JPS613287A - Graphic form input system - Google Patents

Abstract

PURPOSE:To input an accurate handwritten graphic form by extracting a combination of directions of strokes as a feature, and approximating a stroke as a straight line and an arc on the basis of respective sampling points and features at leading and trailing sampling points. CONSTITUTION:A sampling point extracting means 11 extracts (a) both end points (starting and ending points) of a stroke, (b) horizontal and vertical maximal and minimal points of the stroke, and (c) change points of the direction of the stroke as sampling points from the data sequence of a stroke segmenting means 9. The sampling point of (c) is extracted only when a horizontal or vertical line is present before and behind the sampling point. A feature extracting means 12 extracts a combination of directions of strokes before and after a sampling point as a feature according to a feature table. A stroke approximating means 13 approximates a stroke by a straight line and an arc from respective sampling points and features at leading and trailing sampling points.

Description

DETAILED DESCRIPTION OF THE INVENTION (Technical Field) The present invention relates to a figure input method for approximating a line L1 of a hand-drawn geometric figure by a straight line and a circular arc.

(Prior art) Conventionally, regarding the figure input method using the hand m8, there is a method that recognizes symbol figures by template matching, and a method that extracts features from the distribution of stroke changes and normalizes them to grid points and converts them into basic figures. There was a method of approximation.

The former is configured to perform recognition by matching each symbol shape with a standard pattern, so it is not suitable for inputting geometric shapes with 1,000 yen shapes that have many variations in size and aspect ratio. There was a drawback that there was no Further, in the latter method, since the C angle change distribution is used for handwritten strokes that vary widely, there is a drawback that detailed descriptions of figures cannot be accurately described. For example, in Figure 1 (
As shown in a), when simple figures 2 and 3 are written on the tablet surface on which grid points 1 are displayed in advance,
As shown in FIG. 1(b), figure 3 can be recognized as figure 4, but figure 2 does not have figure 5'c, and has the disadvantage of being mistakenly recognized as figure 4.

(Object of the Invention) An object of the present invention is to solve these drawbacks and to realize a figure input method that allows more accurate input of handwritten figures.

(Structure of the Invention) In order to achieve the above-mentioned object, the present invention includes a data reading means for reading handwritten coordinates of a node, a sampling means for sampling the read data, and a sampling means for sampling the read data. a stroke cutting means for cutting out a stroke having a start point and an end point from the stroke; and a sampling point extraction means for extracting both end points and horizontal and vertical maximum and minimum points of the cut stroke 1 cork as sampling points; Strike 1 issued] -
1 stage of sampling point extraction that extracts both end points of the curve, maximum points and minimum points in the horizontal and vertical directions, and points whose direction changes from horizontal or vertical as sampling points, and 1-stroke direction before and after the sampling point. The present invention is characterized by comprising a feature extracting means for extracting a combination of as features, and a means for approximating a stroke to a straight line or circular arc from the features of each sampling point and the sampling points before and after it.

(Example) FIGS. 2 to 11 show an embodiment of the present invention.

FIG. 2 shows the basic configuration of the present invention. In the figure, 6 is a preprocessing section consisting of a data reading means 7, a sampling means 8, and a stroke cutting means 9, and 10 is a sampling point extraction means 11 and a feature extraction means. Means 12 is a feature extraction section, and 13 is a strike 1 cork approximation section (means).

The data reading means 7 reads the handwritten figure data into the tablef 1.
It is read as a series of coordinate values according to the movement of the stylus pen etc. from an input device such as .

The sampling means 8 removes coordinate values that are extremely far from the trajectory of the figure due to hand shake during writing, quantization noise of the tablet, induced noise, etc. from the input data string as noise, and further removes coordinate values that are included due to fluctuations in writing speed. The coordinate values of adjacent redundant writing points are removed to create a data string of coordinate values separated by a predetermined distance.

Stroke 1 The cork cutting means 9 starts from the point where the pen contacts the tablet from the data string sampled by the sampling means 8.
Then, a stroke is cut out whose end point is the point where the pen is farther away from the tablet, and a data string of the coordinate values of one stroke is output.

→The non-bringing point extraction stage f 11 extracts from the data string output from the stroke extraction means 9 a) both end points of the stroke (start point, end point), b) maximum and minimum points in the horizontal and vertical directions in the stroke, and C) stroke. The point of change in the direction of is extracted as a sampling point. Note that the sampling point C) is extracted only when the point before or after the change point is a horizontal line or vertical line. FIG. 3 is a flowchart of the operation.

The feature extraction means 12 extracts the combination U of stroke directions before and after the sampling point as a feature according to a feature table as shown in FIG. This feature map is a combination of directions of strokes before and after the sampling point. The directions are based on 8 directions, but can be arbitrarily set to 4 directions, 16 directions, 32 directions, 64 directions, etc. depending on the approximation accuracy. Moreover, FIG. 5 is an operation flowchart.

The stroke approximation means 13 approximates a stroke to a straight line or circular arc based on the characteristics of each sampling point and the sampling points before and after it. FIG. 6 is a flowchart of the operation.

