JP2021128485A - Target partial shape generation device from composite shape with multiple graphic elements - Google Patents

Abstract

To provide a device capable of more efficiently generating a complicated partial shape from a composite shape.SOLUTION: The device according to the present invention is a device for generating a target partial shape from a composite shape including a plurality of graphic elements. The device includes a representative determination unit that determines representative position information representing a position of a graphic element for each of a plurality of graphic elements based on graphic classification information corresponding to the plurality of graphic elements, an estimation unit that estimates the representative position information of the graphic element corresponding to a target partial shape by a prediction model generated by machine learning based on the composite shape for learning and the partial shape for learning using the representative position information of the plurality of graphic elements, and a shape generation unit that generates the target partial shape based on the estimated representative position information of the graphic element and the graphic classification information of the graphic element.SELECTED DRAWING: Figure 2

Description

The present invention relates to an apparatus that generates a target partial shape from a composite shape having a plurality of graphic elements.

In the fields of CAD (Computer Aided Design) and CG (Computer Graphics), a method of automatically extracting a specific partial shape from a composite shape including a plurality of partial shapes is known. For example, Patent Document 1 discloses a method of dividing a composite shape into each partial shape and extracting a partial shape having a feature amount similar to that of a reference shape registered in advance from each partial shape.

JP-A-2007-280129

However, since the method described in Patent Document 1 uses a reference shape registered in advance by the worker, there is a problem that it cannot be applied to a partial shape having a complicated shape that is difficult for the worker to predict. Further, in order to extract a specific partial shape from the composite shape, it is conceivable to use a prediction model generated by machine learning or the like, but when the partial shape to be extracted is complicated, the amount of information to be input to the prediction model. There is a problem that the calculation cost increases.

The present invention has been made to solve the above problems, and an object of the present invention is to provide an apparatus capable of more efficiently generating a partial shape to be extracted from a composite shape.

The device according to the present invention is a device that generates a target partial shape from a composite shape including a plurality of graphic elements. The apparatus includes a representative determination unit that determines representative position information representing the position of the graphic element for each of the plurality of graphic elements based on the graphic classification information corresponding to the plurality of graphic elements, and the plurality of the representative determination units. Using the representative position information of the graphic element of, the representative position information of the graphic element corresponding to the target partial shape is estimated by the prediction model generated by machine learning based on the composite shape for learning and the partial shape for learning. The estimation unit is provided, and a shape generation unit that generates the target partial shape based on the estimated representative position information of the graphic element and the graphic classification information of the graphic element.

In the present invention, since the target partial shape is estimated using the representative position information, it is possible to suppress an increase in the amount of input information and more efficiently generate the partial shape to be extracted from the composite shape.

INDUSTRIAL APPLICABILITY According to the present invention, it is possible to provide an apparatus capable of more efficiently generating a partial shape to be extracted from a composite shape.

It is the schematic which shows a part of the composite shape which concerns on Embodiment 1. FIG. It is a block diagram of the apparatus which concerns on Embodiment 1. FIG. It is a figure which shows an example of the data structure of the information output by the representative determination part which concerns on Embodiment 1. FIG. It is the schematic of the matrix diagram which concerns on Embodiment 1. FIG. It is a flowchart which shows an example of the target partial shape generation processing of the apparatus which concerns on Embodiment 1. FIG. It is a flowchart which shows an example of the learning data processing of the apparatus which concerns on Embodiment 1. FIG. It is a flowchart which shows an example of the prediction model determination process of the apparatus which concerns on Embodiment 1. FIG. It is a block diagram of the apparatus which concerns on Embodiment 2. FIG. It is a figure for demonstrating an example of the correction process of the apparatus which concerns on Embodiment 2. FIG. It is a block diagram of the apparatus which concerns on Embodiment 3. FIG. It is a schematic diagram of the value stored in the prediction matrix diagram about the predicted value of the representative position information and the predicted value of the duplicate value corresponding to the predicted value of the representative position information according to the third embodiment. It is a flowchart which shows an example of the target partial shape generation processing of the apparatus which concerns on Embodiment 3 and an example of learning data processing. It is a flowchart which shows an example of the duplication process of the apparatus which concerns on Embodiment 3. FIG. It is a block diagram of the computer which concerns on Embodiments 1-3.

Embodiment 1.
First, Embodiment 1 of the present invention will be described with reference to FIGS. 1 to 8. FIG. 1 is a schematic view showing a part of the composite shape 1 according to the first embodiment.
The composite shape 1 is a three-dimensional shape generated by a three-dimensional CAD system using shape data such as three-dimensional CAD data. As an example, the composite shape 1 is a three-dimensional shape representing a set of wall surfaces of an engine when the airflow analysis of oil agitation inside the engine is performed by CAE (Computer Aided Engineering). The composite shape 1 includes a plurality of partial shapes. In this figure, the composite shape 1 includes partial shapes 2, 5 and 7.

The partial shape is the wall surface of the parts constituting the composite shape 1. As an example, the partial shape is the wall surface of an engine component such as a cylinder head and a manifold. The partial shape has one or more graphic elements 3. In this figure, the partial shape 2 has graphic elements 4a, 4b, 4c, the partial shape 5 has a graphic element 6, and the partial shape 7 has a graphic element 8.

Each graphic element 3 has a graphic classification that classifies the shape of the graphic. In this figure, the graphic classification of the graphic elements 4a and 4c is a simple plane, the graphic classification of the graphic element 4b is a fillet, and the graphic classification of the graphic elements 6 and 8 is a cylindrical surface.
For example, when a plurality of graphic elements 3 are included in the partial shape as in the graphic elements 4a, 4b, and 4c included in the partial shape 2 of this figure, any of the graphic elements share the side or the surface. It may have a structure that connects to each other.

In the first embodiment, a specific partial shape (target partial shape) is extracted from the partial shapes constituting the composite shape 1, and shape data related to the target partial shape is generated. The target partial shape is a set of wall surfaces of a space that is an analysis target area. For example, when analyzing the airflow inside the space formed by the composite shape 1, the target partial shape is a shape including only the wall surface of the partial shape which is the boundary surface with the fluid among the partial shapes constituting the composite shape 1. .. For example, in the first embodiment, the partial shapes 2 and 6 are extracted from the composite shape 1 as the target partial shape, and the shape data related to the target partial shape is generated.

