JPH07182509A - Graphic recognition device - Google Patents

Abstract

PURPOSE:To comparatively prevent random noise from affecting by dividing a block based on the dispersion of each sample point sandwiched by two reference points and an elevation angle formed by two vectors linking both the reference points. CONSTITUTION:Point data with the dispersion of the elevation angle and a maximum elevation angle change outputted from a separate elevation angle calculating means 105 are outputted to a reference point setting means 106. At the means 106, the dispersion of the elevation angle is compared with a prescribed first threshold value and when it is larger than the first threshold value, the point data with the maximum elevation angle change are outputted and stored in a reference point storage means 107 as the new reference point to divide the block between two reference points. When no new reference point is set, at the means 107, the curvature of that stored block is compared with the curvature of a block adjacent to that block and when the difference of respective curvatures is smaller than a prescribed second threshold value, two blocks are merged into one block. Therefore, dividing processing is continued until the blocks not to decide the division are eliminated in the means 107. Then, the reference point remained through this processing is defined as a feature point.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a figure recognition apparatus for recognizing a line figure, and more particularly to a handwritten figure recognition apparatus using a digitizer or a tablet.

[0002]

2. Description of the Related Art Generally, when a figure is recognized by a drawing reading device for reading a drawing by optical means or an interactive drawing input device using a digitizer or a tablet,
The lines that make up the figure are subdivided into points (called feature points) that divide the line into basic sections such as straight lines and arcs, and the figure recognition is performed by specifying the line types and the positions of the feature points in that section. There is. For that purpose, it is necessary to determine whether or not all the points forming the figure are characteristic points.

Conventionally, in order to determine whether or not a feature point is present, the following two types of means are roughly classified.

The first conventional example is, for example, Japanese Patent Publication No. 4-62.
As described in Japanese Patent Application Laid-Open No. 107 or Japanese Patent Application Laid-Open No. 63-24473, each point formed into a sequence of points is set as a sample point at a predetermined distance, and three sample points separated by n points can be connected. There is a feature point in which an elevation angle formed by two vectors or a rate of change in elevation angle is greater than a predetermined threshold value.

FIG. 3 is a block diagram showing the structure of the first conventional example. Referring to FIG. 3, this prior art
Graphic input means 10 for reading a graphic by optical means or the like
1, the point sequence reading means 102 for converting the read figure into a point sequence, and the point data input from the point sequence reading means 102 are output as sample points for each point separated by a predetermined distance or more. Thinning means 103, point data storage means 104 for sequentially storing a predetermined number of sample point data output from the thinning means 103, and point data storage means 1
For point data in 04, 3 consecutive points separated by n
Based on an elevation angle formed by two vectors formed by connecting two sample points in order, it is determined whether or not it is a feature point, and the adjacent elevation angle calculation means 150 that outputs point data of the feature point and the adjacent elevation angle calculation means 150 Line type discriminating means 108 for discriminating whether the section between the characteristic points is a straight line or a curved line from the discriminated point data and the sample point data between the characteristic points; and the characteristic point position output by the adjacent elevation angle calculating means 150, It is composed of a graphic shape determination means 109 for recognizing the shape of the graphic based on the line type output by the line output determination means 108, and a recognition result display means 110 for displaying the recognition result output by the graphic shape determination means 109 on a display or the like. .

Next, the operation of the first conventional example will be described with reference to FIG.

The processing contents of the figure input means 101 and the point sequence reading means 102 are somewhat different between the drawing reading device for reading the drawings by optical means and the interactive drawing input device using a digitizer or tablet. Therefore, each will be described.

In the case of a drawing reading device for reading a drawing by optical means, the figure input means 101 converts the drawing into image data by a CCD sensor and supplies it to the point sequence reading means 102. For example, the point sequence reading means 102 that operates under program control converts the image data of the drawing into continuous point sequence data by performing thinning processing.

On the other hand, in the case of an interactive drawing input device using a digitizer or a tablet, the figure input means 101 converts a human handwriting into point data including coordinate values, for example, by reading a sequence of points by a resistive film sensor. The soup stock 102 is supplied. In this case, the point sequence reading means 102 reads the point data from the figure input means 101 at regular time intervals.

