JPS5840785B2 - Shape recognition preprocessing method - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to a preprocessing method for figure recognition, and particularly to a preprocessing method for recognizing figures having a closed loop configuration.

Recently, it has been proposed to automatically read flowcharts and automatically create programs based on this read information.

FIG. 1 shows an example of a flowchart, and FIG. 2 shows an example of flowchart symbols, where 1 is a symbol representing input/output, 2 is a symbol representing a terminal, 3 is a symbol representing judgment, and 4 is a symbol representing a judgment. is a symbol representing paper tape, and 5 is a symbol representing a connector.

These figures are all closed-loop configurations.

In order to create a program from a flowchart, it is necessary to understand what the symbols used in the flowchart mean.

Therefore, as shown in FIG. 3, it is necessary to cut out a region containing the symbol and recognize the symbol within the cutout frame.

Here, the problem when recognizing the shape of a symbol, that is, a figure, is the processing of the connecting line portion cut out together with the closed loop figure.

Originally, when an unnecessary surface or line (corresponding to a connecting line part in this figure) is included in a recognition target (corresponding to a closed loop figure in this figure), the object itself becomes unclear and recognition itself becomes impossible. This is often the case.

The present invention has been made in view of the above considerations, and an object of the present invention is to provide a figure recognition preprocessing method capable of removing unnecessary objects added to a recognition target object.

Therefore, the figure recognition preprocessing method of the present invention includes a video memory, and by performing conversion processing on the contents of the video memory, closed loops are extracted from recognition target figures including connection lines connecting closed loops. In the figure recognition pre-processing method for extracting, an individual figure cutting processing unit cuts out a rectangular individual figure having one closed loop and a connecting line portion coming out of the closed loop, and converts a white mesh in the closed loop of the individual figure to a black mesh. a black mesh processing unit that performs conversion processing; a histogram creation processing unit that performs processing to create a histogram of black mesh points by projecting the individual figures subjected to the black mesh processing in a predetermined direction; The present invention is characterized by comprising a connecting line deletion processing unit that determines that a portion of the histogram smaller than the line width of is a portion based on a connecting line portion, and uses this determination result to delete the connecting line portion. .

Hereinafter, the present invention will be explained with reference to the drawings.

Figure 4 is a diagram explaining the process of removing unnecessary parts from a figure without concavities according to the present invention, Figure 5 is a diagram explaining the principle of converting the inside of a closed loop from white to black, and Figure 6 is FIG. 6 is a diagram for explaining the contents of a histogram of black mesh points.

FIG. 4A shows the input figure.

As the first step in processing unnecessary connecting lines, the inside of the closed loop figure is converted from white to black.

Figure 4 shows the result of this conversion.

FIG. 5 explains this conversion.

First, an operation is performed by scanning the cutting frame in the X direction and making everything between the black mesh points black by marking each point on the Y axis.

Next, the cutting frame is manipulated in the Y-axis direction to make everything between the black mesh points black for each point on the X-axis.

If we take the AND of the two figures obtained in this way, we get
The inside of the closed loop shape is converted from white to black.

FIG. 4C shows a model of unnecessary area removal processing.

■1 is a histogram obtained by accumulating black mesh points from right to left for each point on the left side of the cropping frame, ■2
is a histogram obtained by accumulating black mesh points from left to right for each point on the right side, Hl is a histogram obtained by accumulating black mesh points from bottom to top for each point on the upper side, and Hl is a histogram obtained by accumulating black mesh points from bottom to top for each point on the upper side. A histogram obtained by accumulating black mesh points from top to bottom for each point on the lower side is shown.

FIG. 6 explains the contents of the histogram H2, which includes the number of black meshes within the connecting lines.

If we slice the histogram H2 by the line width values of the connecting lines, we get
A histogram corresponding to the black loop figure can be obtained.

Intersections c and d between the histogram H2 and the line S respectively indicate the end points of the line segment obtained by projecting only the closed loop figure onto the lower side from above.

Points a and b are the intersection points when histogram H1 is sliced by line width values, points e and f are the intersection points when histogram ■1 is sliced by line width values, and points g and h are the intersection points when histogram ■2 is sliced by line width values. The intersection points obtained when slicing are shown.

A straight line passing through points a and c, point S, a straight line passing through d, point e.

If a quadrilateral frame surrounded by a straight line passing through g and a straight line passing through points f and h is found, a closed loop figure to be recognized exists within this quadrilateral frame.

FIG. 4 2 shows the quadrilateral frame obtained in this way.

FIG. 7 is a diagram illustrating the process of removing unnecessary parts from a figure having a concave according to the present invention.

FIG. 7A shows an input figure.

The opening in FIG. 7 shows a figure in which the inside of a closed loop figure is converted from white to black.

FIG. 7C shows histograms H1 and H2 of black mesh points, and histogram H4 of white mesh points, respectively.

Histogram H1 is obtained by accumulating black mesh points from bottom to top for each point on the upper side of the cropping frame, and histogram H2 is obtained by accumulating black mesh points from left to right for each point on the right side of the cropping frame. It is obtained by accumulating black mesh points in the direction.

Histogram 4 is obtained by accumulating only the white mesh points existing between black points from left to right for each point on the right side of the cropping frame.

By slicing the histogram H1 at an appropriate value
Therefore, the X coordinate value of the two connecting line parts can be known.

Further, by using the coordinate values obtained by slicing the histograms (2) and (4) by the line width values of the connecting lines, it is possible to know the Y coordinate values of the end points of the two connecting line parts.

By using the X and Y coordinate values obtained in this manner, unnecessary connecting line portions can be deleted.