Next, the apparatus 10 used for generating shape data relating to such a target partial shape will be described with reference to FIGS. 2 to 7. FIG. 2 is a configuration diagram of the device 10 according to the first embodiment. The device 10 generates a target partial shape to be extracted from the composite shape 1. The device 10 is a personal computer, a notebook computer, a smartphone or other device capable of input / output. The device 10 includes a storage unit 100, an acquisition unit 110, a conversion processing unit 120, an estimation unit 140, a shape generation unit 160, a prediction model generation unit 180, and a learning database 190.

The storage unit 100 is a storage medium that stores various information such as shape data related to the composite shape 1. The storage unit 100 is connected to the acquisition unit 110 and provides the acquisition unit 110 with various information such as shape data related to the composite shape 1. Here, the shape data includes, for each graphic element 3, information indicating the graphic classification and the positions of the vertices forming the graphic element 3. In the first embodiment, the graphic classification is, for example, a simple plane, a fillet, a cylindrical surface, a free curved surface, or the like. _{Further, the information indicating the position of the apex in the first embodiment is the position coordinates (x 0} , y ₀ , z ₀ ) of the apex in the space using the X-axis, the Y-axis, and the Z-axis. The format of the shape data is, for example, STEP (Standard for the Exchange model data), IGES (Initial Graphics Expert Specialization), DWG (DRAWING), etc., and DX (DRAWING) and DX.

The acquisition unit 110 acquires various information such as shape data related to the composite shape 1 from the storage unit 100. The acquisition unit 110 is connected to the conversion processing unit 120, and outputs various information such as shape data related to the composite shape 1 to the conversion processing unit 120.

The conversion processing unit 120 is connected to the estimation unit 140 and performs conversion processing from the shape data related to the composite shape 1 to the input data of the prediction model used by the estimation unit 140. The conversion processing unit 120 has a representative determination unit 122 and a mapping processing unit 124.

The representative determination unit 122 determines representative position information representing the position of the graphic element 3 for each of the plurality of graphic elements 3 based on the shape data. The representative position information will be described later. The representative determination unit 122 is connected to the mapping processing unit 124, and outputs the representative position information to the mapping processing unit 124 together with the graphic classification information.

The mapping processing unit 124 maps the representative position information of the plurality of graphic elements 3 or the graphic classification information of the plurality of graphic elements 3 to one or a plurality of matrix diagrams. The matrix diagram will be described later. The mapping processing unit 124 is connected to the estimation unit 140 and outputs a matrix diagram to the estimation unit 140.

The estimation unit 140 uses a matrix diagram including representative position information and graphic classification information of a plurality of graphic elements 3 of the composite shape 1, and uses a prediction model to obtain representative position information of the graphic element 3 corresponding to the target partial shape and its graphic. Estimate classification information. Here, the prediction model includes, for example, a neural network. In the first embodiment, the neural network is a convolutional neural network (CNN). In the first embodiment, the prediction model has a number of prediction models according to the type of matrix diagram. The estimation unit 140 is connected to the shape generation unit 160, and outputs the representative position information of the graphic element 3 corresponding to the target partial shape and the graphic classification information thereof to the shape generation unit 160.

The shape generation unit 160 generates shape data of the target partial shape based on the estimated representative position information of the graphic element 3 and the graphic classification information of the graphic element 3.

The prediction model generation unit 180 generates a prediction model by machine learning based on the composite shape for learning and the target partial shape for learning. Further, the prediction model generation unit 180 is connected to the conversion processing unit 120 and the estimation unit 140, and inputs and outputs various information between them. The prediction model generation unit 180 includes a learning processing unit 182 and a prediction model determination unit 184.

The learning processing unit 182 manages the learning data of the prediction model. The learning processing unit 182 is connected to the learning database 190, the conversion processing unit 120, and the prediction model determination unit 184. The learning processing unit 182 requests the conversion processing unit 120 to convert the shape data of the composite shape for learning and the target partial shape for learning acquired from the learning database 190 into the learning data for the prediction model. Then, the learning processing unit 182 outputs the learning data output from the conversion processing unit 120 to the prediction model determination unit 184.
In the first embodiment, the learning data is one or more generated based on the representative position information or the graphic classification information acquired from each shape data for each of the compound shape for learning and the target partial shape for learning. It is a matrix diagram.

The prediction model determination unit 184 determines the prediction model by machine learning using the training data. The prediction model determination unit 184 is connected to the estimation unit 140 and outputs the determined prediction model to the estimation unit 140.

The learning database 190 stores various information such as shape data of a composite shape for learning and a target partial shape for learning, and learning data provided by the learning processing unit 182 corresponding thereto.

FIG. 3 is a diagram showing an example of a data structure of information output by the representative determination unit 122 according to the first embodiment. The information includes a graphic element identification number, graphic classification information K, and representative position information R.

The graphic element identification number is a number that uniquely identifies each graphic element 3. In the first embodiment, the graphic element identification number is assigned by the representative determination unit 122 based on the shape data.
The graphic classification information is information indicating the graphic classification of the graphic element 3. The graphic classification information is a number assigned based on the graphic element 3 included in the shape data. In the first embodiment, when the types of graphic classification are N (N is a natural number), the graphic classification information is represented by K = kn (k is a natural number less than or equal to N, and n is a constant of an arbitrary natural number). .. In this figure, n = 10.
For example, when the graphic element 3 is a simple plane, the value K of the graphic classification information is 10 (k = 1), and when the graphic element 3 is a fillet, the value K of the graphic classification information is 20 (k = 2). When the graphic element 3 is a cylindrical surface, the value K of the graphic classification information is 30 (k = 3).

The representative position information is the coordinates obtained by standardizing the position coordinates of the representative points and discretizing them, and is represented by R = (R _x , R _y , R _z ). Here, the representative point is a point representing the position of the corresponding graphic element 3. The position coordinates (x ₁ , y ₁ , z ₁ ) of the representative point are determined based on the figure classification information K and the position coordinates (x ₀ , y ₀ , z ₀ ) of the vertices of the corresponding graphic element 3.
For example, when the graphic element 3 is a simple plane or a fillet, the representative point may be the center of gravity of the simple plane. As an example, when the graphic element 3 is a simple plane formed by four corners (vertices), each component of the position coordinate of the representative point is the number of vertices (here) by adding the four vertices for each component. Then, it may be a value divided by 4). Further, for example, when the graphic element 3 is a cylindrical surface, the representative point may be a midpoint on the central axis of the cylindrical surface. Instead, the representative point may be any one of the plurality of vertices forming the graphic element 3.