Since the subsequent processes are almost the same, they will be described collectively. It should be noted that the means thereafter is assumed to be configured by program control, except where described.

The thinning means 103 is a point sequence reading means 1
02 and the point data storage means 10
Find the distance from the last output sample point for 4,
When the distance is shorter than the predetermined threshold value, the point is discarded, and when the distance is equal to or larger than the predetermined threshold value, the point is output to the point data storage means 104 as a sample point.

The point data storage means 104 is, for example, a RAM.
It is a FIFO configured by, and can store a predetermined number of point data sequentially. When the point data exceeding the predetermined number is input, the old point data are discarded in order.

The adjacent elevation angle calculating means 150 determines whether or not the point data in the point data storing means 104 is a feature point based on the elevation angle formed by two vectors formed by connecting three consecutive sample points separated by n. To judge.

When n = 1, the elevation angle formed by two vectors formed by sequentially connecting three adjacent sample points is shown.

Further, the elevation angle may be subjected to arithmetic mean with the elevation angles at points within m points before and after, that is, smoothing processing by a moving average method.

In order to obtain the feature point based on this elevation angle, it may be judged as a feature point when the magnitude of the elevation angle is larger than a predetermined threshold value, or when the difference between two adjacent elevation angles is larger than the predetermined threshold value. You may judge it as a characteristic point.

In the line type discriminating means 108, for example, it is determined whether the average value of the elevation angles corresponding to the respective sample points sandwiched by two consecutive characteristic points output from the adjacent elevation angle calculating means 150 is larger than a predetermined value. , Determine whether the section is a straight line or a curve. Further, when it is determined to be a curve, it is determined whether or not the variance of each elevation angle corresponding to each sample point sandwiched between two feature points is smaller than a predetermined value to form an arc.

The figure shape determining means 109 is defined in advance for each figure category to be recognized, based on the characteristic point positions determined to be characteristic points by the adjacent elevation angle calculating means 150 and the line types determined by the line type determining means 108. Matching with the recognized recognition dictionary data is obtained, the shape of the figure is recognized according to the degree of matching, and the result is supplied to the recognition result display means 110.

The recognition result display means 110 displays the recognition result graphic. In addition, it is also possible to make a clean copy by correcting the shape or line type of a figure based on the recognition result.

The second conventional example is, for example, Japanese Patent Laid-Open No. 58-6.
As described in Japanese Patent No. 2767, a straight line formed by connecting two arbitrary points with the configuration shown in FIG.
After dividing the section so that the maximum value of the distance between the point sequence data of the section sandwiched between the points is less than or equal to a predetermined threshold value, and performing line approximation to connect the divided sections, two adjacent straight lines For each part of the polygonal line consisting of, the judgment is made based on the magnitude of the elevation angle at the center, such as "is a corner", "is a part that should be combined into one curve", and "is a part that should be combined into a straight line". Here, the division points remaining until the end correspond to the feature points.

Similarly, as described in Japanese Patent Application Laid-Open No. 4-1606887, the area of a region surrounded by a straight line formed by connecting arbitrary two points and a point sequence of a section sandwiched between the two points There is also a method of determining whether or not to divide the section depending on the size.

FIG. 4 is a block diagram showing an embodiment of the second conventional example.

Referring to FIG. 4, this embodiment is different from the first conventional example shown in FIG. 3 in the following points.

In this embodiment, instead of the adjacent elevation angle calculating means 150 of the first conventional example which receives the point data output from the point data storing means 104 and outputs the characteristic points, the reference point storing means 153 outputs 2 One reference point and the point sequence data output from the point data storage means 104 between the two reference points are input, and the maximum value of the distance between the straight line connecting the reference points and the point data and the maximum distance point sequence data Is provided, for example, the deviation distance calculating unit 151 configured by program control is provided, and the deviation distance calculating unit 1 is provided in addition to the configuration of the first conventional example.
When the maximum value of the distance output by 51 is equal to or larger than a predetermined threshold value, the reference point setting means 152 configured by, for example, program control, which outputs point data of the maximum distance output by the deviation distance calculating means 151, and, for example, RAM And a reference point storage unit 153 for storing the reference point configured by.

Next, with reference to FIG. 4, an operation relating to a part of the operation of the second conventional example different from that of the first conventional example will be described.