FIG. 8 shows the overall structure of an embodiment of the present invention in which unnecessary parts are removed from a figure having no concavities.

In FIG. 8, reference numeral 6 indicates an individual figure cutout processing section, 7 a black mesh processing section, 8 a histogram creation processing section, and 9 a connecting line deletion processing section.

The individual figure cutout processing section 6 cuts out a flowchart figure including a closed loop in the form of a large frame including unnecessary connecting line parts.

The black mesh processing section 7 performs a process of converting a white area in a closed loop into a black mesh.

The histogram creation processing unit 8 performs horizontal and vertical direction projection on the cut-out large frame graphic data that has been subjected to the black mesh processing, and creates a histogram of the number of black meshes.

The connecting line deletion processing unit 9 determines that the part of the histogram smaller than the line width S in the histogram data created by the histogram creation processing unit 8 is based on the connecting line part, and deletes that part, thereby changing the flowchart. This method extracts closed-loop graphic parts.

Figure 9A is for explaining the black mesh processing unit 7 that converts the inside of a closed loop figure from white to black.
0 is an address generation circuit, 11 is a video memory, 12 is an incandescent-thermal conversion circuit, 13 to 15 are memories, and 16 is an AN.
D circuits are shown respectively.

The address generation circuit 10 generates an input X coordinate value and a Y coordinate value.
It converts coordinate values into memory addresses.

The white-to-thermal conversion circuit stores video information for -scanning, and if there is two or more pieces of black mesh point information, it converts all the information between the end points into black mesh point information. It is something.

The cutting frame has coordinate values X09X1, and coordinate values YO,
It shall be specified by Y1.

FIG. 9A will be explained.

By setting Y=YO and changing X from XO to Xl, video information is read from the video memory 11 and inputted to the incandescent-thermal conversion circuit 12.

The white-to-heat conversion circuit 12 performs white-to-heat conversion.

A similar operation of storing the incandescent-thermal converted video information at the corresponding address in the memory 13 is performed for each point between MO and Y1.

After the above operations are completed, X=XO is set, Y is changed from ¥1 to YO, and the video information read from the memory 11 is input to the incandescent-thermal conversion circuit 12.

The incandescent-to-thermal conversion circuit 12 performs incandescent-to-thermal conversion and stores the incandescent-to-thermal converted video information at a corresponding address in the memory 14 .

The above operations are performed for each point between XO and X1.

Next, the opening in FIG. 9 will be explained.

The coordinates of each point within the cutting frame are input to the address conversion circuit 10, and video information is read from the memos IJ 13 and 14.

, bit information sequentially read from the memories 13 and 14 is input to an AND circuit 16.

The AND circuit 16 takes the AND of both bit information.

The ANDed bit information is stored at a corresponding address in the memory 15.

FIG. 10 illustrates the histogram creation processing unit 8 that creates a histogram of black mesh points, in which 17 represents an accumulation circuit and 18 represents a memory.

For example, when creating a histogram H1, X=X
0, Y is changed from YO to ¥1, video information is read from the memory 15, the read video information is input to the accumulation circuit 17, and the accumulation circuit 17 stores the black included in the - scan. Accumulate mesh points.

This cumulative value is stored in the memory 18 at addresses corresponding to the four coordinates.

If the above scanning is performed for each point between XO and X1, a histogram H1 can be obtained.

Note that the coordinates of the intersection points c and d between the line width value and the histogram H1 are determined by comparing the read data from the memory 18 and the line width value.
It is determined by storing the X coordinate value when the read data becomes larger than the line width value.

As is clear from the above description, according to the present invention, unnecessary information can be easily removed from the graphic information input to the graphic recognition device.

[Brief explanation of the drawing]

FIG. 1 is a diagram showing an example of a flowchart, FIG. 2 is a diagram showing an example of flowchart symbols, FIG. 3 is a diagram showing a cut-out figure, and FIG. 4 is a figure from a figure without a concave according to the present invention. Figure 5 is a diagram explaining the process of removing unnecessary parts, Figure 5 is a diagram explaining the principle of converting the closed loop from white to black, Figure 6 is a diagram explaining the contents of the histogram of black mesh points, and Figure 7 is a diagram explaining the contents of the histogram of black mesh points. The figure is a diagram explaining the process of removing unnecessary parts from a shape with a concave according to the present invention, FIG. 8 is a diagram showing the overall configuration of one embodiment of the present invention, and FIG. 1 is a diagram illustrating a black mesh processing unit that converts into
FIG. 0 is a diagram illustrating a histogram creation processing section that creates a histogram of black mesh points. 6...Individual figure cutting processing section, 7...
Black mesh processing section, 8... Histogram creation processing section, 9... Connecting line deletion processing section, 10...
...Address generation circuit, 11 Knee...Video memory, 12...White-thermal conversion circuit, 13 to 15
...Memory, 16...AND circuit, 1
7... Accumulation circuit, 18... Memory.

Claims (1)

[Claims] 1 A figure recognition preprocessing method that includes a video memory and extracts closed loops from recognition target figures that include connection lines that connect closed loops by performing conversion processing on the contents of the video memory. , an individual figure cutting processing unit that cuts out a rectangular individual figure having one closed loop and a connecting line portion coming out of the closed loop, and a black mesh that converts a white mesh in the closed loop of the individual figure into a black mesh. a mesh processing unit; a histogram creation processing unit that performs processing to create a histogram of black mesh points by projecting the individual figures subjected to the black mesh processing in a predetermined direction;
A connecting line deletion processing unit that determines that a portion of the histogram smaller than the line width of the connecting line in the histogram is a portion based on the connecting line portion, and uses this determination result to erase the connecting line portion. A shape recognition preprocessing method that uses