
) is read out and given to the angle calculation unit (8) to calculate the angle in step 3. If the angle is not within a predetermined range, the processing for the boundary point (ik, jk) is terminated, and the image scanning unit (5) is instructed to restart rask scanning of the digital image in order to obtain a new boundary point. .

If the angle is within a predetermined range, proceed to the next process.

The point sequence obtained in step 2 is read from the point sequence memory unit (]4) and given to the counterclockwise closed boundary tracking unit (9) to perform closed boundary extraction in step 4, and the resulting closed boundary point sequence B1 is Write to the point sequence memory section (14).

Next, the closed boundary point sequence B1 is read out from the point sequence memory unit (14) and given to the feature calculation unit (10) to calculate the area, boundary length, circularity, etc. in step 1, and the closed boundary point sequence B1 is
calculate the characteristics of

Next, if it can be determined from the obtained features that the figure forms a closed region to be extracted, the closed boundary point sequence B1 is registered in the point sequence memory unit (14) as the figure to be extracted, and the boundary point (l
ht Jk) and instructs the image scanning unit (5) to resume rask scanning of the digital image in order to obtain a new boundary point.

In other cases, proceed to the first processing.

Read the closed boundary B1 from the point sequence memory section (14), give it and the coordinates (ik, jk) to the edge detection section (11), extract the partial boundaries B2 and 83 according to step 6, and store it in the point sequence memory section (14). ).

Next, the end of the line is determined to be B2 or
If detected in only one of 3,
The narrow portion detection unit (12) is given information on whether or not the end of the line is detected in the point sequence B2, point sequence B3, and B2, and whether or not the end of the line is detected at +33, and is made to perform step 7. The data is written to the point sequence memory section (14).

Next, if the end of the line is not detected in either B2 or 83, the process for the boundary point (life jk) is finished and the image scanning unit (5) is sent to obtain a new Xt boundary point. Instructs to resume rask scanning of digital images. If the end of the line is detected at both B2 and 83, give the line interpolator (13) the point sequence B2 and 83 and perform step 8.
is executed, and the resulting point sequence B4 forming a loop is written into the point sequence memory section (14).

Next, read B4 from the point sequence memory section (14), ! 1% l
, to the feature calculation unit (10) to calculate the area, boundary length, circularity, etc. to obtain the features of the point sequence B4.

Next, if the point sequence B4 can be determined to be a figure forming a closed region to be extracted from the obtained features, that point sequence is registered in the point sequence memory unit (14) as a figure to be extracted, and the boundary point (" k
, Jk), and instructs the image scanning unit (5) to resume rask scanning of the digital image in order to obtain a new lrt boundary point.

When one rask scan of the digital image is completed, 9. The value of the pixel on the @ point sequence is 111 II
, the image output unit (1
5) and instruct it to be output as an image on paper, etc.

In addition, for data obtained by vectorizing line figure data,
The above procedure can be simplified as follows.

Step 1 is to extract the nodes (ik, jk) of the back 1-le.

Step 2 is to calculate one vector connected to *(lk, Jk) (referred to as vector 1) and the adjacent vector (referred to as vector 1) when viewed counterclockwise from vector 1 with (lb+jh) as the center. It will be extracted.

Since there are generally multiple pairs of vectors, Step 3 and subsequent steps are executed for all of them.

In step 3, the nodes in vector 1 that are not (il,, jh) are set as b++j+), the nodes that are not as (Lk+ jk) in vector 2 are set as (ln+ Jn), and (Lk-++ jh
-+) may be applied as the same point as (jk, jk).

Steps 4 to 9 can be left as they are. However, tracking of boundary points on an image becomes boundary node tracking on a vector. Also,
Detecting the end of a line involves extracting a node with only one connected vector.

(Effects of the Invention) As described above, according to the present invention, characteristic corners of a figure forming a closed region to be extracted are first detected, and only a limited range is processed from there. Without extracting the completely closed region of the vectorized data,
It is possible to extract figures that form closed regions that have certain characteristics. In addition, it is possible to extract figures that are incompletely closed regions due to noise or the like.

Therefore, there is an advantage that a figure forming a closed region having characteristic corners can be efficiently and accurately extracted.

[Brief explanation of drawings]

FIG. 1 is a diagram for explaining the present invention in detail, FIG. 2 is a diagram for explaining clockwise IIt boundary tracing, FIG. 3 is a diagram for explaining counterclockwise boundary tracing, and FIG. 4 is a diagram for explaining counterclockwise boundary tracing. FIG. 2 is a diagram showing the configuration of an example. 1... Line figure, 2... Background, 3... Image input section, 4... Image memory section, 5...
- Image scanning unit, 6... Clockwise partial boundary tracking unit, 7... Counterclockwise partial boundary tracking unit, 8... Angle calculation unit, 9... Counterclockwise closed boundary tracking unit, lO. . . . Feature calculation section; 11. . . Edge detection section. 12... Narrow portion detection section, 13... Linear interpolation section. 14... Point sequence memory section, 15... Image output section, 1
6...Control unit.

Claims (2)

[Claims] (1) In a method that automatically extracts figures constituting a closed region with straight line parts from binary line figure data or vectorized data, the characteristic corners of the figure are detected at the boundary lines, and the A figure extraction method that uses a corner as a starting point, performs boundary tracking from there, and extracts closed boundaries. (2) In a method that automatically extracts a figure in which a part of the figure constituting a closed region having a straight line part is missing due to noise etc. from binary line figure data or data obtained by converting it into a vector, Either detect a characteristic corner and detect the end of a straight line from the corner, or perform boundary tracking until the length of the tracked partial boundary exceeds a certain value to extract the partial boundary, and then extract the partial boundary from the corner. Detect the end of the straight line in the opposite direction from the first tracing, or perform boundary tracing until the length of the tracked partial boundary exceeds a certain value to extract the partial boundary, and then If only
When tracing a partial boundary whose end part has not been detected again and detecting a boundary point within a certain distance from the detected end part and using it as the end part of the line, or if there is a length of the traced partial boundary. A figure extraction method characterized in that tracking is terminated when a value exceeds a value, and when two end portions are detected in the above processing, a partial boundary that connects the two end portions and passes through the corner is extracted.