JPS63220370A - Automatic layer recognizing system for cad data based on knowledge processing - Google Patents

Abstract

PURPOSE:To reduce input cost and to speed up processing by sorting vector data on a screen by means of knowledge and constituting hierarchical data. CONSTITUTION:Vector data outputted from a graphic input device 2 are written in a vector information storing means 4 and a frame forming part 5 describes the properties of graphic elements constituting a picture or relation among the graphic elements as abstract data structure based upon a frame model on the basis of graphic constituting data based on vector expression. The frame data formed by the part 5 are written in a frame data storing means 7 and a knowledge base 8 stores plural IF-THEN type production rules describing restricting conditions inherent in the graphic. An estimation engine 9 executes the production rules and the abstract frame data stored in the means 7 are sorted based on the contents of the knowledge base 8 and the sorted result is converted into a CAD data format. Consequently, the multimedia formation of a data base can be easily executed.

Description

[Detailed Description of the Invention] [Summary] A map or the like is input to a drawing input device, the input map is expressed as a collection of vector information by the drawing input device, the vector information is input to a frame generation section, and the frame generation section generates The attributes of the figure to be recognized (area, coordinate values, etc.) and the relationships between objects (structural relationships between surfaces and lines, parallel relationships between lines, etc.) are described as an abstract data structure using a frame model.
Houses are included in blocks. '', ``Houses are independent and contain nothing inside,'' and ``There are roads between adjacent blocks.'' - It is configured to classify data and output CAD data of houses, CAD data of roads, CAD data of railways, etc.

[Industrial application field]

The present invention describes input drawings using an abstract data structure based on a frame model, and utilizes a knowledge base to create frame data that represents houses, frames data that represents roads, and frame data that represents railways. The present invention relates to a CAD data automatic layer recognition system using knowledge processing that is capable of finding frame data, etc., that are being used in a building, and outputting CAD data of houses, CAD data of roads, CAD data of railways, etc.

[Conventional technology]

In recent years, the development of multi-mezzo databases has rapidly progressed. Even for graphic data, traditional CAD/C
In addition to the AM field, local governments and niche companies have recently been actively creating map databases.

In conventional technology, when inputting houses, roads, railways, etc. into a computer, the shapes of the house, roads, and railways are input individually using a digitizer or the like.

[Problem to be solved]

In conventional technology, the initial input of drawing utilization systems relies on manual digitizing input, which has become a major bottleneck.

The present invention was created in view of this point, and uses knowledge to classify vector data of drawings, construct hierarchical data, and output CAD data using knowledge processing. It aims to provide an automatic layer recognition system.

[Means for solving problems]

FIG. 1 is a diagram illustrating the principle of the present invention. A drawing input device 2 converts image data of a drawing into vector data.

The vector data output from the drawing input device 2 is written into the vector information storage means 4. The frame generation unit 5 generates an abstract data structure based on a frame model based on the vector representation drawing structure data stored in the vector information storage means 4, and the attributes of the graphical elements constituting the drawing and the relationships between the graphical elements. Describe it as Frame generation section 5
The frame data generated by is written to the frame data storage means 7. The knowledge base 8 stores a plurality of IF-THEN type production rules that describe drawing-specific constraints. Inference engine 9 executes production rules. The abstract frame data stored in the frame data storage means 7 is classified based on the contents of the knowledge base 8, and the classification results are converted into a CAD data format.

〔Example〕

FIG. 2 is a functional block diagram of one embodiment of the present invention. In the figure, 1 is a calculator, 2 is a drawing input device, and 3 is a TS
S terminal, 4 is a vector data storage unit, 5 is a frame generation unit, 6 is a graphic operation library, 7 is a frame data storage unit, 8 is a knowledge base, 9 is an inference engine, 10 is a C
Each AD format converter is shown.

Computer 1 includes O8, TSS, vector data storage unit 4
The drawing input device 2, which includes a frame generation section 5, a graphical operation Lypra I 76, a frame data storage section 7, a knowledge base 8, an inference engine 9, a CAD data format conversion section 10, etc., inputs images of drawings. 0 drawing input device 2 that converts data into vector data
The vector data output from the vector data storage section 4 is stored in the vector data storage section 4. Vector data has start point information and end point information. The starting point information has an X coordinate, a Y coordinate, and end point information. The end point information indicates whether the vector is open or connected to another vector. The end point information has the same format as the start point information. The frame generation unit 5 generates frame data based on vector data stored in the vector data storage unit 4. The graphic operation library 6 stores programs for examining neighboring relationships, inclusion relationships, circumscribed relationships, inscribed relationships, intersection relationships, and partial relationships of graphics. Frame data is stored in frame data storage section 7. Knowledge base 8 includes the knowledge that ``A house is included in a city block.'' and the knowledge that ``A house is independent and does not contain anything inside.''
There are roads between adjacent blocks. ” is stored.