_{As shown in this figure, the representative position information has components R x} , R _y , and R _z in the X-axis, Y-axis, and Z-axis directions. In this figure, the graphic element of the graphic element identification number 1 has the graphic classification information 10 and the representative position information (1, 2, 2), and the graphic element of the graphic element identification number 2 has the graphic classification information 20 and the representative position. It has information (1, 2, 1), and the graphic element of the graphic element identification number 3 has graphic classification information 10 and representative position information (5, 5, 3).
In the first embodiment, details of standardization and discretization will be described later.
By converting the graphic element 3 of the composite shape 1 into the representative position information in this way, the efficiency of the estimation process using the prediction model described later is improved.

FIG. 4 is a schematic diagram of a matrix diagram according to the first embodiment. FIG. 4A shows a value assigned to each pixel of the matrix diagram, and FIG. 4B shows an imaged matrix diagram based on the value of each pixel. In the first embodiment, the matrix diagram has the first to third matrix diagrams 50 and the fourth to sixth matrix diagrams 52.

As shown in FIGS. 4A and 4B, each matrix diagram has pixels specified using rows and columns. In this figure, the number of rows and columns of each matrix diagram is S, and the number of pixels of one matrix diagram is S × S. Here, the number of pixels indicates the resolution of the matrix diagram. That is, in this figure, the matrix diagram has a resolution of S × S.

Here, the row and column values correspond to _{the values R 1} and R ₂ of the coordinate components in the first and second directions of the representative position information R of the graphic element 3, respectively. For example, the first and second directions are a pair of axial directions selected from the X, Y, and Z axis directions. In the first embodiment, the first and second directions of the first and fourth matrix diagrams 50a and 52a are the X-axis direction and the Y-axis direction, and the first and second directions of the second and fifth matrix diagrams 50b and 52b. The two directions are the Y-axis direction and the Z-axis direction, and the first and second directions in the third and sixth matrix diagrams 50c and 52c are the Z-axis direction and the X-axis direction. That is, the first and fourth matrix Figure 50a, _R 1, _{R 2} of 52a _is, R x, a _{R y,} second and fifth matrix Figure 50b, _R 1, _{R 2} and 52b _{is, R} y, R _{z , and R 1} and R ₂ in the third and sixth matrix diagrams 50c and 52c are R _z and R _x .
Then, for each graphic element 3, the positions of the pixels in each matrix diagram are assigned based on _{the coordinate components R 1} and R ₂ in the first and second directions of the representative position information R.

Next, the pixel values shown in FIG. 4A will be described. In the first embodiment, the definition of the pixel value is different between the first to third matrix diagrams 50 and the fourth to sixth matrix diagrams 52.
The value of the first to third matrix Figure 50 pixels corresponds to the sum of the values R ₃ of coordinate components in the third direction of the representative position information R of all the graphic elements 3 which is assigned to the position of that pixel. In the first embodiment, the third directions of the first, second and third matrix diagrams 50a, 50b and 50c are the Z, X and Y axis directions, respectively. _{That is, the values of the coordinate component R 3 in} the third direction of the representative position information R of the first, second, and third matrix diagrams 50a, 50b, and 50c are R _z , R _x , and R _y , respectively.
The values of the pixels in FIGS. 52 of the 4th to 6th matrices correspond to the sum of the values K of the graphic classification information of all the graphic elements assigned to the positions of the pixels.

Here, a specific example of the value of each pixel in FIG. 4A will be described using the information shown in FIG. 3 as an example. For example, the positions of the pixels in the second row and the first column correspond to the graphic element identification numbers 1 and 2 in which (x, y) = (1, 2). Therefore, the value of the pixel in the second row and the first column of the first matrix diagram 50a is 3 (= 2 + 1), which is the sum of _{the Z-axis coordinates Rz of the representative position information of the graphic element identification numbers 1 and 2.} Further, the value of the pixel in the second row and the first column of the fourth matrix FIG. 52a is 30 (= 10 + 20), which is the sum of the values K of the graphic classification information of the graphic element identification numbers 1 and 2.
On the other hand, the positions of the pixels in the 5th row and 5th column correspond only to the graphic element identification number 3 in which (x, y) = (5, 5). Therefore, the value of the pixel in the 5th row and 5th column of the first matrix diagram 50a is 3, which is _{the Z-axis coordinate Rz of the representative position information of the graphic element identification number 3.} Further, the value of the pixel in the 5th row and 5th column of the fourth matrix diagram 52a is 10, which is the value K of the graphic classification information of the graphic element identification number 3.

Next, FIG. 4B will be described. As shown in FIG. 4B, the matrix diagram is a diagram in which the values stored in each pixel are converted into color brightness (contrast) and imaged. For example, the color of a pixel may have a higher brightness as the value of the pixel is larger, and may have a lower brightness as the value of the pixel is smaller. In the first embodiment, the lightness is used to image FIG. 4A, but the present invention is not limited to this, and hue, saturation, and the like may be used.
By generating the matrix diagram imaged in this way, the efficiency of the estimation process by the prediction model using the CNN described later is improved.

FIG. 5 is a flowchart showing an example of the target partial shape generation process of the device 10 according to the first embodiment.
First, in S10, the acquisition unit 110 acquires the shape data related to the composite shape 1 from the storage unit 100, and outputs the shape data to the representative determination unit 122 of the conversion processing unit 120. At this time, the acquisition unit 110 also acquires a predetermined resolution S × S value from the storage unit 100 and outputs the value to the representative determination unit 122 of the conversion processing unit 120.

Next, in S11, the representative determination unit 122 determines each graphic element 3 based on the graphic classification information K of each graphic element 3 included in the acquired shape data and the position coordinates (x ₀ , y ₀ , z ₀ ) of the vertices. The position coordinates of the representative point (x ₁ , y ₁ , z ₁ ) are determined, and the representative position information R = (R _x , R _y , R _z ) is determined based on the position coordinates of the representative point. At this time, the representative determination unit 122 _{standardizes and discretizes each component of the position coordinates (x 1} , y ₁ , z ₁ ) of the representative point based on the resolution S × S, so that the representative position information R = ( Each component of R _x , R _y , R _z ) is determined.
_{For example, the value of the coordinate x 1 in} the X-axis direction of the representative point is standardized as follows.
x'= S × (x ₁ −X _{0 min} ) / (X _0max −X _{0 min} )
Here, X _0min is the minimum value of the acquired vertices of the graphic elements 3 included in the shape data X-axis direction of the coordinates (x _0), X _0max is the graphical elements 3 included in the acquired shape data It is the maximum value of _{the coordinates (x 0} ) of the vertices in the X-axis direction. S is the square root value of the resolution. As a result, x'becomes a value less than or equal to S.