The reference point storage means 153 stores point data of arbitrary two different points such as a stroke start point and a stroke end point as initial values.

The deviation distance calculating means 151 receives two consecutive reference points from the reference point storing means 153, and then
Point data of a point sequence sandwiched between two reference points is read from the point data storage means 104. The distance between the straight line connecting the two reference points and each read point data is obtained, the maximum value of the distance and the point data having the maximum distance are calculated, and the reference point setting means 15
Output to 2.

The reference point setting means 152 compares the maximum value of the distance input from the reference point setting means 152 with a predetermined threshold value, and when the maximum value of the distance is equal to or larger than the threshold value, the reference point setting means 152 inputs the value. The reference point storage means 153 stores the point data having the maximum distance.

The above processing is repeated as long as two continuous reference points that are not evaluated by the deviation distance calculating means 151 exist in the reference point storing means 153.

After performing a polygonal line approximation connecting the sections divided in this way, for each part of the polygonal line consisting of two adjacent straight lines, from the magnitude of the central elevation angle,
The determination is made as to "which is the corner", "the part that should be grouped into one curve", and "the part that should be grouped into one straight line", and the sections are integrated.

[0031]

When an optical reading device is used in the graphic input means 101, random noise called edge noise occurs when the analog waveform output from the CCD sensor is binarized.

Also, when using a device such as a tablet for reading human handwritten loci, random noise such as camera shake occurs. Furthermore, especially in the case of a display-integrated tablet in which an LCD and a tablet are integrated, the LCD
Burst noise occurs due to the effect of electromagnetic waves emitted by the device.

Here, an example of the output result of the graphic input means 101 containing random noise and burst noise is schematically shown in FIG. 5 (a).

Therefore, in the first conventional graphic recognition apparatus described above, the thinning-out determination means 103 uses a distance long enough to eliminate the influence of noise at the time of graphic input, or the adjacent elevation angle calculation means 150. In, the number of sample points (n) and the number of points used in the moving average method (m) are adjusted to adjust to the noise state.

However, this method is difficult to adjust, and the influence of noise may not be sufficiently removed.

That is, in the case of adjustment in the thinning process,
If the thinning distance is made too large in order to increase the noise removal capability, points near the true feature points will be thinned, and the feature points will be misaligned, making it impossible to make a clean copy with the correct shape. The point-likeness is calculated to be low, which causes a mistake in the number of characteristic points and line type.

As shown in FIG. 4B, when a distance having a length that removes only random noise is set as a threshold value, burst noise cannot be removed by the thinning-out determination means, and its position is erroneously recognized as a feature point. To do.

Further, as shown in FIG. 4C, if a distance having a length that also removes burst noise is set as a threshold value, points near the true feature point may be thinned out. The elevation angle formed between the vectors connecting the adjacent sample points in the data storage unit 104 becomes small, and the adjacent elevation angle calculation unit 150 calculates the feature point likelihood as low.

In the case of adjustment by the moving average method, if the number of points (m) used for smoothing is increased too much in order to increase the noise removal capability, the curve is smoothed too much and a curve other than a circular arc is generated. It may be erroneously recognized as an arc or it may be difficult to determine the line type of a short line.

The elevation function obtained by the moving average method is
When there is little noise, for example, as shown in FIG. 6A, the portion corresponding to the arc portion or the straight portion of the input figure becomes a horizontal line, and the portion corresponding to the connection point has a step or a peak. Occurs. Further, the portion corresponding to the curved portion other than the circular arc of the input figure becomes a straight line or curved line having an inclination.

However, when the burst noise or the like occurs to such an extent that it cannot be removed, for example, as shown in FIG. 6B, it becomes difficult to extract the above-described relationship between the line type of the input figure and the feature of the elevation angle function. , Will be lost.

Further, in the case of adjusting the number (n) between sample points, the effect of removing the burst noise is small. That is, when the central sample point is burst noise, the elevation angle becomes unreasonably large, and that portion is often mistakenly recognized as a feature point.

Further, if the number (n) is too large in order to increase the random noise removing ability, the elevation angle becomes unreasonably large in the curved portion having a large curvature. Therefore, characteristic points may occur in the curved portion.

