JP3088088B2 - Drawing symbol recognition method - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

[0001] 1. Field of the Invention [0002] The present invention relates to a method for automatically recognizing the meaning of a direction sign or the like written on a drawing from a handwritten or printed drawing input through an input unit such as an image scanner. It can be used as an automatic input tool for digitizing drawings (CAD input).

[0002]

2. Description of the Related Art Conventionally, manual input is generally used for CAD input of drawings, and almost no automatic input is performed. Even when automatic input is performed for a limited number of drawings, there are restrictions such as the need to write in a fixed manner and to write clearly.

[0003]

SUMMARY OF THE INVENTION The present invention has been made in view of the above-described circumstances, and solves the above-mentioned problem that there are many restrictions on drawing input even when the conventional drawing recognition technology is used. It is an object of the present invention to provide a drawing symbol recognition method for resolving symbol information written roughly by handwriting.

[0004]

SUMMARY OF THE INVENTION In order to achieve the above object, a drawing symbol recognition method of the present invention is a method for recognizing a direction symbol or the like drawn on a drawing such as a map or a plan view inputted from an input unit such as an image scanner. Thus, in a method of recognizing a sign having a meaning in a drawn direction, a step of extracting a feature of a portion having direction information from the sign and a step of patterning the extracted feature for database registration And a step of recognizing the direction of the symbol by comparing the database with the input pattern.

According to the present invention, in the step of extracting a feature of a drawing symbol having directional information in the drawing symbol recognition method, a pixel representing the symbol is extracted from a donut-shaped region which is separated from the center of the symbol by a predetermined value in a radiation direction. It is characterized by doing.

Further, the present invention is characterized in that a circle symbol or the like divided by several line segments passing through the center is printed in advance in a dropout color on a paper on which the symbol of the drawing symbol recognition method is drawn. I do.

The present invention also provides a method for patterning a drawing symbol having directional information according to the drawing symbol recognition method.
The method is characterized in that the pixels constituting the symbol are clustered based on the positional information, and the symbol is patterned based on the clustered pixel information.

That is, the present invention provides (1) a method of acquiring features of a portion which is hardly affected by distortion or the like from handwritten symbol data prepared in advance for learning;
(2) Use a method of converting acquired features into numerical patterns that are easy to compare.

[0009]

DESCRIPTION OF THE PREFERRED EMBODIMENTS Embodiments of the present invention will be described below in detail with reference to the drawings. INDUSTRIAL APPLICABILITY The drawing symbol recognition method according to the present invention can be used for automatically recognizing a symbol indicating a direction or the like, which is drawn in a plan view or the like indicating a layout of a general house.
The following is an explanation of an embodiment of the present invention applied to automatic recognition of a direction sign written on a plan view for architecture.

FIG. 1 shows a processing procedure of the present invention. In 10 drawing symbol registration processing, an angle table used in recognition processing is generated from learning data, and in 20 drawing symbol recognition processing, symbols are recognized by collating the angle table with input data.

In 10 drawing symbol registration processing, 11 learning data is input, and 14 angle tables are generated by 12 feature extraction processing and 13 pattern registration. 11 learning data is
As shown in FIG. 3, a symbol representing the azimuth is drawn on a circle symbol previously printed on a drawing sheet in a dropout color. FIGS. 4A, 4B, and 4C show three examples of how to write the direction symbol. Here, FIG.
The symbols shown in (b) and (c) all represent states in which the upper right direction is expressed. These symbols are read as 11 learning data from a data input device such as an image scanner.

In the 12 feature extraction process, black pixels are extracted from the 11 learning data read from the data input device, and are patterned as azimuth mark data from the angle information of the position. FIG. 2 shows this processing flow. First, 3
As shown in FIG. 5, the one-black-pixel extraction process extracts black pixels of a donut-shaped gray portion (represented by dots in the drawing) from the orientation symbol drawn on the drawing sheet. As shown in the partial enlarged view of FIG. 5, there is a noise removal (10 Dot) portion around the projection range (20 Dot). In FIG. 5, a portion where the azimuth mark and the oblique line overlap is extracted. Next, the angle from the center of the circle is calculated for the obtained black pixel as shown in FIG. 6 in the 32 angle detection processing. That is, if X and Y are the coordinates of the black pixel, θ is the angle, r is the radius, and Xr and Yr are the center coordinates of the circle, θ = tan ⁻¹ {(Y−Yr) / (X−Xr)} θ < In the case of 0, θ + = 360. Then, in the 33 clustering process, the angles obtained by the 32 angle detection process are allocated to, for example, 16 angle divisions of numbers (Number) 0 to 15 as shown in FIGS. 7A and 7B, and 8 (a),
As shown in (b), the number of detected black pixels is counted for each angle range of numbers 0 to 15. The example of FIGS. 8A and 8B is a case of 45 degrees. This is performed for each symbol data drawn in the drawing, and an angle range in which the number equal to or more than a certain value is counted is set to 1, and the other bit strings are set to 0 to form a pattern as 34 output data as 34 output data.