The knowledge stored in knowledge base 8 is
It is a rule type thing. The inference engine 9 checks whether the IF part of the production rule is true or not, and if it is true, T
II This is to execute the processing of the EN part. CA
The D data format conversion section 10 converts graphic data representing houses, graphic data representing roads, and graphic data representing railways into a form that can be understood by the CAD' system and outputs the data.

FIG. 3 is a diagram showing an example of a graphic model. One embodiment of the invention uses the following graphical models: lines, faces, trees, and complexes. A line is a set of vectors connected based on straightness,
A face is a collection of lines with a clockwise order and direction within a loop discovered by the left-handed algorithm, and a tree is a collection of lines that have no faces (however, when creating a virtual face, The direction may change), and a complex is a figure composed of a plurality of the above figure models.

FIG. 4 is a diagram showing an example of a frame logical structure.

In the same figure, for example, a frame called the structural map world has a slot called CLASS and a slot called PART, and the slot called CLASS has a slot called \VAL.
The slot has an attribute called UE and a value called GENERIC. A slot called PART has an attribute called ¥LINK, and has values called symbol string, independent figure, and map frame.

FIG. 5 is a diagram showing an example of a frame hierarchical structure.

In the figure, square frames represent prototype frames, and nine frames represent instance frames. As shown in the figure, the frame called the structural map world is linked to a prototype frame called a symbol string, a prototype frame called a map frame, and a prototype frame called an independent figure. A prototype called a symbol string.
Frames are linked to prototype frames called symbols. Further, a prototype frame called a symbol string has a symbol string 01 as an instance frame. The following is as shown.

The knowledge base 8 stores production rules. The production rule is (rule number (r
p condition part) (THEN result part). In the knowledge base 8, for example, (#1 (IF (inclusive block figure A) (= figure A "house"))) (#2 (IF (neighborhood block figure A) (- figure A block) ) are stored.

FIG. 6 is a diagram explaining the function of the recognition section. In the figure, reference numeral 11 indicates a meta controller, 12 a recognition processing unit based on geometric features, and 13 a recognition processing unit based on two-dimensional relationships. The recognition unit can be considered to be composed of a knowledge base 8 and an inference engine 9. The meta controller 11 performs general management of processing and management of opening and termination. The recognition processing unit 12 based on geometric features examines the shape of the figure and determines the recognition starting point. The two-dimensional relationship recognition processing unit 13 forms a hypothesis as to what the figure represents, and performs processing to verify the hypothesis.

FIG. 7 is a diagram showing an example of system processing. The drawing input device 2 reads drawings and outputs vector data. The frame generation unit 5 receives vector data as input and generates a frame. Frame generation includes initial processing, input processing, map frame processing, character/symbol processing, independent figure processing,
Structure recognition processing and frame output processing are performed in this order. In initial processing, various parameters are set up, in input processing, vector drawing data is read, in map frame processing, map frames are recognized, and in character/symbol processing, characters and symbols are recognized. In the independent figure processing, continuous lines are treated as a group of figures. In the structure recognition process, structure recognition, virtual processing, relational processing, etc. are performed. Structural recognition involves recognizing the structure of figures (for example, extracting loops), virtual processing involves virtually connecting figures that are open due to disconnections, etc., and relational processing involves recognizing the relationships between element figures (neighborhood, etc.). Inclusion, circumscription.

(inscribed, intersecting two parts, etc.) are extracted. In the frame output process, the frame graphic data processed by the recognition unit is output to the frame data storage unit 7.

After the frame data is generated, the meaning of the shape (house, road, etc.) is determined using the frame data using the knowledge base.
Recognize. The meta controller 11 controls geometric feature processing and two-dimensional feature processing, and manages the start and end of the processing. The general flow of the process is as follows.

1) Find the recognition starting point in the drawing by geometric feature recognition (a place that seems to be a house based on its shape).

2) Use the figure as a “fact” for generating a hypothesis.

3) Transfer control to two-dimensional relationship recognition and generate hypotheses based on "facts." The hypothesis is generated using the knowledge base 8, such as "If it is a house, it is included in the block."

4) If the reliability of the rules applied to the generated hypothesis is low, the generated hypothesis is checked for consistency with the surroundings through a verification process.

5) If verification fails, hypothesis generation is performed again.

6) If the verification is successful, the next hypothesis is generated using it as a "fact".

7) If a two-dimensional relationship is not established, control is transferred to geometric feature recognition.

8) Repeat the same process from 1) until the end.

The result of semantically classifying the drawings is a layered structure that is converted into target CAD data by the CAD data format conversion unit 10.