Next, its operation and effect will be explained in detail according to the drawings. First, the data of the handwritten figure is processed by the preprocessing unit 6.
receive immediate punishment. 15 in Figure 7(a), 1 in Figure 8(a)
6 is an example of cut out S1-〇-ri (
Here, X represents the writing point.

). Next, sampling points are extracted in the one non-ringing point extracting means 11, which will be explained in order below. Both end points of the stroke in a) have already been extracted by the stroke extraction means 9. 17.18 in FIG. 7(b) are the starting and ending points of the stroke 15, and the 8th
19°20 in the figure (b) are the starting and ending points of the stroke 16.

Extraction of maximum and minimum points in the horizontal and vertical directions is performed by sequentially comparing the coordinate values of the writing points from the starting point to the ending point. FIG. 9 shows an example of extraction processing using a mask. Pen point pn on the target stroke (n=1.2.3...
・) A square mask of 2D on each side is applied. 9th
In Figure (a), 21 represents a mask of pen point p0.22 represents pn. Regarding the intersections between the boundary line of the mask 22 and the stroke point, the intersection inputted temporally before the writing point pn is assumed to be p, and the intersection inputted after the writing point pn is assumed to be pn. In FIG. 9 αゝ(a), 23 represents p(x and 24 represents pn).

The horizontal coordinates at writing point p are X(p), jl

Let the coordinate value in the n vertical direction be Y(pn), dH,=X (p,) -X (p,)d)l =X
'(p)-X (pn)pn dV1=Y (pn)-Y (p(x)dV =Y(
p)-Y(pn) pn. At this point, if di-1>O and dl-(2>O), p is the maximum point in the horizontal force direction, and if dl-1<0 and d1-12<O, pn is the minimum point in the horizontal force direction. If dV > 0 and dV > O, then p is the maximum point in the vertical direction, and if dV < 0 and dV 2 < O, then p is the minimum point in the vertical force direction. 25 in ) are the maximum and minimum points in the direction perpendicular to the torque on the stroke, 26 are the maximum and minimum points in the horizontal direction, and 2 in FIG. 8(b)
7.28.29 is the vertical maximum point of stroke 16.
It is a minimum point.

If you try to approximate by extracting only the end points of the stroke and the maximum and minimum points in the horizontal and vertical directions as sampling points, you will notice that part of the stroke consists of horizontal line components or vertical line components as shown in 30. If so, it is not possible to approximate it accurately. For example, in the case of a rectangle, these are the four corners. Therefore, points of change in the stroke direction are also extracted as sampling points.

To identify the point of change in direction, for example, the method using a mask used for extracting maximum and minimum points in the horizontal and vertical directions is used. This will be explained with reference to FIG. 9(b).

Let sp be the stroke component within the mask of writing point p. At this time, if dH-0 and dH2=O, then Sp n is! I!
If the straight line component, dV -0th dv2=0, then t, r sp, is written as the horizontal line component. sp that becomes a vertical line component or a horizontal line component
A point p with . If there are two or more in a row like 31, the writing point 34 next to both ends 32.33 of those dots
, 35 are extracted as points of change in the direction of the stroke. Therefore, in stroke 15, writing points 36 and 37 are also extracted as sampling points.

Next, the feature extraction means 12 determines the stroke directions before and after the sampling point, and roughly determines the features from this. Here, sampling point "n (n=1.2.3・-
...-) before (D'), point P to find the h direction Cb(P) of the stroke. and point P
Find the slope of the straight line "1" that passes through the writing point B[)n that was input just before , that is, the ratio of the horizontal coordinate difference to the vertical coordinate difference, and from this determine which of the eight directions above should be approximated. To find the stroke direction Ca (P,) after point P, find the slope of the straight line that passes through point P and the writing point APn input immediately after point P, and from this, similarly decide.

Furthermore, the combination of stroke directions before and after the obtained sampling point is compared with the feature daple shown in FIG. 4, and the corresponding feature is determined. In FIG. 4, 38 is before the sampling point, and 39 is after lean pulling.

40 represents the direction of the stroke before and after the sampling point.

Further, 41 is a number given for convenience to represent the direction.

Figures 7(C) and 8(C) show examples of features determined at each sampling point, where 42.43 indicates the sign of each rimbling point, and 44.45 indicates the sign of each sampling point. 46.4 indicates the direction sequence of the determined stroke;
7 shows a feature number sequence for each numbering point according to the feature table of FIG.

Next, the stroke approximation means 13 approximates the stroke with a straight line or circular arc. To explain this, as shown in FIG. 10, for each sampling point of the stroke, from the start point to the end point, Pl, P2, P3, .・・・・・・・・・The name (=I [pu, also, the stroke component between Pl and 22 is S, the stroke component between P2.P3 is S2...
Name it... Also, 1=(P,) is the sampling point P, the direction of the previous stroke, and G(P) is P. In the direction of the later stroke.