In registering 13 patterns, a 14-angle table is created based on the obtained patterns. The 14-angle table is created for collating with the azimuth mark data used for detecting the azimuth angle. The number of tables exists for the data of the divided patterns. As in this case, 16
If divided, 2 ¹⁶ = 65536. First, an angle table having a size of 65536 lines shown in FIG. 9 is prepared,
The initial values in the fixed range column are all set to -1. Next, azimuth marks, which are learning data, are sequentially registered. For example, for the azimuth mark in FIG. 8, azimuth mark data “001111100000100” is obtained by calculation, and the direction defined by the mark is defined in advance by the operator as “2”. When registering, enter “0011111000000010” on the angle table.
0 "is accessed, and" 2 "defined in advance is written as a fixed range. A numerical value (0
15) are written. As for the pattern of the learning data to be registered, data in all directions can be registered by generating data obtained by rotating one data by 360 degrees at a fixed frequency with respect to the center of the circle.

In 20 drawing symbol recognition processing, 21 evaluation data is input, and 24 recognition results are output by 22 feature extraction processing and 23 angle recognition processing. 21 evaluation data is 11
It is a direction symbol drawn in the same format as the learning data. This is read from a data input device such as an image scanner, and a 22 feature extraction process is performed. The 22-feature extraction process is almost the same as the 12-feature extraction process.
After the angle detection processing, processing for creating the following black pixel detection table is added. The black pixel detection table is a table that stores the number of black pixels counted for each detected angle. In the 23 angle recognition process, the pattern obtained by the 22 feature extraction process is compared with the 14 angle table. At this time, if the determined range is not registered (−1), it is output as an error. If not registered, the peak value of the black pixel is detected with reference to the black pixel detection table for the registered angle range, and this is set as the 24 recognition result. FIGS. 10A, 10B, and 10C show examples of this processing. As shown in FIG. 10A, when the direction mark data is “001111100000100”, the determined range of the direction mark is 2. Since this is the peak detection range of 33.75 to 56.25 degrees as shown in FIG. 10B, the black pixel detection table 34 to 56 degrees of the azimuth mark 2 as shown in FIG. The peak value is determined from the range. In this case, the peak is detected from 43 degrees to 47 degrees, and the black pixel is most detected at 45 degrees.
The degree is determined as the angle of the direction mark, and the recognition result is 45 degrees.

[0015]

As described above, according to the present invention, the time required for CAD input and the like can be reduced by automatically recognizing the azimuth symbol handwritten on a plan view or a map for construction.

[Brief description of the drawings]

FIG. 1 is a configuration explanatory diagram showing a processing flow according to an embodiment of the present invention.

FIG. 2 is a configuration explanatory diagram showing a feature extraction processing flow according to an embodiment of the present invention;

FIG. 3 is a configuration explanatory view showing display of an azimuth mark according to the embodiment of the present invention.

FIGS. 4A, 4B, and 4C are explanatory diagrams illustrating the format of an azimuth mark according to an embodiment of the present invention.

FIG. 5 is a configuration explanatory diagram showing a black pixel extraction range according to the embodiment of the present invention.

FIG. 6 is an explanatory diagram showing a method for calculating an angle according to the embodiment of the present invention.

FIG. 7 is a configuration explanatory view showing an azimuth mark table range according to the embodiment of the present invention.

FIG. 8 is an explanatory diagram showing creation of azimuth mark data according to the embodiment of the present invention.

FIG. 9 is a configuration explanatory view showing an angle table according to the embodiment of the present invention.

FIGS. 10A, 10B, and 10C are explanatory diagrams illustrating determination of an angle according to the embodiment of the present invention.

[Explanation of symbols]

 Reference Signs List 10 drawing symbol registration processing 11 learning data 12 feature extraction processing 13 pattern registration 14 angle table 20 drawing symbol recognition processing 21 evaluation data 22 feature extraction processing 23 angle recognition processing 24 recognition result 31 black pixel extraction processing 32 angle detection processing 33 clustering processing 34 Output data X, Y coordinates of black pixel θ angle r radius Xr, Yr center coordinates of circle

────────────────────────────────────────────────── ─── Continuation of the front page (56) References JP-A-4-263285 (JP, A) JP-A-6-176192 (JP, A) JP-A-5-81464 (JP, A) (58) Field (Int.Cl. ⁷ , DB name) G06T 7/60 G06F 17/50 G06K 9/00 G06T 1/00 G06T 7/00 JICST file (JOIS)

Claims (4)

(57) [Claims] 1. A method for recognizing a sign having a meaning in a drawn direction, such as a direction sign drawn on a drawing such as a map or a plan view input from an input unit such as an image scanner. Extracting a feature of a portion having direction information from among the above, patterning the extracted feature in a database, and recognizing the direction of the symbol by comparing the database with an input pattern. A drawing symbol recognition method comprising: 2. The method according to claim 1, wherein, in the step of extracting a feature of the drawing symbol having the direction information, a pixel representing the symbol is extracted from a donut-shaped region which is separated from the center of the symbol by a predetermined value in the radial direction. The method of recognizing the drawing symbols described. 3. The drawing according to claim 1, wherein, on a sheet for drawing a symbol, a circle symbol or the like divided in advance by several line segments passing through the center is printed in a dropout color in advance. Symbol recognition method. 4. A method for patterning a drawing symbol having directional information, comprising: clustering pixels constituting the symbol by its position information; and patterning the symbol based on the clustered pixel information. The drawing symbol recognition method according to claim 1, 2 or 3, wherein: