JPH0927035A - Linear graphic recognition system - Google Patents

Abstract

PROBLEM TO BE SOLVED: To detect a closed figure and recognize its attribute such as the kind even if a segment is partially broken in a binary two-dimensional digital image or a broken line or alternate long and short dash line. SOLUTION: The linear graphic recognition system which recognizes a closed figure drawn with lines in a binary two-dimensional digital image is equipped with an initial polygon generation part 11 which inputs the two-dimensional digital image and generates a polygon in the closed figure in the two-dimensional digital image, a polygon fitting part 12 which inputs the polygon and two-dimensional digital image and expands and deforms the polygon until it is inscribed in the closed figure, and a figure detection part 13 which decides an attribute such as the kind of the closed figure by inputting the polygon inscribed in the closed figure as a result of the expansion deformation and measuring figure features of the polygon.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a line segment on a binary two-dimensional digital image obtained by inputting a map, a drawing, a flow chart, etc. drawn in a line figure on a paper by an image scanner or the like. The present invention relates to a line figure recognition method capable of detecting a closed figure and recognizing attributes such as the type of the closed figure even when there is a break or when the type of the line is a broken line or an alternate long and short dash line.

This line figure recognition method can be widely used as a basic technique for realizing automatic reading of maps and drawings.

[0003]

2. Description of the Related Art Conventionally, a binary two-dimensional digital image I in which a graphic portion is represented by pixels having a value of 1 and a background portion is represented by pixels having a value of 0
The following method is used to detect a closed figure and recognize its attribute.

First, the image I is thinned to generate a thinned image I '. The thinning process is a process in which, in the process of raster-scanning an image, an operation of converting pixels included in a graphic portion and in contact with the background to 0 is repeated until the graphic becomes a core line graphic having a width of 1.

Next, the skeletonized figure on the thinned image I'is traced and approximated by a series of coordinate values. After that, in consideration of the detection of the closed figure having a break and the closed figure drawn by the broken line and the alternate long and short dash line, the connection processing of the gap having a short distance is performed. The connection process is a process of selecting from the combination of open end point pairs that the distance between them is sufficiently small and that a large angle corner cannot be created when connecting, and interpolating the gap between these open end points to connect. is there. The distance between the open end points and the threshold value of the corner angle for determining whether or not to connect are given in advance.

Then, a closed loop portion is detected on the series data of the coordinate values obtained as a result of the gap connection processing.

Next, an approximation process is applied to the series of coordinate values forming a closed loop in order to remove the influence of digitization error. For the closed loop that has been subjected to the approximation processing, the shape determination is performed by measuring the number of vertices, the size of the internal angle, the length of the side, the parallelism of the opposite sides, and the like. The shape determination result is used as the recognition result of the closed figure corresponding to the closed loop.

[0008]

However, in the conventional method of detecting a closed figure after interpolating and connecting a discontinuity of a line figure, as shown in FIG. 5, in the interpolating and connecting process, as shown in FIG. It may be difficult to find the correct pair of endpoints to connect when is detected. As a result, there is a problem that a closed figure cannot be detected or an incorrect closed figure is detected.

[0009]

According to the present invention, in a line figure recognition system for recognizing a closed figure drawn by a closed line on a binary two-dimensional digital image, the two-dimensional digital image is used as an input. An initial polygon generation unit for generating a polygon inside a closed figure on a three-dimensional digital image, and a polygon fitting section for expanding and deforming the polygon until the polygon and the two-dimensional digital image are inscribed in the closed figure. It is characterized by further comprising a figure detecting unit for determining an attribute such as a type of the closed figure by measuring the figure feature of the polygon with the polygon inscribed in the closed figure as an input as a result of the expansion deformation.

[0010]

The operation of the present invention will be described with reference to FIG.

When a binary image I containing a closed figure is input to the initial polygon generator 11, the initial polygon generator 11 first generates an image G obtained by subjecting the input image I to a figure expansion process. Here, in the graphic expansion processing, while raster scanning the image, it is determined whether or not it is a background pixel adjacent to the contour pixel based on the arrangement of pixel values in a 3 × 3 pixel area centered on the point of interest. If it is a background pixel adjacent to, the pixel value is rewritten from 0 to 1. The contour pixel is a pixel included in the graphic portion and in contact with the background. The graphic expands to the background side by one pixel by one processing. The image G is an image that has been subjected to about 1/2 of the gap distance to which the graphic expansion process is to be connected, and the gap between the line segments is in a filled state.

Next, of the contours of each figure on the image G, the inner contour existing inside the figure is detected. Further, the detected inner contour is approximated, and the obtained closed coordinate value series is output as an initial polygon.

Next, in the polygon fitting section 12, the initial polygon is expanded and deformed on the input image I to be inscribed in a closed figure on the image. A method of expanding and deforming the initial polygon will be described in an embodiment.

Next, in the figure detecting unit 13, the shape is determined by measuring the number of vertices of the polygon obtained as a result of the expansion and deformation, the size of the internal angle, the length of the side, the parallelism of the opposite sides, and the like. The shape determination result of the polygon is used as the determination result of the closed figure with which the polygon is in contact.

As described above, first, after the polygon is generated in the closed figure, the polygon is expanded and deformed and inscribed in the closed figure, so that the closed point is not connected without connecting a gap or a broken line gap. It becomes possible to detect the shape of the figure. As a result, it is possible to solve the problem that a correct polygon is not detected due to an error in the conventional connection processing.

[0016]

First, the overall structure of the present invention will be described. FIG. 2 is a configuration diagram of an embodiment of the present invention, and the same parts as those in FIG. 1 are designated by the same reference numerals.

Reference numeral 11 denotes an initial polygon generation unit for generating and outputting an initial polygon included in a closed figure drawn by a solid line, a broken line, a dash-dotted line or the like or a closed figure drawn by a solid line on the input binary image. Is. The initial polygon is a polygon formed by connecting N nodes Q1, Q2, ..., QN in order. Here, the number of nodes N is given in advance. The initial polygon generator 11 includes a graphic expansion processor 11a, an inner contour extractor 11b, and a polygon generator 11c.

The graphic expansion processing unit 11a performs expansion processing a predetermined number of times on the input binary image I. The image I input to the graphic expansion processing unit 11a is first raster-scanned, and marks are added to pixels adjacent to the contour pixels. Here, the mark is added to a plane different from the plane that stores the pixel value of the image I. Whether or not the pixel is adjacent to the contour pixel is determined from the arrangement of pixel values in a 3 × 3 pixel area centered on the pixel of interest during raster scanning.
When the pixel value of the pixel of interest is 0 and any of the 8 pixels in the vicinity thereof is 1, the pixel of interest is determined to be the background pixel to be obtained. After the raster scanning of the entire image I is completed, the value of the marked pixel is rewritten from 0 to 1. By the above processing, the image G in which the figure is expanded by one pixel toward the background side is generated.

The inner contour extracting unit 11b extracts the contour of the black pixel connected component in the input binary image G by tracing the contour, and then selects and outputs only the inner contour. The contour tracking is a process in which after detecting a contour pixel of a black pixel connected component which has not been found by raster scanning, a contour pixel adjacent to the contour pixel on the black pixel connected component is sequentially tracked and detected. Tracked pixels are given a tracked mark on another plane of image G. When the contour for one round is detected, raster scanning is performed again, and when undetected contour pixels are detected, the same processing as described above is performed. As a result of the above processing, the contours in all the images are detected. The contour detection result is stored as a coordinate value sequence of contour pixels.