Further, for example, discretization, 'for, m-1 <x' x normalized to _{x 1} (m is a natural number equal to or less than S) ≦ m if it is _to convert the _R x = m. As a result, R _x becomes a natural number less than or equal to S.

In the first embodiment, the representative determination unit 122 calculates the position coordinates (x ₁ , y ₁ , z _{1 ) of the representative point from the position coordinates (x 0} , y ₀ , z ₀ ) of the apex, and (x _1). , Y ₁ , z ₁ ) were standardized and discriminated to calculate the representative position information R = (R _x , R _y , R _z). However, the present invention is not limited thereto, the representative determining unit 122, the vertex coordinates _{(x 0, y 0, z} 0) after normalization, the position coordinates of the representative point of the graphic element in the normalized spatial _(x 1, _{y 1} , Z ₁ ) may be calculated, and this may be differentiated to calculate the representative position information R = (R _x , R _y , R _z ).

Then, the representative determination unit 122 outputs the calculated representative position information R = (R _x , R _y , R _z ) to the mapping processing unit 124 together with the graphic classification information.

Next, in S12, the mapping processing unit 124 generates a matrix diagram having rows and columns corresponding to _{the coordinate components R 1} and R _{2 in the first and second directions of the representative position information, respectively.} In the first embodiment, the mapping processing unit 124 generates the first to third matrix FIGS. 50. For example, the mapping processing unit 124 assigns pixel positions corresponding to _{the coordinates R 1} and R ₂ in the first and second directions of the corresponding representative position information to each of the plurality of graphic elements 3. Then, the mapping processing unit 124 takes the sum _{of the coordinate values R 3 in} the third direction of the assigned graphic element 3 for each pixel. The brightness of the pixels is determined according to the sum, and the first to third matrix diagrams 50 of the matrix diagram are generated.
Further, the mapping processing unit 124 generates the fourth to sixth matrix FIGS. 52. For example, the mapping processing unit 124 takes the sum of the values K of the graphic classification information of the assigned graphic element 3 for each pixel. The mapping processing unit 124 determines the brightness of the pixels according to the sum thereof, and generates the fourth to sixth matrix diagrams 52 of the matrix diagram.
In this way, the mapping processing unit 124 generates six matrix diagrams. Then, the mapping processing unit 124 outputs the six matrix diagrams to the estimation unit 140.

In S13, the estimation unit 140 acquires a prediction model from the prediction model determination unit 184. In the first embodiment, the estimation unit 140 acquires one prediction model (CNN) for each matrix diagram. That is, the estimation unit 140 determines the first to third prediction models corresponding to the first to third matrix FIGS. 50 and the fourth to sixth prediction models corresponding to the fourth to sixth matrix FIGS. 52. Obtained from 184.

Next, in S14, the estimation unit 140 uses a matrix diagram based on the representative position information R and the graphic classification information K of the plurality of graphic elements 3, and the representative position of the graphic element 3 corresponding to the target partial shape by the prediction model. The information R'and the graphic classification information K'are estimated. For example, the estimation unit 140 acquires the value of each pixel from the brightness of each pixel of each matrix diagram, inputs the value of each pixel into the input layer of the corresponding prediction model, and predicts each pixel from the output layer of the prediction model. Get the value. In the first embodiment, the estimation unit 140 performs this process on the first to sixth matrix diagrams. As a result, the estimation unit 140 obtains the first to sixth prediction matrix diagrams corresponding to the first to sixth matrix diagrams. That is, the estimation unit 140 includes the predicted value R'of the representative position information of the graphic element corresponding to the target partial shape and the predicted value K'of the graphic classification information using the values of each pixel in the first to sixth matrix diagrams. The first to sixth prediction matrix diagrams are acquired.
Then, the estimation unit 140 outputs these prediction matrix diagrams as shape generation data to the shape generation unit 160.

In S15, the shape generation unit 160 generates shape data of the target partial shape based on the acquired shape generation data. The shape generation unit 160 generates a prediction matrix diagram of the shape to be deleted based on the shape generation data, and based on the prediction matrix diagram, generates the shape data of the target partial shape from the shape data of the composite shape 1. It may be extracted. At this time, the shape generation unit 160 takes a difference between the matrix diagram used for inputting the prediction model and the prediction matrix diagram output from the prediction model included in the shape generation data, so that the shape generation unit 160 has a prediction matrix of the shape to be deleted. You may get the figure. Then, the shape generation unit 160 may generate shape data of the shape to be deleted based on the prediction matrix diagram of the shape to be deleted, and subtract the shape data of the shape to be deleted from the shape data of the composite shape 1. .. However, the present invention is not limited to this, and the shape generation unit 160 may extract the shape data of the target partial shape from the shape data of the composite shape 1 based on the prediction matrix diagram of the shape to be deleted by any method. As a result, the shape data of the target partial shape is obtained by removing only unnecessary parts from the shape data of the original composite shape 1 while retaining the shape data, so that the target partial shape can be extracted more accurately. Can be done.

As described above, according to the first embodiment, the device 10 uses the prediction model based on machine learning (particularly deep learning), and the graphic element 3 corresponding to the target partial shape from the representative position information R of the plurality of graphic elements 3. Estimate the representative position coordinate R'of. Therefore, the simplified data is the input data of the prediction model, not the shape data of the composite shape 1. As a result, the calculation cost can be suppressed while maintaining the accuracy of the prediction model. Therefore, it is possible to more efficiently generate a complicated target partial shape from the composite shape 1.

Further, according to the first embodiment, the input data of the prediction model includes the graphic classification information K of the plurality of graphic elements 3 in addition to the representative position information R of the plurality of graphic elements 3. Therefore, the prediction accuracy of the shape data of the target partial shape can be further improved.

Further, according to the first embodiment, the apparatus 10 generates a matrix diagram having a good affinity with the CNN which is a prediction model based on the representative position information R and the graphic classification information K of the plurality of graphic elements 3, and this is used as a CNN. Used as input data for. Therefore, it is possible to more efficiently generate a complicated target partial shape from the composite shape 1.

