JPH022189B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION Field of Industrial Application The present invention relates to a stroke extraction device for extracting strokes as line segment elements representing the shape characteristics of a line figure from a line figure when recognizing or understanding figures or drawings. It is something.

Conventional structure and problems Conventionally, to input information on line figures (logic circuit drawings, diagrams, etc.) to a computer, the coordinate information of the line figure was coded and input using a card.The line figure was input using a digitizer, tablet, etc. There are methods such as specifying and inputting coordinates, inputting by drawing a line figure on a graphic display using a keyboard or tablet, and reading the line figure as image data to obtain information through pattern recognition. This input method can be expected to have great effects, such as being able to be automated using a computer, reducing the number of processing steps, and reducing errors caused by manual intervention. Several devices have been devised for this purpose.

However, in conventional methods, in order to extract the constituent elements of strokes and line figures, a tracking method is used on all the data of the original image, so a large capacity storage is required to store all the data of the original image. section is required. Furthermore, thinning processing may be necessary as preprocessing for the tracking method, and in this case, processing time is long. Furthermore, in order to perform extraction reliably and efficiently, it is necessary to draw straight lines using a ruler,
This method has the disadvantage that there are significant restrictions on drawing creation, such as the need to draw the shapes of drawing components using templates.

Purpose of the invention The present invention solves the above-mentioned conventional problems.
Even without having a memory unit to store all the original image data, it is possible to extract the horizontal strokes, vertical strokes, diagonal strokes that make up a line figure, and their connection relationships, and there is no need for preprocessing such as line thinning processing, and the target It is an object of the present invention to provide a stroke extracting device for a line figure that does not require large restrictions on the figure.

Structure of the Invention The present invention provides a reading means for sequentially binarizing observed values while observing a two-dimensional line figure expressed on a two-dimensional plane by line scanning using a line scanner, and A process of converting binary data representing white pixels and black pixels into coordinate value data of the starting point and end point on the two-dimensional plane for each set of consecutive black pixels on one scanning line, and scan list generation means that performs the process of counting the number of consecutive black pixel sets; and coordinate value data generated by the scan list generation means and the number of consecutive black pixel sets, starting from the latest one. Based on a scan list storage unit that stores two scan lines, and the data stored in the scan list storage unit, the length of a set of consecutive black pixels on one scan line on two adjacent scan lines is determined. , pattern processing calling means which examines the mutual connection relationship between two adjacent scanning lines of the continuous black pixels and distributes processing for each connection pattern set in advance; a pattern processing means for performing pattern processing by generating strokes, which are line portions constituting a figure, and nodes representing connection portions between the strokes, from the pattern;
The coordinate value data of the start point and end point of the set of consecutive black pixels on one scanning line and the number of the set of consecutive black pixels existing on one scanning line are updated for one scanning line, and the data is updated for each scanning line. By repeating the process of distributing the processing to the pattern processing and pattern processing, three types of strokes, horizontal, vertical, and diagonal, and their connection relationships are extracted as the line parts that make up the figure.

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

FIG. 1 is a block diagram of a line figure stroke extraction device according to an embodiment of the present invention.

In the figure, reference numeral 1 denotes a raster scanning type reading unit such as a facsimile, which performs binarization processing on a two-dimensional line figure and outputs binarized data representing white pixels and black pixels. 2 is a scan list generation unit, which converts the start point and end point of each black run into coordinate value data based on the output information of the reading unit 1, counts the number of black runs on one scan line, and selects horizontal, vertical, and diagonal strokes. The black runs that make up the area are identified. 3 is a scan list generation unit, which receives coordinate value data of the start point and end point of the black run sent from the scan list generation unit 2;
The type of black run and the number of black runs on one scanning line are always stored for two scanning lines starting from the newest one. Reference numeral 4 denotes a pattern processing calling section which examines the connection relationship of the black runs on the upper and lower two scanning lines, classifies which pattern it belongs to among preset patterns, and calls the corresponding pattern processing section 5. A list processing section 6 performs basic stroke extraction processing in response to a call from the pattern processing section 5. 7a, 7b, and 7c are a stroke list storage section, a node list storage section, and a run list storage section, respectively, which are successively updated to the latest data according to the processing results of the list processing section 6, and finally connect strokes to each other. Remember relationships. Reference numeral 8 denotes an output unit that outputs the stroke extraction result of a line figure.

The operation of the apparatus configured as described above will be explained below.

First, a target two-dimensional line figure is observed by a raster scanning type reading unit 1 such as a facsimile, and subjected to a binarization process to become binarized data representing white pixels and black pixels. This binarized data is sequentially input to the scan list generation unit 12, and for each black run, 2
It is converted into coordinate value data of the start point and end point on the dimensional plane and output.