In the second conventional figure recognition apparatus, the influence of random noise is relatively small. In particular, in the example of using the area of the region surrounded by the straight line connecting the reference points and the sequence of points sandwiched between the reference points as the determination guideline of the division, the effect of burst noise is small.

However, with this method, there is no problem in the case of a figure composed of only straight lines, but there is a problem in the case of a figure with a curved part because the curved section is approximated by a polygonal line composed of several straight lines. .

That is, when the curved line portion is approximated by a polygonal line, even if it is a part of the same circular arc, the elevation angles of the polygonal line vary due to variations in division due to the threshold value.

Performing the integration processing into a curve using such elevation angle and elevation angle transition causes erroneous recognition.

Further, since many unnecessary divisions are performed on the curved portion, the processing speed is also disadvantageous.

[0049]

In order to solve the above-mentioned problems, a figure recognition apparatus according to the present invention uses an arbitrary two points in a point sequence forming a line figure as an initial setting value of a reference point, The reference point and the curvature obtained from an average value of elevation angles described later corresponding to a section sandwiched between two consecutive reference points are stored in the order of connection of columns, and the reference point is sandwiched between the two.
When the difference in curvature corresponding to one section is less than or equal to a predetermined second threshold value, the reference point is deleted, the two sections are integrated, and the curvature of the integrated section is calculated from the curvatures of the two original sections. An interval coefficient k which is determined by the number of point sequences in the section sandwiched between the reference point storage means and the two consecutive reference points stored in the reference point storage means.
According to the above, regarding the point sequence of the section excluding the points in the vicinity of the both reference points, two line segments from the one reference point to the other reference point are sequentially set as the attention points at k intervals. Of the elevation angle made by, the point data of the point of interest where the change of the elevation angle is the largest, the separation elevation angle calculating means for obtaining the average value of the elevation angle, the dispersion of the elevation angle output by the separation elevation angle calculating means , Greater than a predetermined first threshold,
As a new reference point for dividing the corresponding section, there is provided a reference point setting means for storing the point data of the point of interest having the largest change in elevation angle in the reference storage means.

[0050]

DESCRIPTION OF THE PREFERRED EMBODIMENTS The present invention will be described with reference to the drawings.

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

Referring to FIG. 1, the embodiment of the present invention has a configuration different from the conventional example shown in FIG. 4 in the following points.

In this embodiment, the conventional deviation distance calculating means 1 is used.
In place of 51, a separation elevation angle calculating means 105 is provided, and in place of the conventional reference point setting means 152, reference point setting means 106.
In addition to the conventional reference point storage means 153, the reference point storage means 107 is provided.

The separation elevation angle calculating means 105 and the reference point setting means 106 operate under program control, for example. The reference point storage means 107 includes, for example, a control unit that operates under program control and a storage unit that has a linear list structure or a tree structure.

Next, with reference to FIG. 1, an operation relating to a part of the operation of this embodiment different from the above-mentioned conventional example will be described.

At the start of recognition, the reference point storage means 107
Stores, as initial values, point data of arbitrary two different points such as a stroke start point and a stroke end point.

Division processing for determining whether or not to divide a section sandwiched between two consecutive reference points in the reference point storage means 107 into finer sections, and the reference point storage means 1
An integration process for determining whether or not to integrate two continuous sections in 07 into one section is performed as follows.

The separation elevation angle calculation means 105 receives two consecutive reference points from the reference point storage means 107, and then 2
Point data of a point sequence excluding the vicinity of both reference points sandwiched between two reference points is read from the point data storage means 104, and is set as a point of interest in order.

At this time, setting all points as the points of interest means
Since it is redundant and wastes calculation time, it may be thinned out at regular intervals. Furthermore, the width of thinning may be increased or decreased according to the number of point sequences in the section so that the required number of points of interest can be obtained even in a short section and the thinning effect can be obtained even in a long section.

For each point of interest, the elevation angle formed by the two line segments from one reference point to the other reference point via the point of interest is determined, the average and variance are determined, and the change in elevation angle is the largest. Obtain the point data of the point of interest.

The point data output by the separated elevation angle calculation means 105 at which the variance of the elevation angle and the change in elevation angle are the maximum are output to the reference point setting means 106.