In the above embodiment, the figure is divided into five figure elements such as line, face, tree, complex, and symbol.
It can also be divided into two graphical elements. Additionally, six concepts such as neighborhood, inclusion, circumscription, inscription, intersection, and part can also be used as relationships between graphic elements. FIG. 8 is a diagram showing the relationships between the graphical elements that appear when the topographic map is divided into the graphical elements.

FIG. 9(a) is a diagram showing the inclusion relation extraction procedure.

■ The shape for which you want to find a relationship is called the target shape, and the object whose inclusion relationship is to be investigated is called a candidate shape.

(2) As shown in FIG. 9(b), a figure whose circumscribed rectangle is included in the circumscribed rectangle of the figure of interest is selected from the group of candidate figures.

■ Check whether all the vertices of the candidate figure are on the figure of interest. If so, it will be excluded.

(2) As shown in FIG. 9(C), a large vector is drawn horizontally from one end point of the candidate figure, and it is set as vo.

■ All lines and vectors that make up the figure of interest v
Check the intersection with 0 and count the number of intersections.

■ If the number of intersections is an odd number, it is determined that it is included, and if it is an even number, it is determined that it is not included. However, at the stage of cross checking, there are times when the value of the cross product becomes 0, as shown in Figure 9 (dl, (e)).In this case, when the value of the cross product becomes positive O, the number of times If the count is odd, add 1 to the number of crossings; if the count is even, leave the number of crossings as is.

FIG. 10+a) is a diagram showing the neighborhood relationship extraction procedure.

■ The shape for which you want to find a relationship is called the target shape, and the object whose neighborhood relationships are to be investigated is called a candidate shape.

■ Check whether the following conditions are satisfied for all the lines that make up the figure of interest and each line of the candidate figure. If the conditions are met, the relationship is a neighbor relationship.

1) At least one of the legs dropped from both end points of the candidate line to the line of interest is below the darkness value.

2) Linearity with the line of interest is below a threshold.

3) The candidate line whose parallel length to the line of interest is the largest among the candidate lines (see Figure 11).

FIG. 10(b) is a diagram illustrating calculation of the perpendicular length.

Here, the leg lengths hl and h2 of the perpendicular line are given by the following formulas (
XI-X2) ・(Y3-Yl)-(Y2-Yl)
(X3-XI) - FIG. 10(C) is a diagram illustrating calculation of parallel length.

Vector A and a point p that is not on the straight line on which vector A rides (
x3. Y3), and let the intersection of the perpendicular line drawn straight down from point P and the straight line be Q (X4. Y4). Here, the coordinates of point Q are given by the following equation.

X4. E/F Y4=G/F It is necessary to judge whether the intersection Q is a point on the vector A on a straight line. Similarly, for point R, the intersection point S with the straight line is determined.

When the relationship between the two intersection points and the vector A thus determined is sorted out, it is classified into five patterns as shown in FIG. Calculate the parallel length for each of them.

FIG. 12 (al is a diagram showing the intersection relation extraction procedure.

■ The figure whose intersection relationship you want to find is called the target figure, and the object whose intersection relationship is to be investigated is called the candidate figure.

(2) Obtain circumscribing rectangles for all the lines constituting the figure of interest and each line of the candidate figure, and perform intersection checking only on lines that overlap the circumscribed rectangles of the constituent lines of the figure of interest. In a case like the one shown in the 12th issue), a cross check is performed, and in a case like FIG. 12 (C), a cross check is not performed.

■ As shown in Figure 12(d), find the cross product of "a" Perform the following processing as an intersection candidate,
If it does not exist, it is determined that they do not intersect.

(2) As shown in FIG. 12(e), the same process as before is performed using the candidate line as a reference. Based on the obtained value, it is checked whether the end points p1 and p2 exist on both sides of the vector of interest. If they exist, it is determined that they intersect, and if they do not exist, it is determined that they do not intersect.

FIG. 12(f) is a diagram illustrating cross product calculation. The cross product can be found as follows.

ZL□ Xtt-X++ ZY+= Lz-Lt Z)h= lk ZXt
ZBz =ZL Umbrella ZYs -ZY+ * ZXsZB
, is larger than O and ZBg is smaller than 0, or vice versa, it is determined that the end points of the candidate line exist on both sides of the line of interest.

FIG. 13 (al is a diagram showing the inscribed relation extraction procedure.

■ The figure for which you want to find a relationship is called the target figure, and the object whose inscribed relationship is to be investigated is called a candidate figure.

(2) As shown in FIG. 13(b), a figure whose circumscribed rectangle is included in or overlaps with the circumscribed rectangle of the figure of interest is selected from the group of candidate figures.

■ Select candidate shapes with one or more vertices on the target shape.