In the first embodiment, the input data of the prediction model is a matrix diagram, but is not limited to this, and includes the representative position information R generated by the representative determination unit 122 and the corresponding graphic classification information K. It may be text data. In this case, the process of determining the lightness in S12 and the process of acquiring the pixel value from the lightness in S14 are omitted.

Further, in the first embodiment, in S13, the estimation unit 140 acquires six prediction models corresponding to the first to sixth matrix diagrams from the prediction model determination unit 184. However, the present invention is not limited to this, and for example, the estimation unit 140 may acquire one prediction model corresponding to all of the first to sixth matrix diagrams, and the prediction model corresponding to the first to third matrix diagrams and the prediction model. Two prediction models corresponding to the 4th to 6th matrix diagrams may be acquired.

Next, the process for generating the prediction model will be described with reference to FIGS. 6 to 7. FIG. 6 is a flowchart showing an example of learning data processing of the apparatus 10 according to the first embodiment.
First, in S20, the learning processing unit 182 acquires shape data of the composite shape for learning and the target partial shape for learning from the learning database 190. Then, the learning processing unit 182 outputs the shape data acquired by the representative determination unit 122 to the representative determination unit 122.

In S21, the representative determination unit 122 determines the representative position information R of the graphic element 3 by performing the same processing as the processing shown in S11 of FIG. 5 on the acquired shape data. The representative determination unit 122 outputs the representative position information R of the graphic element 3 determined by the mapping processing unit 124 together with the graphic classification information K.

In S22, the mapping processing unit 124 generates 6 types of matrix diagrams for each shape data by performing the same processing as in S12 shown in FIG. 5 on the acquired representative position information R and graphic classification information K. .. The mapping processing unit 124 outputs these matrix diagrams to the learning processing unit 182.

In S23, the learning processing unit 182 stores the acquired matrix diagram as the learning data of the prediction model in the record of the corresponding shape data of the learning database 190. When performing cross-validation at the time of predictive model determination processing, the learning processing unit 182 may classify the learning data into training data and test data. The learning processing unit 182 may determine whether the acquired matrix diagram is training data or test data, and store the acquired matrix diagram together with the classification in the learning database 190. The matrix diagram corresponding to the composite shape for learning is the input data input to the prediction model. Further, the matrix diagram corresponding to the target partial shape for learning becomes the teacher data corresponding to the input data.

As described above, according to the first embodiment, since the apparatus 10 stores the learning data of the prediction model in the learning database 190 in advance, the learning data can be immediately acquired in the subsequent prediction model determination process.

FIG. 7 is a flowchart showing an example of the prediction model determination process of the apparatus 10 according to the first embodiment.
First, in S30, the learning processing unit 182 acquires training input data regarding the composite shape for learning and training teacher data regarding the target partial shape for learning as training data from the learning database 190. Further, the learning processing unit 182 acquires test input data regarding the composite shape for learning and test teacher data regarding the target partial shape for learning as test data from the learning database 190. The learning processing unit 182 outputs various data to the prediction model determination unit 184.

Next, in S31, the prediction model determination unit 184 selects one of the six matrix diagrams of the training input data and acquires the prediction model corresponding to the selected matrix diagram type.
Next, in S32, the prediction model determination unit 184 acquires various parameters of the prediction model acquired in S31.
Then, in S33, the prediction model determination unit 184 acquires the pixel value from the brightness of the pixel of the selected matrix diagram, inputs the value to the input layer of the prediction model, and inputs the prediction value of the pixel from the output layer of the prediction model. get.

In S34, the prediction model determination unit 184 acquires the pixel value from the pixel brightness of the matrix diagram in the training teacher data corresponding to the selected matrix diagram, and calculates the error from the prediction value.
In S35, the prediction model determination unit 184 determines whether or not the error is less than the threshold value. If the prediction model determination unit 184 is less than the threshold value (Y in S35), the process proceeds to S37, and if not (N in S35), the process proceeds to S36.

In S36, the prediction model determination unit 184 updates various parameters of the prediction model based on the error. At this time, the prediction model determination unit 184 may update various parameters by using, for example, the backpropagation method. Then, the prediction model determination unit 184 returns the process to S32.

In S37, the prediction model determination unit 184 performs the same processing as the processing of S31 to S34 for the test data, and calculates an error between the prediction value and the pixel value of the test teacher data.
In S38, the prediction model determination unit 184 determines whether or not the error is less than the threshold value. If the prediction model determination unit 184 is less than the threshold value (Y in S38), the process proceeds to S40, and if not (N in S38), the process proceeds to S39.

In S39, the prediction model determination unit 184 updates the prediction model based on the error. The prediction model determination unit 184 returns the process to S31.

In S40, the prediction model determination unit 184 determines the prediction model corresponding to the selected matrix diagram. Then, the prediction model determination unit 184 performs the same processing as in S31 to S40 for the remaining matrix diagrams, and determines all the prediction models of the prediction model.

As described above, according to the first embodiment, the apparatus 10 can generate a prediction model for predicting the matrix diagram for the target partial shape from the matrix diagram for the composite shape by using the learning data.

In the first embodiment, the learning database 190 does not have to be included in the device 10. For example, the learning database 190 may be included in an external device (not shown). In this case, the prediction model generation unit 180 may be connected to an external device via a communication means (not shown) to send and receive information stored in the learning database 190 to and from the external device.

Embodiment 2.
Next, Embodiment 2 of the present invention will be described with reference to FIGS. 8 to 9. The second embodiment is characterized in that correction processing is performed on the predicted value of the representative position information generated by the estimation unit 140.
FIG. 8 is a configuration diagram of the device 20 according to the second embodiment. The device 20 basically has the same configuration and function as the device 10 shown in FIG. 2 of the first embodiment, but is different in that it includes a correction processing unit 250 in addition to the same structure and function. In FIG. 8, the same parts as those in FIG. 2 are designated by the same reference numerals, and the description thereof will be omitted as appropriate.

The correction processing unit 250 sets the value assigned to the pixel corresponding to the representative position information of the plurality of graphic elements 3 of the composite shape 1 and the pixel corresponding to the representative position information R of the graphic element 3 corresponding to the estimated target partial shape. The difference from the assigned value (predicted value R') is taken, and the predicted value R'of the representative position information of the graphic element 3 corresponding to the target partial shape estimated based on the difference is corrected. That is, the correction processing unit 250 acquires the input data R of the prediction model and the shape generation data including the prediction value R'which is the output data of the prediction model from the estimation unit 140, and the input data R and the prediction value R' The shape generation data is corrected based on the comparison with. Then, the correction processing unit 250 outputs the corrected shape generation data to the shape generation unit 160.