The reference point setting means 106 compares the variance of the elevation angle with a predetermined first threshold value. Then, when the variance of the elevation angle is equal to or greater than the predetermined first threshold value, the point data with the largest change in elevation angle is output, and is used as a new reference point for dividing the section between the two reference points. Reference point storage means 1
It is stored in 07.

The average value of the elevation angles output from the separation elevation angle calculation means 105 is output to the reference point storage means 107. It is converted into the curvature of the section sandwiched between the two reference points and stored.

In the reference point storage means 107, when a new reference point that divides the section sandwiched between the two reference points is not set, the curvature of the section stored earlier and the adjacent section are stored. Compare with the curvature of the section.

When the difference between the curvatures is equal to or smaller than the predetermined second threshold value, the two sections are integrated into one section.
That is, the reference point sandwiched between the two sections is deleted,
Calculate and set the curvature of the integrated section from the curvature of the original section (by taking the average, etc.).

The method of obtaining the curvature from the average elevation angle may be as follows, for example. First, a straight line connecting both reference points is taken as a base, and a first isosceles triangle whose apex angle is equal to the average elevation angle is obtained. Next, using either one of the first isosceles triangles as the base, the second isosceles triangle whose low angle is half the average elevation angle is obtained. Then, the length of the hypotenuse of the second isosceles triangle becomes the curvature.

As described above, the reference point storage means 107
The division process is performed until there is no section in which division determination is not performed. After such processing, the remaining reference points are set as feature points.

A specific description will be given with reference to FIG.

If the input figure is as shown in FIG. 2A, the characteristic points are three points A, D and B.

Since the two points A and B are end points, detection is easy. Therefore, these two points are stored in the reference point storage means 107 as initial values of the reference points, and the division processing is started.

The reference point storage means 107 outputs the points A and B to the separated elevation angle calculation means 105 as reference points.

The spaced-apart elevation angle calculation means 105 sets the point sequence excluding the vicinity of both AB reference points in order at an appropriate interval according to the number of point sequences between AB, and sets the reference point A to the reference point. The elevation angle of the line segment reaching the point B is obtained, the average value and variance of the elevation angles are calculated, and the point of interest with the largest change in elevation angle is calculated.

The average value of the elevation angles is output to the reference point storage means 107, converted into the curvature of the section sandwiched between the reference points AB, and stored.

The elevation angle with AB as both reference points is as shown in FIG. 2B in the ideal state where there is no noise.

Since the fluctuation due to random noise is very small compared to the distance between the reference point and the point of interest, it is not greatly affected, but the fluctuation due to burst noise is not negligible compared with the distance between the reference point and the point of interest. Is the size of.

If there is a burst noise at the point C between the curved portions A and D, the elevation angle with AB as both reference points is
It becomes like FIG.2 (c).

Hereinafter, description will be made assuming that the input is in the state as shown in FIG.

The separated elevation angle calculating means 105 selects the point C as the point of greatest interest in the change in elevation angle.

The reference point setting means 106 compares the variance of the elevation angle output from the separated elevation angle calculating means 105 with a predetermined first threshold value. As can be seen from FIG. 2 (c), the elevation angle varies considerably, and the dispersion of the elevation angle becomes larger than the predetermined first threshold value.

Therefore, the reference point setting means 106 stores the point C in the reference point storage means 107 as a new reference point.

At this point, the reference point storage means 107 stores
Three points, point A, point C, and point B, are stored, and two sections, AC and CB, exist as sections for which division determination has not been performed.

Next, the reference point storage means 107 outputs the points A and C to the separated elevation angle calculation means 105 as reference points.

Similarly, the separation elevation angle calculation means 105 sets the point sequence excluding the vicinity of both AC reference points in order at appropriate intervals according to the number of point sequences between ACs, and determines the reference point from the reference point A. Then, the elevation angle of the line segment reaching the reference point C is obtained, the average value and variance of the elevation angles are calculated, and the point of interest with the largest change in elevation angle is calculated.

The average value of the elevation angles is output to the reference point storage means 107, converted into the curvature of the section sandwiched between the reference points AC, and stored.

The elevation angle with AC as both reference points is shown in FIG.
become that way.