■ Draw a large vector horizontally from one vertex of a candidate figure that is not on the target figure and set it as vo.

■゛All lines and vectors that make up the figure of interest■
. Check the intersections and count the number of intersections.

■ If the number of intersections is an odd number, it is determined that it is inscribed, and if the number of intersections is an even number, it is determined that it is not inscribed.

FIG. 14(a) is a diagram showing a circumscribed relationship extraction procedure.

■ The figure for which you want to find a circumscribed relationship is called the figure of interest, and the object whose circumscribed relationship is to be investigated is called a candidate figure.

(2) As shown in FIG. 14(b), check the circumscribed rectangle of the target figure and the circumscribed rectangle of the candidate figure, and select those that overlap or are in contact with each other.

■ For each vertex of the target figure and the candidate figure, 1) one or more of the vertices is on the candidate figure.

2) Other vertices exist outside the candidate figure.

Check whether the condition is satisfied, and if it is, it is determined that there is a circumscribed relationship.

FIG. 15(a) is a diagram showing a partial relationship extraction procedure.

■ The figure for which you want to find a partial relationship is called the target figure, and the object whose partial relationship is to be investigated is called a candidate figure.

■ As shown in FIG. 15 (bl), figures that are included in the circumscribed rectangle of the figure of interest or have the same circumscribed rectangle are selected from the group of candidate figures.

■ Check the relationship between all the lines composing the target figure and each line of the candidate figure, and if the relationship is other than intersecting, not intersecting, or touching, it is determined that there is a partial relationship.

〔Effect of the invention〕

As is clear from the above description, according to the present invention, it is possible to fill the gap between vector information from a drawing input device and data at the level of a usage system such as a CAD system. Further, according to the present invention, it is possible to achieve effects such as reducing input costs and speeding up processing, so it becomes easy to create a multi-mezzo layer of a database.

[Brief explanation of the drawing]

FIG. 1 is a diagram showing the principle of the present invention, FIG. 2 is a block diagram of an embodiment of the present invention, FIG. 3 is a diagram showing an example of a graphical model, FIG. 4 is a diagram showing an example of a frame logical structure, and FIG. Figure 5 is a diagram showing an example of the frame hierarchical structure, Figure 6 is a diagram explaining the function of the recognition unit, Figure 7 is a diagram showing an example of system processing, and Figure 8 is a diagram showing the relationship between elements in a topographic map. 9 is a diagram explaining the inclusion relationship extraction procedure, FIG. 10 is a diagram explaining the neighborhood relationship extraction procedure, FIG. 11 is a diagram showing the parallel length pattern,
Figure 2 is a diagram for explaining the intersecting relationship extraction procedure, Figure 13 is a diagram for explaining the inscribed relationship extraction procedure, Figure 14 is a diagram for explaining the circumscribed relationship extraction procedure, and Figure 15 is a diagram for explaining the partial relationship extraction procedure. It is a diagram. l...Calculator, 2...Drawing input device, 3...TS
S terminal, 4... Vector data storage unit, 5... Frame generation unit, 6... Graphic calculation library, 7...
Frame data storage unit, 8...Knowledge base, 9...
- Inference engine, IO...CAD format conversion unit. Unexploded Moe t Zetsukou 1'' ---1 Figure @ E-type H 7 - 4 ゜ V - row table start turn % // figure (11 shots) 1) ite, chestnut company 1 hand cough 13 jump 4 town days 8'l! If Deci7th Thousand Sansin) Iito 1D Butsuichi Yoshin No. 14-Figure

Claims (2)

[Claims] (1) A drawing input device (2) that converts image data of a drawing into vector data, a vector data storage means (4) that stores vector data output from the drawing input device (2), and a vector data storage means (4) that stores vector data output from the drawing input device (2).・A frame that describes the attributes of the graphical elements composing the drawing and the relationships between the graphical elements as an abstract data structure using a frame model based on the vector representation drawing configuration data stored in the data storage means (4) A generation unit (5), a frame data storage means (7) for storing frame data output from the frame generation unit (5), and an IF-THEN type production rule that describes drawing-specific constraints. A knowledge base that stores multiple information (8) and an inference engine that executes production rules (9)
and a frame using the contents of the knowledge base (8) above.
A CAD data automatic layer recognition system using knowledge processing, characterized in that it classifies abstract frame data stored in a data storage means (7) and converts the classification results into a CAD data format. (2) The frame data generation unit (5) classifies the vector representation drawing configuration data stored in the vector data storage unit (7) into four graphic elements: planes, lines, points, and symbols. Claim 1 is characterized in that the system is configured to extract relationships between graphical elements using six relationship concepts: , neighborhood, inclusion, circumscription, inscription, intersection, and part. CAD data automatic layer recognition system based on knowledge processing.