FIG. 9 is a diagram for explaining an example of the correction process of the apparatus 20 according to the second embodiment.
First, the correction processing unit 250 classifies each pixel of the prediction matrix diagram output from the estimation unit 140 into groups A to D.
(Group A)
For example, as shown in FIG. 9A, when the input value R of a pixel is 0 and the predicted value R'is not 0, that is, when R = 0 and R-R'<0, the pixel is classified into group A. .. The pixels of group A are pixels that are predicted to contain graphic element 3 even though they originally did not contain graphic element 3, and there is a possibility that misalignment may have occurred in the process of prediction, so-called. It is a "misaligned pixel".
(Group B)
Further, when the input value R of the pixel is not 0 and the predicted value R'has a value different from R, and when R ≠ 0 and | R-R'|> 0, the pixel is classified into the B group. Pixels in Group B include pixels that were successfully deleted (ie, were not extracted) or were accidentally deleted due to misalignment during the prediction process. Therefore, the pixels of group B are so-called "deleted successful pixels" or "misaligned source pixels".
(Group C)
Further, when the input value R of the pixel is not 0, the predicted value R'is not 0, and R'and R have the same value, that is, R ≠ 0 and R'≠ 0 and R-R'=. If it is 0, the pixel is classified into C group. Pixels in group C are "extracted successful pixels" that have been successfully extracted.
(Group D)
Finally, the pixels other than the A to C groups are classified into the D group. For example, the pixels of the D group originally do not include the graphic element 3, and include pixels that have not been deleted or extracted as a result of prediction.
The pixels of the D group may be registered in the <deleted pixel data> related to the shape to be deleted in the shape generation data in association with the predicted value R'.

Next, the correction processing flow of the apparatus 20 according to the second embodiment will be described with reference to FIG. 9B.
First, in S60, the correction processing unit 250 selects an arbitrary pixel X in the matrix diagram. Then, the correction processing unit 250 determines whether or not the pixel X belongs to the C group. The correction processing unit 250 advances the processing to S67 when the pixel X belongs to the C group (Y in S60), that is, the pixel X is an “extraction successful pixel”. If not (N in S60), the correction processing unit 250 advances the processing to S62.

In S62, the correction processing unit 250 determines whether or not the pixel X belongs to the B group. The correction processing unit 250 advances the processing to S64 when the pixel X belongs to the B group (Y in S62), that is, the pixel X is a “delete successful pixel” or a “misalignment source pixel”. If not (N in S62), the correction processing unit 250 advances the processing to S69.

In S64, the correction processing unit 250 searches for pixels around the pixel X. For example, the correction processing unit 250 may search for pixels in the range of ε × ε (ε is a natural number of S or less) centered on pixel A. Note that ε may be predetermined and may be set by input from the user. Then, the correction processing unit 250 has pixels in group A (that is, "misaligned pixels") around pixel X, and the predicted value R'of the pixels in group A has a value equal to the input value R of pixel X. Judge whether or not. If the condition is satisfied (Y in S64), the correction processing unit 250 considers the pixel X to be the “positional deviation source pixel” output by shifting the pixel X to the pixel of the A group, and proceeds to the process in S66. If the condition is not satisfied (N in S64), the correction processing unit 250 considers the pixel X as a "deleted successful pixel" and proceeds to the process in S68.

Next, in S66, the correction processing unit 250 changes the value of the predicted value R'of the pixel X to the value of the input value R of the pixel X, and corrects the value of the predicted value R'of the pixel X. Then, the correction processing unit 250 advances the processing to S67.

In S67, the correction processing unit 250 registers the pixel X in the <extracted pixel data> related to the shape to be extracted of the shape generation data in association with the predicted value R'. The correction processing unit 250 advances the processing to S69.

In S68, the correction processing unit 250 registers the pixel X, which is the “deleted successful pixel”, in the <deleted pixel data> in association with the predicted value R'. Then, the correction processing unit 250 advances the processing to S69.
In S69, the correction processing unit 250 determines whether or not the processing has been completed for all the pixels. The correction processing unit 250 ends the processing if it has ended, and returns the processing to S60 otherwise.
In response to the completion of the processing of the correction processing unit 250, the shape generation unit 160 is based on at least one of the <extracted pixel data> and the <deleted pixel data> of the shape generation data output from the correction processing unit 250. To generate shape data of the target partial shape.

According to the second embodiment, when the predicted value R'of the shape generation data has a misalignment, the apparatus 20 can appropriately correct the misalignment. As a result, the accuracy of the shape data of the target partial shape generated based on the shape generation data is improved.

Embodiment 3.
Next, Embodiment 3 of the present invention will be described with reference to FIGS. 10 to 13. The third embodiment is characterized in that the shape data of the target partial shape is generated in consideration of the information reduced at the time of mapping to the matrix diagram.
FIG. 10 is a configuration diagram of the device 30 according to the third embodiment. The apparatus 30 basically has the same configuration and functions as the apparatus 20 shown in FIG. 8 of the second embodiment, but instead of the conversion processing unit 120, the estimation unit 140, and the prediction model generation unit 180, the conversion processing unit 320 , The estimation unit 340, the duplication processing unit 350, and the prediction model generation unit 380 are provided. In FIG. 10, the same parts as those in FIG. 8 are designated by the same reference numerals, and the description thereof will be omitted as appropriate.

The conversion processing unit 320 basically has the same configuration and function as the conversion processing unit 120, but has a mapping processing unit 324 instead of the mapping processing unit 124.
The mapping processing unit 324 basically has the same configuration and function as the mapping processing unit 124. The mapping processing unit 324 assigns pixel positions to each graphic element 3 in the same manner as the mapping processing unit 124. Next, the mapping processing unit 324 calculates the sum of the values of the representative position information R or the graphic classification information K of the graphic element corresponding to the pixel for each pixel, and generates a matrix diagram. However, the mapping processing unit 324 includes the duplication preprocessing unit 330 in addition to the mapping processing unit 124.

The duplication preprocessing unit 330 calculates the number of graphic elements 3 (overlapping value m) corresponding to each pixel of the matrix diagram, and generates duplicate data in which the duplication value m is converted into the matrix diagram.