The separated elevation angle calculation means 105 selects the point D as the maximum focus point of the change in elevation angle.

The reference point setting means 106 compares the variance of the elevation angle output from the separated elevation angle calculating means 105 with a predetermined first threshold value. As can be seen from FIG. 2 (d), the elevation angle varies considerably, and the dispersion of the elevation angle becomes larger than the predetermined first threshold value.

Therefore, the reference point setting means 106 stores the point D in the reference point storage means 107 as a new reference point.

At this point, the reference point storage means 107 stores
Four points, point A, point D, point C, and point B are stored, and there are three sections, AD, DC, and CB, which are not subjected to division determination.

Next, the reference point storage means 107 outputs points A and D as reference points to the separation elevation angle calculation means 105.

The separated elevation angle calculation means 105 similarly calculates the average value and variance of the elevation angles, and further calculates the point of interest with the largest change in elevation angle.

The average value of the elevation angles is output to the reference point storage means 107, converted into the curvature of the section sandwiched between the reference points AD, and stored.

The elevation angles with AD as both reference points are all substantially 0 because the section AD is a straight section.

That is, since the variance of the elevation angle becomes a very small value, the reference point setting means 106 does not divide the section AD.

At this point, the reference point storage means 107 stores
Four points, point A, point D, point C, and point B are stored, and two sections, DC and CB, exist as sections for which division determination has not been performed.

Next, the reference point storage means 107 outputs the points D and C as the reference points to the separation elevation angle calculation means 105.

The separated elevation angle calculation means 105 similarly calculates the average value and variance of the elevation angles, and further calculates the point of interest with the largest change in elevation angle change.

The average value of elevation angles is output to the reference point storage means 107, converted into the curvature of the section sandwiched between the reference points DC, and stored.

The elevation angle with DC as both reference points has a substantially constant value because the section DC is an arc section. That is, since the variance of the elevation angle is a very small value, the reference point setting unit 106 does not divide the section DC.

Here, in the reference point storage means 107,
Although the curvature of the section AD and the curvature of the section DC are compared, they are not integrated because they are the infinity and the radius of the section DC.

At this point, the reference point storage means 107 stores
Four points, point A, point D, point C, and point B are stored, and one section CB exists as a section for which division determination is not performed.

Next, the reference point storage means 107 outputs points C and B as reference points to the separated elevation angle calculation means 105.

The separated elevation angle calculation means 105 similarly calculates the average value and variance of the elevation angles, and further calculates the point of interest with the largest change in elevation angle change.

The average value of the elevation angles is output to the reference point storage means 107, converted into the curvature of the section sandwiched between the reference points CB, and stored.

The elevation angle with CB as both reference points takes a substantially constant value because the section CB is an arc section.

That is, since the variance of the elevation angle becomes a very small value, the reference point setting means 106 does not divide the section CB.

Here, the curvature of the section DC and the curvature of the section CB are compared. Since the section DC and the section CB are originally the same arc, the difference between the two curvatures is a value smaller than the second threshold value, and the section DC and the section CB are integrated. That is, the reference point C is deleted, and the curvature of the section DB is obtained from the curvature of the section DC and the curvature of the section CB.

At this point, since there is no section in the reference point storage means 107 for which division determination has not been made, three stored reference points--point A, point D, and point B--are output as feature points. It

[0109]

As described above, in the figure recognition apparatus according to the present invention, the interval of the section is determined based on the variance of the elevation angle formed by the two vectors connecting the sample points sandwiched between the two reference points and the two reference points. Make a split.

Since the vector connecting the sample points between the reference points and the reference point is much longer than the vector connecting the adjacent sample points, the elevation angle variation is small even if the position of the sample point varies. That is, it is relatively unaffected by random noise.

Further, for the same reason, the elevation angle, which is the basis for the division determination of the section, can be obtained with high accuracy from the input figure having the same resolution.

Since the division of the section is determined by taking the variance of these elevation angles, it is relatively unaffected by the burst noise, which rarely occurs even if the variation distance is large.

Therefore, excellent feature points can be extracted which are relatively unaffected by random noise and burst noise.

Further, in the case of not only a straight line but also a curve having the same curvature, the variance of the elevation angle is small, so that the arc is not divided and an erroneous feature point is not generated in the middle of the curved portion.