In addition, the scan list generation unit 2 determines whether a black run constitutes a horizontal stroke (hereinafter referred to as a horizontal run), or a black run constitutes a vertical stroke or a diagonal stroke (hereinafter referred to as a non-horizontal run), depending on the length of the black run. orchid). That is, as shown in FIG. 2, if the representative line width of the two-dimensional line figure is W and the inclination of the stroke 21 from the horizontal direction is θ, then the run length L is as follows: L=(1/sinθ)W. Therefore, if the maximum θ that is regarded as a horizontal stroke is θ _H , then a black run where the run length L is L≧(1/sin θ _H )W is defined as a horizontal run. By selecting θ _H to be about 15°, for example, good results can be obtained for ordinary handwritten figures.

Then, the scan list generation unit 2 outputs the result of identifying whether the black run considered as a target is a horizontal run or a non-horizontal run, and the number of black runs existing on one scan line.

Next, the coordinate value data of the start point and end point of the black run output from the scan list generation unit 2, the type of black run, and the number of black runs on one scanning line are always stored in the scan list storage for two scanning lines from the newest one. Registered in 3. (Hereinafter, the contents of the scan list storage unit 3 will be simply referred to as a scan list, the scan list for the latest scan line will be referred to as the scan list for the second scan line, and the scan list for the previous scan line will be referred to as the scan list for the first scan line. ) Then, in the pattern processing calling unit 14, based on the scan list, the black run on the second scanning line,
The black runs on the first scanning line are made to correspond in order, and the connection relationship between the black runs on the upper and lower two scan lines is checked. It classifies which pattern it belongs to among the set patterns, and calls the corresponding pattern processing section 5.

The connection relationship of the above black runs is checked as follows based on the scan list.

That is, as shown in FIG. 3, the x coordinates of the starting point and ending point of the black run on the second scanning line are x _S,2 , x _E,2 , and
The x coordinates of the starting point and ending point of the black run on the first scanning line are x _S,1 ,
Assuming x _E,1 , if there is any one of the following relationships between them, it is determined that the black runs on the upper and lower two scanning lines are connected.

x _S,2 ≦x _S,1 ≦x _E,2 +1 x _S,2 −1≦x _E,1 ≦x E _,2 x _S,1 ＜x _S,2 ≦x _E,2 ＜x _E,1 Here, the coordinates are taken as follows. Let the main scanning direction of the line scanner be the x coordinate, and let the sub-scanning direction (perpendicular to the main scanning direction) be the y coordinate. It is assumed that the coordinate value increases by 1 for each pixel in the x direction and for each scanning line in the y direction.

There are many types of connection patterns that are set in advance, but the following seven types are the basic ones.

A: As shown in FIG. 4a, the second scanning line has a horizontal run that is not connected to the black run of the first scanning line.

B: As shown in FIG. 4b, the second scanning line has a non-horizontal run that is not connected to the black run of the first scanning line.

C: As shown in FIG. 4c, the horizontal run of the second scanning line and the horizontal run of the first scanning line are connected.

D: A case where the non-horizontal run of the second scanning line and the non-horizontal run of the first scanning line are connected as shown in FIG. 4d.

E: As shown in FIG. 4e, the horizontal run of the second scanning line and the non-horizontal run of the first scanning line are connected.

F: As shown in FIG. 4f, the non-horizontal run of the second scanning line and the horizontal run of the first scanning line are connected.

G: As shown in FIG. 4g, there is a non-horizontal run in the first scanning line and there is no black run to be connected to it in the second scanning line.

As described above, the connection relationships of black runs can be classified based on the scan list.

Now, the pattern processing section 5 called by the pattern processing calling section 4 calls a plurality of list processing sections 6, which will be described later, according to the processing procedure according to the pattern.
The stroke list storage section 7a, node list storage section 7b, and run list storage section 7c, which will be described later, are successively updated to the latest data according to the processing results of the list processing section 6, and finally the connection relationship between strokes is determined. The data is stored in these storage units 7a to 7c.

Here, the pattern processing section 5 and the pattern processing section 5
The basic process of extracting strokes performed by the list processing section 6 called from the list processing section 6 will be explained in relation to the above-mentioned storage section.

(b) When horizontal runs are connected as shown in Figure 5a, when scanning is performed sequentially, the scan list for the second scan line, which is the latest scan list of the scan list for the second scan line, is displayed. First, the i-th run appears. In this case, it is determined that the horizontal stroke starts from this black run, and the stroke list storage unit 7
The type of stroke, the y-coordinate of the i-th run, and the x-coordinates of the start point and end point are registered in a.