The estimation unit 340 basically has the same configuration and function as the estimation unit 140. However, in addition to this, the estimation unit 340 uses the duplication data acquired from the duplication preprocessing unit 330 and the prediction model acquired from the prediction model determination unit 384 of the prediction model generation unit 380 to predict the duplication value (overlapping prediction). Value) m'is estimated. The estimation unit 340 outputs the overlap prediction value m'to the overlap processing unit 350 via the correction processing unit 250.

The duplication processing unit 350 calculates the number of combinations of the graphic elements 3 corresponding to one pixel from the overlapping predicted value m', the representative position information R'of the graphic element 3 and the predicted value K'of the graphic classification information, and calculates the number of combinations. Determine one combination. The details of the duplication processing of the duplication processing unit 350 will be described later.

The prediction model generation unit 380 is basically the same as the prediction model generation unit 180, but has a learning processing unit 382 and a prediction model determination unit 384 in place of the learning processing unit 182 and the prediction model determination unit 184.
The learning processing unit 382 is basically the same as the learning processing unit 182, but manages a matrix diagram having an overlapping value m as learning data of the prediction model in addition to the learning data of the second embodiment.
The prediction model determination unit 384 is basically the same as the prediction model determination unit 184, but in addition to the matrix diagram of the second embodiment, the prediction model corresponding to the matrix diagram of the overlapping value m is determined and determined. The model is output to the estimation unit 340.

FIG. 11A is a schematic diagram of the values stored in the prediction matrix diagram regarding the predicted value R'of the representative position information according to the third embodiment. FIG. 11B is a schematic diagram of the values stored in the prediction matrix diagram regarding the overlapping predicted value m'corresponding to the predicted value R'of the representative position information according to the third embodiment.
In the third embodiment, these matrix diagrams are prediction matrix diagrams included in the shape generation data output from the estimation unit 340 via the correction processing unit 250. The prediction matrix diagram in this figure has a resolution of 8 × 8 (that is, S = 8).

As shown in FIGS. 11A and 11B, for example, the value of the representative position information R'of the pixel in the 7th row and the 2nd column is 10, and the overlap prediction value m'is 3. Therefore, three graphic elements 3 whose total value of the representative position information R'is 10 are assigned to this pixel. In addition, where the value of the representative position information R'of one graphic element 3 is S or less, since the value of the representative position information R'is larger than S in this pixel, a plurality of graphic elements 3 are surely assigned. This can be seen from the value of R'.
Similarly, the value of the representative position information R'of the pixels in the second row and the fifth column is 7, and the overlap prediction value m'is 2. Therefore, two graphic elements 3 whose total value of the representative position information R'is 7 are assigned to this pixel. As described above, the duplicate prediction value m'can indicate the number of graphic elements 3 assigned to the pixels even when the value of R'is S or less.

Next, the target partial shape generation process and the learning data process according to the third embodiment will be described with reference to FIG. FIG. 12A is a flowchart showing an example of the target partial shape generation process of the apparatus 30 according to the third embodiment, and FIG. 12B shows an example of the learning data processing of the apparatus 30 according to the third embodiment. It is a flowchart. The same steps as those in FIGS. 5 and 6 are designated by the same reference numerals and the description thereof will be omitted.

Regarding FIG. 12A, in S70, the overlap preprocessing unit 330 determines the overlap value m for each pixel of each matrix diagram, and generates a matrix diagram of the overlap value m for each matrix diagram. Then, the overlap preprocessing unit 330 outputs all the matrix diagrams to the estimation unit 340.
In S72, the estimation unit 340 acquires the corresponding prediction model from the prediction model determination unit 384 for the matrix diagram of the overlapping value m in addition to S13 in FIG.

In S74, the estimation unit 340 estimates the predicted value m'using the prediction model for the matrix diagram having the overlapping value m in addition to S14 in FIG. 5, and generates the prediction matrix diagram. Then, the estimation unit 340 outputs a prediction matrix diagram of the predicted value m'to the overlapping processing unit 350 via the correction processing unit 250.
In S75, the correction processing unit 250 performs the same correction processing as the step shown in FIG. 9B on the shape generation data including the prediction matrix diagram of the representative position information output from the estimation unit 340, and after the correction. Generate data for shape generation.

In S76, the duplication processing unit 350 executes the duplication processing described later using the prediction matrix diagram of the acquired predicted value m'and the shape generation data output from the correction processing unit 250. As a result, the duplication processing unit 350 determines the combination of the plurality of graphic elements 3 reduced to one pixel. Then, the duplication processing unit 350 registers the data corresponding to the determined graphic element 3 in the shape generation data, and updates the shape generation data. The duplication processing unit 350 outputs the shape generation data to the shape generation unit 160. Then, the duplicate processing unit 350 advances the processing to S15.

Further, as shown in FIG. 12B, in S80, the overlap pretreatment unit 330 performs the same processing as in S70.

FIG. 13 is a flowchart showing an example of duplicate processing of the apparatus 30 according to the third embodiment.
First, in S90, the duplication processing unit 350 acquires <extracted pixel data> of shape generation data from the correction processing unit 250.

Next, in S91, the duplication processing unit 350 determines whether or not the duplication prediction value m'is 2 or more for any pixel included in the extraction pixel data of the shape generation data. If m'is 2 or more (Y in S91), the duplicate processing unit 350 advances the process to S93, and if not (N in S91), advances the process to S92.

In S92, the duplication processing unit 350 leaves the predicted value R'of the representative position information as <extracted pixel data> of the shape generation data. Then, the duplicate processing unit 350 advances the processing to S98.
In S93, the duplication processing unit 350 deletes the prediction value R'of the representative position information from the <extracted pixel data> of the shape generation data according to the duplication prediction value m'being 2 or more.

In S94, the duplication processing unit 350 acquires the predicted value K'of the graphic classification information corresponding to the pixel of the deleted representative position information.
In S95, the duplication processing unit 350 calculates the combination of the duplication graphic elements 3 included in the pixel based on the duplication prediction value m'and the prediction value K'of the graphic classification information. For example, when m'= 2, the pixel contains two graphic elements 3. Here, when K'= 40, the pixel contains the graphic element 3 having the graphic classification information of 10 and 30, and the pixel contains the graphic element 3 having the graphic classification information of 20 and 20. It is possible to calculate two combinations of and.