[Brief description of drawings]

FIG. 1 is a block diagram showing an embodiment of a figure recognition device according to the present invention.

FIG. 2 is an explanatory diagram of an operation of the embodiment shown in FIG.

FIG. 3 is a block diagram showing an embodiment of a first conventional graphic recognition apparatus.

FIG. 4 is a block diagram showing an embodiment of a second conventional graphic recognition apparatus.

FIG. 5 is an explanatory diagram of an output of the thinning means 103 shown in FIGS.

6 is an explanatory diagram of an adjacent elevation angle calculating means 150 shown in FIG.

[Explanation of symbols]

 101 Graphic Input Means 102 Point Sequence Reading Means 103 Thinning Means 104 Point Data Storage Means 105 Separation and Elevation Angle Calculation Means 106 Present Reference Point Setting Means 107 Present Reference Point Storage Means 108 Line Type Discrimination Means 109 Graphic Shapes Judgment means 110 Recognition result display means 150 Adjacent elevation angle calculation means 151 according to first conventional example 151 Deviation distance calculation means according to second conventional example 152 Reference point setting means 153 according to second conventional example Reference point storage according to second conventional example means

─────────────────────────────────────────────────── ─── Continuation of the front page (51) Int.Cl. ⁶ Identification code Office reference number FI Technical display location G06K 9/48 9289-5L 9061-5L G06F 15/70 365

Claims (6)

[Claims] 1. An input line figure is converted into an array of point data including at least coordinate values, characteristic points (end points, bending points and inflection points) are calculated from the arrangement of the point data, and basic points such as straight lines and arcs are calculated. The line figure is divided at the feature point into a section that can be regarded as a typical line type, and the line figure is approximated by correcting the positional relationship of the feature points or the line type of the section, or the feature point or In the figure recognition device for recognizing the type of the line figure by calculating the consistency between the line type of the section and predetermined dictionary data, any two points in the point data are set as initial setting values, and A reference point storage unit that stores reference points in the order of connection, and two consecutive reference points stored in the reference point storage unit are reference points of interest, and k is a point sequence in a section sandwiched between the reference points of interest. Wear one by one at intervals And a separation elevation angle calculating unit that obtains a dispersion degree of an elevation angle formed by two line segments extending from the two attention reference points to the attention point, and a dispersion degree of the elevation angle output by the separation elevation angle calculation unit is a predetermined number. A reference point setting unit that generates a new reference point that divides a corresponding section when the threshold value is greater than 1 and stores the new reference point in the reference point storage unit. 2. The distance elevation angle calculating means sequentially sets k points of interest as point-of-interest points for point sequences excluding two points near the reference point of interest in the point sequence of the section sandwiched between the reference points of interest, and the two points The pattern recognition apparatus according to claim 1, wherein the degree of elevation angle dispersion made by two line segments from the reference point of interest to the target point is obtained. 3. The separation elevation angle calculating means increases or decreases the value of k when selecting the point of interest in accordance with the number of point sequences in the section sandwiched by the reference points of interest. The pattern recognition device according to items 1 or 2. 4. The separation elevation angle calculating means outputs point data of the point of interest that has the largest change in elevation angle in addition to the dispersion degree, and the reference point setting means generates a new reference point. 4. The figure recognition device according to claim 1, wherein the point data of the point of interest at which the change in elevation angle is the largest is generated as the new reference point when the condition is satisfied. 5. The separated elevation angle calculating means outputs a representative value of elevation angle in addition to the point data of the point of interest at which the degree of dispersion or the change in elevation angle is the largest, and the reference point storage means, In addition to the reference point, a curvature obtained from a representative value of the elevation angle corresponding to a section sandwiched between two consecutive reference points is stored, and the curvature corresponding to two sections sandwiching any reference point is stored. When the difference of is less than or equal to a predetermined second threshold value, the reference point is deleted, the two sections are integrated, and the curvature of the integrated section is recalculated from the curvatures of the original two sections. The pattern recognition device according to claim 1. 6. The figure recognition device according to claim 1, wherein the dispersion degree calculated by the separation elevation angle calculation means is dispersion, and the representative value is an average value.