The contour pixels are detected by raster scanning by checking the values of the pixel of interest p and its neighboring pixels {p0, p2, p4, p6}. FIG. 3 shows the relative positional relationship between the pixel of interest p and its neighboring pixels {p0, p2, p4, p6}. The pixel of interest p has a value of 1, and neighboring pixels {p0, p
2, p4, p6} has the value 0, the pixel of interest p
Is a contour pixel.

For each contour detected by raster scanning and contour pixel tracking processing, the sum of the deviation angles when the series of coordinate values is regarded as a polygon is calculated, and the outer contour of the figure is the inner contour ( Inner contour). The sum of declination is +2
When it becomes π, it is the outer contour, and when it becomes −2π, it is the inner contour. The series of coordinate values of the inner contour of the figure detected by the above processing is output to the polygon generator 11c.

The polygon generator 11c is a process for inputting a series of coordinate values of the inner contour and approximating the inner contour with a polygon having a predetermined number N of vertices. The input coordinate value series of each contour is subjected to thinning processing to be converted into a polygon having the number of vertices N.

The above is the configuration of the initial polygon generator 11.

Reference numeral 12 expands and deforms the polygon generated by the initial polygon generator 11 on the input image I to inscribe the closed figure, and then outputs the polygon inscribed to the closed figure to the figure detector 13. This is the polygon fitting section. The polygon fitting unit 12 includes a polygon deforming unit 12a and a deformation stop determining unit 12b.

The polygon deforming section 12a expands and deforms the input polygon. The polygon is input from the initial polygon generation unit 11 or the deformation stop determination unit 12b. The expansion deformation was such that, for three consecutive nodes Qn-1, Qn, Qn + 1 among the nodes Q1, Q2, ..., QN of the polygon, the interior angle of the polygon formed by Qn-1, Qn, Qn + 1 was less than 180 degrees. At this time, the positions of Qn-1 and Qn + 1 are moved by a minute distance in the direction of expanding the interior angle. However, with respect to the polygon re-input from the deformation stop determination unit 12b, only the permitted nodes are moved.
The above operation is performed for all combinations of three consecutive nodes. The polygon after expansion and deformation is the deformation stop determination unit 1
2b is output.

The deformation stop determination unit 12b calculates the sum of the distances and interior angles of the input polygon with respect to the figure containing the polygon, and the respective nodes Q1, Q2, ... Of the polygon.
.., Determine whether to allow further movement of QN. If the distance dn between the node Qn and the figure is dn> 0, further movement of the node Qn is permitted. When there is no deformable node, it is determined that the polygon is inscribed in the closed figure, and the expansion process is completed. The polygon whose expansion processing has been completed is output to the figure detection unit 13.

The above is the configuration of the polygon fitting section 12.

A polygon 13 output from the polygon fitting part is first approximated, and then the number of corners, the corner angle, the length of the side, the parallelism of the opposite sides, etc. are measured, and the polygon is inscribed. The figure determination unit determines the attribute of the closed figure and finally outputs the result. The type of polygon is determined from the number of corners, and the type of special polygon such as a parallelogram or a square is determined from the parallelism of opposite sides and the length of the sides. The conditions necessary for these attribute determinations are registered in advance as dictionary data.

The above is the overall configuration.

Next, the overall operation of the present invention will be described.

When a binary image I containing a closed figure is input to the initial polygon detecting section 11, the figure expanding section 11a performs a figure expanding process to generate an image G in which gaps such as breaks and broken lines are filled. It At this time, it is assumed that the number of times of the graphic expansion processing is preset to about 1/2 of the gap distance to be connected. The image G is the inner contour detection unit 11
b.

The inner contour detecting section 11b performs a process of detecting the inner contour in the image G, and outputs the detection result. The detection result of the inner contour is input to the polygon generation unit 11c, where it is approximated, and the polygon fitting unit 12 determines that the polygon has N predetermined nodes.
Output to This polygon is called an initial polygon. The initial polygon and the binary image I are input to the polygon fitting unit 12.