For each sampling point except the end point, the stroke direction G (P) after the sampling point Pn in order from the starting point,
n before the next sampling point P
Direction of stroke of n+1 F(P)
The stroke component S between and is classified into circular arcs and center lines.

Specifically, as shown in Figure 11(a), G(P) and r) If F(P) does not match, it is assumed to be a circular arc.

As the method of approximation using n +1 circular arcs, a conventionally used means is applied. - As an example, Pn =] passes through n-11, P,
Apply a method of approximating it with an arc whose tangent has the same slope as the slope of the stroke after . In addition, in the case of coincidence and overlapping, the slope of the stroke after P is KP, the slope of the stroke before P is 1 I n -11, and the slope of the straight line passing through P and P is 3 and υ.
When n n+1, K 1. as shown in FIG. 11(b). = K 3 and K = K, then S is a straight 2 3 n n n+1 line segment passing through P and P ,
As shown in Fig. 11 (C), S is classified as two arcs connected in such a way that they touch 'I:l' at their connection point. In this case, As an example of an approximation method, an arc whose tangent slope is 1 through P nM J5 and an arc whose tangent slope is 2 through Pn-41 and Pn+1 are tangent to each other at their connection point. A means for calculating the centers and connection points of both circular arcs is applied as shown in FIG. 2.Furthermore, whether each sampling point has an angle is determined from the direction of the strokes before and after it.

Approximate examples are shown in Figure m7 (d) and Figure 8 (d).

49 is the approximation result g of stroke 1-15, and 50 is the approximation result of stroke 16.

(!R, effect of the Ming dynasty and Song dynasty) As explained above, according to the present invention, there are provided a data reading means for reading handwritten coordinates of points of contention, a sampling means for sampling the read data, and a sampling means for sampling the read data. A stroke cutting means cuts out the lines 1 to ○ to the right of the starting point and ending point from the converted data, and the both end points of the cut out stroke, the horizontal and vertical poles, the large fish and the minimum point are extracted. There is a feature extraction means for extracting a combination of pull stroke directions before and after the sampling point as a feature, and a feature extraction means for extracting a combination of pull stroke directions as a feature. Since the method for approximating the stroke to a straight line or circular arc is provided, any drawn figure can be input correctly, including its size, direction, curved part, and stroke connections. There are many different types of pulsating point extraction means for extracting both endpoints, maximum and minimum points in the horizontal and vertical directions, and points where the direction changes from horizontal or vertical as sampling points. In addition, since this method can accurately represent handwritten figures into geometric figures, it can be applied not only to the input of a single handwritten figure but also to feature extraction for handwritten figure recognition.

[Brief explanation of drawings]

The drawings serve to explain the present invention (", Figure 1 (a) (
b) is a diagram for explaining an example of erroneous recognition that occurs when the conventional handwritten figure input method is implemented; FIGS. 2 to 11 show an embodiment of the present invention; FIG. 3(a) to 3(C) are flowcharts of the sampling point extraction process, FIG. 4 is a diagram showing an example of the feature table input to the feature extraction means, and FIG. Figure (a) ~
(C) is a flowchart 1-1 for feature extraction processing; FIG. 6 is a flowchart 17-1 for stroke approximation processing;
(d), Figures 8 (a) to (d) are explanatory diagrams of specific examples of input processing, and Figure 9 (a) (1)) uses [Noruma x Suff] as the sampling point extraction means. l [Diagram for explaining an example of the extraction process, × Figure 10 is a diagram for explaining the stroke approximation means, × Figures 11 (a) to (C) explain the method for classifying arcs and straight lines.' This is a diagram for understanding. × 7...Data reading means, 8...41
Main forming means, × 9... Stroke cutting means, 11...
・Samp × Ring point extraction means, 12...Feature extraction means, × × Patent applicant Patent attorney representing Nippon Telegraph and Telephone Public Corporation 1) Takashi Takashi xxxx × × Senior × × Father × ×
KX KW Figure 6 JI7 Figure M8FI4 (C) (d) Figure 9 (G) j! Figure 10 n

Claims (2)

[Claims] (1) a data reading means for reading the coordinates of a pen point input by hand; a sampling means for sampling the read data;
A stroke extraction means that extracts a stroke having a start point and an end point from the sampled data, and a sampling point extraction means that extracts both end points and horizontal and vertical maximum and minimum points of the extracted stroke as sampling points. and,
It is characterized by comprising a feature extraction means for extracting a combination of stroke directions before and after the sampling point as a feature, and a means for approximating the stroke to a straight line or circular arc from the characteristics of each sampling point and the sampling points before and after the sampling point. Graphical input method. (2) a data reading means for reading the coordinates of a pen point input by hand; a sampling means for sampling the read data;
A stroke cutting means for cutting out a stroke having a start point and an end point from sampled data, both end points of the cut out stroke, maximum points and minimum points in the horizontal and vertical directions, and points where the direction changes from horizontal or vertical. sampling point extraction means for extracting as a sampling point; feature extraction means for extracting a combination of stroke directions before and after the sampling point as a feature; A graphic input method comprising: means for approximating a circular arc.