Next, when the scan list is updated by one line, the i-th run is added to the scan list for the first scan line.
The (i+1)th run appears in the scan list of scan lines. Here, the stroke list storage unit 7
The x-coordinates of the starting point and ending point of a, the x-coordinates of the starting point and ending point of the (i+1)th run, and the x-coordinates of the starting point and ending point of the (i+1)th run are compared, and the stroke list storage unit 7a is updated. That is, x _S =MiN (x _S , x _S , I ₊₁ ) is the starting point x coordinate, and x _E =MAX (x _E , x _E , I ₊₁ ) is the ending point x coordinate.

However, x _S , x _E : Stroke list storage section 17a
The x-coordinates of the start and end points of the horizontal stroke registered in .

x _S , I ₊₁ , x _E , I ₊₁ : x coordinates of the start and end points of the (i+1)th run.

Thereafter, the same process is repeated while sequentially updating the scan list, and when the (i+k)th run appears in the scan list on the first scan line, the horizontal stroke cannot be extracted because there is no subsequent black run on the second scan line. Complete. That is, as shown in FIG. 5b, a stroke 51 connecting the minimum starting point coordinate x _S and the maximum ending point coordinate x _E is detected.

(b) When non-horizontal runs are continuous as shown in FIG. 6a, when sequential scanning is performed, the i-th run first appears in the scan list of the second scanning line. At this time, it is determined that a vertical stroke or diagonal stroke has started from this black run, and the run list storage unit 7
The y-coordinate of the i-th run and the x-coordinates of the starting point and ending point are registered in c. After that, when the scan list is updated sequentially, the scan list for the second scan line includes the (i+1)th run, the (i+2)th run, etc.
appears, but each time the run list storage unit 17
The run information (for example, the coordinates of the start point and end point) is registered in c. When the (i+k)th run appears in the scan list of the first scanning line, it is determined that the vertical stroke or diagonal stroke has ended because there is no continuing black run on the second scanning line, and the memory is sequentially stored. Based on the information in the run list, it is determined whether the stroke is a vertical stroke, diagonal stroke, or noise, and the type of stroke and the coordinates of the starting point and ending point are registered in the stroke list storage section 7a. As a result, strokes are extracted as shown in FIG. 6b61.

Here, the discrimination between a vertical stroke and a diagonal stroke can be determined, for example, from the positional relationship between the i-th run and the (i+k)-th run in FIG. 6A, which constitute the stroke. That is, if the midpoint of the starting run (first black run) is (x _S , Y _S ) and the midpoint of the ending run (last black run) is (x _E , y _E ), then
Set a certain value θ _V , and if y _S −y _E /x _S −x _E ≧tanθ _V , it is a vertical stroke, otherwise it is a diagonal stroke. Here, tan θ _V can be set arbitrarily, but in the case of a normal line figure, good results can be obtained by setting θ _V = about 80°, for example.

(C) As shown in Figure 7a, when a horizontal run and a non-horizontal run are connected, the scan list is updated by repeating sequential scanning, and the i-th run is added to the scan list of the second scan line.
When the (i-1)th run appears in the scan list of the first scanning line, it is first determined that the horizontal stroke has ended in the (i-1)th run, and the stroke is updated in the stroke list storage section 7a. finish. Then, it is determined that a new vertical stroke or diagonal stroke has appeared from the i-th run, and the y-coordinate and x-coordinates of the start point and end point of the i-th run are registered in the run list storage section 7c.

Furthermore, in order to express the connection relationship between these two types of strokes, the concept of "node" is used, and the coordinates of the node are stored in the node list storage section 7b.
The strokes connected at that node are registered. The extracted strokes and nodes are shown in Figure 7b.
It becomes as shown in . Here, a position approximately corresponding to the center of the black run appearing in the i-th run is used as the coordinates of the node.

(D) When a non-horizontal run and a horizontal run are connected as shown in Figure 8a, the scan list is updated by repeating sequential scanning, and the i-th run is added to the scan list of the second scan line.
When the (i-1)th run appears in the scan list of the first scanning line, information about the (i-1)th run is first registered in the run list storage section 7c, and
1) It is determined that a vertical or diagonal stroke has ended with a run, and based on the information in the run list storage unit that constitutes this stroke, it is determined whether the vertical stroke is a diagonal stroke or noise, and the stroke is stored in the stroke list storage unit 7a. Register the type and start point and end point coordinates. Furthermore, it is determined that the horizontal stroke starts from the i-th run in the scan list of the second scanning line, and the stroke type of the i-th run is added to the stroke list.
Register the y-coordinate, the y-coordinates of the starting point, and the ending point.
At the same time, in order to represent the connection relationship between these two types of strokes, the coordinates of the node, that is, the coordinates of the center equivalent position of the (i-1)th run, and the strokes connected to the node are registered in the node list storage section 7b. As a result, strokes and nodes are extracted as shown in FIG. 8b.

Note that in the processes (a) to (d) above, input/output processing with the stroke list storage section 7a, node storage section 7b, and run list storage section 7c is performed by the list processing section.