The polygon deforming section 12a inside the polygon fitting section 12 expands and deforms the input initial polygon by a small amount. The polygon which has undergone a slight amount of expansion and deformation is judged by the deformation stop judgment unit 12b using the binary image I whether or not it is inscribed in the closed figure. When it is determined that the polygon is not inscribed, the evaluated polygon is re-input to the polygon deforming unit together with the information of the node that can be further moved. At this time, the polygon deforming unit 12a further expands and deforms the re-input polygon. As a result of repeating the above processing, when the deformation stop determination unit 12b determines that the polygon is inscribed in the closed figure, the polygon is output to the figure detection unit 13. Examples of the initial polygon and the final polygon output to the figure detection unit 13 are shown in FIG.
And (b).

Next, in the figure detecting section 13, the polygon obtained as a result of the expansion and deformation is approximated. For the approximated polygon, the number of vertices, the size of the internal angle, the length of the side, the parallelism of the opposite sides, etc. are measured, and the shape of the closed figure inscribed by the polygon is determined.

[0035]

As described above, according to the present invention, a gap between line elements such as broken lines caused by a broken portion of a closed figure or a closed figure being drawn by a broken line as in the conventional art is used as a line figure on a thinned image. Rather than interpolating from the open end point pair search, the polygon generated inside the closed figure on the input image is expanded and deformed until it is inscribed in the closed figure, and the attribute of the closed figure is determined by the shape of the obtained polygon. As configured. as a result,
In the conventional method, it is possible to correctly detect and recognize the attribute of a closed figure that cannot be detected because the discontinuity cannot be uniquely determined. Therefore, it is possible to perform highly accurate closed figure recognition that exceeds the conventional method.

[Brief description of drawings]

FIG. 1 is a block diagram showing a configuration of the present invention.

FIG. 2 is a block diagram showing one embodiment of the present invention.

FIG. 3 is a diagram for explaining a method of detecting contour pixels.

4A is a diagram showing a relationship between a closed figure and an initial polygon, and FIG. 4B is a diagram showing a result of expanding and deforming the initial polygon.

FIG. 5 is a diagram showing an example of an image in which it is difficult to detect a closed figure by a conventional method.

[Explanation of symbols]

 Reference Signs List 11 initial polygon generation unit 11a graphic expansion processing unit 11b inner contour extraction unit 11c polygon generation unit 12 polygon fitting unit 12a polygon deformation unit 12b deformation stop determination unit 13 graphic detection unit

Claims (4)

[Claims] 1. A line figure recognition system for recognizing a closed figure drawn by a closed line on a binary two-dimensional digital image,
An initial polygon generation unit that generates a polygon inside a closed figure on the two-dimensional digital image by using the two-dimensional digital image as an input, and the polygon and the two-dimensional digital image as inputs until the inscribed shape is applied to the closed figure. A polygon fitting part that expands and deforms a polygon, and a graphic detection that determines attributes such as the type of the closed graphic by measuring the graphic features of the polygon by inputting the polygon inscribed in the closed graphic as a result of the expansion and deformation. And a line figure recognition method. 2. The initial polygon generation unit, a graphic expansion processing means for generating an image in which a gap is filled, an inner contour detection means for detecting an inner contour in the image, an approximation processing for the inner contour, 2. The line figure recognition system according to claim 1, further comprising a polygon generation means for generating a polygon having a predetermined number of N nodes. 3. The polygon fitting section, a polygon deforming means for slightly expanding and deforming an input initial polygon, and a deformation stop determination for judging whether or not the minutely expanded and deformed polygon is inscribed in a closed figure. 3. The line figure recognition system according to claim 1, further comprising: 4. The figure detecting unit approximates a polygon obtained as a result of expansion and deformation, and calculates the number of vertices, the size of the internal angle, the length of the side, the parallelism of the facing sides, etc. with respect to the approximated polygon. The line figure recognition method according to claim 1, 2 or 3, wherein the shape of the closed figure in which the polygon is inscribed is measured.