According to this embodiment, for example, when scanning an A4 size original figure at a sub-scan of 8 lines/mm, this method is approximately 1/2 times faster than the method of accumulating all image data of the original figure.
The image data storage capacity is only 1,200, making it possible to realize a low-cost device.

Although the basic parts have been explained in this embodiment, in actual cases, short black runs are often detected at the beginning and end of a horizontal run, as shown in FIG. Such black runs are once registered in the run list as non-horizontal runs, but if they do not appear continuously for a certain number of scanning lines or more, they are treated as noise and the information is discarded. For example, even if two or less consecutive black runs are discarded as noise, not only will this method not have any negative effect on stroke extraction using this method, but the noise component will be reduced and good results will be obtained.

Further, in this embodiment, a case where a single black run exists on one scanning line has been described, but strokes can be extracted in exactly the same way even when a plurality of black runs exist on one scanning line.

Effects of the Invention As described above, the present invention converts the information of the black run in each scanning line into coordinate value data at the same time as reading the original figure, and converts the information of the black run in the scanning line into coordinate value data. By simply providing a scan list storage unit that stores scan lists consisting of coordinate values as the main storage unit, it is possible to extract the horizontal strokes, vertical strokes, and diagonal strokes that make up a line figure, as well as their connection relationships. It has the advantage of requiring less storage capacity, less amount of data, and being able to process with high efficiency and high speed.

In addition, in the part of the series of processing that is closest to the figure input part, image data is converted into encoded and compressed information of strokes and their connection relationships, so compared to conventional line thinning methods, etc. It is possible to significantly reduce the overall processing speed.

[Brief explanation of drawings]

Fig. 1 is a block wiring diagram of a line figure stroke extraction device according to an embodiment of the present invention, Fig. 2 is a diagram showing the distinction between horizontal runs and non-horizontal runs, and Fig. 3 is a diagram showing the connection relationship between black runs. Figure 4 is a diagram showing the basics of the connection relationship between black runs, Figure 5 is a diagram showing the extraction of horizontal strokes from the connection relationship between horizontal runs,
FIG. 6 is a diagram showing the extraction of non-horizontal (vertical or diagonal) strokes from the connection relationship between non-horizontal runs;
Figure 7 shows the extraction method when a horizontal run is followed by a non-horizontal run to form a node, Figure 8 is a diagram showing the extraction method when a non-horizontal run is followed by a horizontal run to form a node, and Figure 9. is a diagram showing a case where a horizontal run is followed by a noisy non-horizontal run. 1...Reading section, 2...Scan list generation section,
3... Scan list storage section, 4... Pattern processing calling section, 5... Pattern processing section, 6... List processing section, 7a... Stroke list storage section, 7b
... Node list storage section, 7c... Run list storage section.

Claims (1)

[Scope of Claims] 1. A reading means for sequentially binarizing observed values while observing a two-dimensional line figure expressed on a two-dimensional plane by line scanning such as a line scanner; A process of converting binary data representing white pixels and black pixels into coordinate value data of the start point and end point on the two-dimensional plane for each set of consecutive black pixels on one scanning line, and scan list generation means for performing a process of counting the number of sets of continuous black pixels that exist;
a scan list storage unit that stores the coordinate value data generated by the scan list generation means and the number of consecutive black pixel sets for two scanning lines starting from the latest one; and the scan list storage unit stores the coordinate value data and the number of consecutive black pixel sets. Based on the data, examine the length of a set of consecutive black pixels on one scanning line on two adjacent scanning lines and the connection relationship between the two adjacent scanning lines of the continuous black pixels, A pattern processing calling means that distributes processing for each connection pattern set in advance, and a node that is activated by the pattern processing calling means and represents strokes that are line portions forming a figure and connecting portions between the strokes. and pattern processing means for performing pattern processing by generating the coordinate value data of the start point and end point of a set of consecutive black pixels on one scanning line of the two scanning lines and the continuous black pixels existing on one scanning line. By updating the number of black pixel sets for one scanning line, distributing processing for each pattern, and repeating pattern processing, horizontal, vertical, diagonal A stroke extraction device for line figures, characterized in that it extracts three types of strokes and their connection relationships. 2. The pattern processing means includes a stroke list storage unit that registers the coordinates of the start and end points of strokes, which are line parts constituting a figure, and the type of stroke, and the coordinates of nodes representing connecting parts between strokes, as well as their coordinates. A node list storage unit that registers strokes connected by nodes, a run list storage unit that registers the coordinates of the start and end points of a set of black pixels that are considered to constitute a vertical or diagonal stroke, and each 2. The line graphic stroke extraction device according to claim 1, further comprising list processing means for manipulating data in the stroke list storage section, the node list storage section, and the run list storage section for each pattern.