In S96, the duplication processing unit 350 determines the combination of the overlapping graphic elements 3. In the third embodiment, the duplication processing unit 350 generates as many graphic elements 3 as the number of combinations according to the combination, and selects and determines the combination of the graphic elements 3 by comparing with the shape data of the composite shape 1. The combination of the graphic elements 3 may be determined by receiving an input from the user.

In S97, the duplication processing unit 350 registers the data corresponding to the determined combination of the graphic elements 3 in the <extracted pixel data> of the shape generation data, and updates the shape generation data.
In S98, the duplicate processing unit 350 determines whether or not processing has been performed on the predicted values R'of all the representative position information. The duplicate processing unit 350 ends the processing if it is being performed (Y in S98), and returns the processing to S91 if it is not (N in S98).
Then, in response to the completion of the processing of the duplication processing unit 350, the shape generation unit 160 generates the shape data of the target partial shape based on the <extracted pixel data> of the output shape generation data. When the shape generation unit 160 generates the shape data of the target partial shape based on the <deleted pixel data> of the shape generation data, the duplication processing unit 350 performs each step shown in FIG. 13 as <extracted pixels. It may be executed by reading from <data> to <deleted pixel data>.

As described above, according to the third embodiment, since the apparatus 30 performs the duplication processing, it is possible to generate the shape data of the target partial shape in consideration of the information reduced at the time of mapping to the matrix diagram. Therefore, the accuracy of the shape data of the generated target partial shape is improved.

In the third embodiment, the estimation unit 340 estimates the overlap prediction value m', but this estimation may be omitted. In this case, the duplication processing unit 350 shown in FIG. 10 has one duplication value m instead of the duplication prediction value m', the representative position information R'of the graphic element 3, and the prediction value K'of the graphic classification information. The combination is determined after calculating the number of combinations of the graphic elements 3 corresponding to the pixels. At this time, the duplication processing unit 350 may read each step shown in FIG. 13 from the duplication prediction value m'to the duplication value m and execute the step.

FIG. 14 is a configuration diagram of the computer 1900 according to the first to third embodiments. As shown in FIG. 14, the computer 1900 includes a control unit 1000 for controlling the entire system. A storage device 1200, a storage medium drive device 1300, a communication control device 1400, and an input / output I / F 1500 are connected to the control unit 1000 via a bus line such as a data bus.

The control unit 1000 includes a CPU 1010, a ROM 1020, and a RAM 1030.
The CPU 1010 performs various information processing and control according to a program stored in various storage units such as the ROM 1020 and the storage device 1200.
The ROM 1020 is a read-only memory in which various programs and data for the CPU 1010 to perform various controls and calculations are stored in advance.

The RAM 1030 is a random access memory used by the CPU 1010 as a working memory. In the RAM 1030, various areas for performing various processes according to the first to third embodiments can be secured.

The input device 1050 is an input device that receives input from a user such as a keyboard, a mouse, and a touch panel. For example, the keyboard is provided with various keys such as a numeric keypad, function keys for executing various functions, and cursor keys. The mouse is a pointing device, and is an input device that specifies a corresponding function by clicking a key, an icon, or the like displayed on the display device 1100. The touch panel is an input device arranged on the surface of the display device 1100. It identifies a user's touch position corresponding to various operation keys displayed on the screen of the display device 1100, and an operation displayed corresponding to the touch position. Accepts key input.

As the display device 1100, for example, a CRT, a liquid crystal display, or the like is used. The display device displays the input result by the keyboard and the mouse, and displays the finally searched image information. Further, the display device 1100 displays an image of operation keys for performing various necessary operations from the touch panel according to various functions of the computer 1900.

The storage device 1200 includes a readable and writable storage medium and a drive device for reading and writing various information such as programs and data to and from the storage medium.
A hard disk is mainly used as the storage medium used in the storage device 1200, but a literate storage medium among various storage media used in the storage medium drive device 1300 described later may be used. good.
The storage device 1200 has a data storage unit 1210, a program storage unit 1220, and other storage units (for example, a storage unit for backing up programs and data stored in the storage device 1200) and the like (for example, a storage unit for backing up programs and data stored in the storage device 1200). ing. The program storage unit 1220 stores programs for realizing various processes according to the first to third embodiments. The data storage unit 1210 stores various data of various databases according to the first to third embodiments.

The storage medium drive device 1300 is a drive device for the CPU 1010 to read data including computer programs and documents from an external storage medium (external storage medium).
Here, the external storage medium means a storage medium in which a computer program, data, or the like is stored. Specific examples of the external storage medium include magnetic storage media such as floppy disks, hard disks, and magnetic tapes, semiconductor storage media such as memory chips and IC cards, and CD-ROMs, MOs, PDs (phase change rewritable optical discs), and the like. A storage medium from which information can be read optically may be included. Further, the external storage medium may include a storage medium using paper (and a medium having a function equivalent to paper) such as a paper card or paper tape, and a storage medium in which a computer program or the like is stored by various other methods. ..
The storage medium drive device 1300 can read a computer program from these various external storage media. In addition, the storage medium drive device 1300 can write data and the like stored in the RAM 1030 and the storage device 1200 to a writable external storage medium such as a floppy disk.

In the computer 1900, the CPU 1010 of the control unit 1000 reads a computer program from an external storage medium set in the storage medium driving device 1300 and stores (installs) it in each unit of the storage device 1200.

Then, when the computer 1900 executes various processes, the corresponding program is read from the storage device 1200 into the RAM 1030 and executed. However, the computer 1900 can also read and execute the program directly from the external storage medium into the RAM 1030 by the storage medium driving device 1300 instead of from the storage device 1200. Further, depending on the computer, various programs and the like may be stored in the ROM 1020 in advance, and the CPU 1010 may execute the programs. Further, the computer 1900 may download various programs and data from another storage medium via the communication control device 1400 and execute the programs and data.

The communication control device 1400 is a control device for connecting the computer 1900 to various external electronic devices such as other personal computers and word processors via a network. The communication control device 1400 makes it possible to access the computer 1900 from these various external electronic devices.

The input / output I / F 1500 is an interface for connecting various input / output devices via a parallel port, a serial port, a keyboard port, a mouse port, and the like.

In the first to third embodiments, the computer 1900 is composed of a computer system including a personal computer, a word processor, and the like. However, the present invention is not limited to this, and the computer 1900 can be configured by a LAN (local area network) server, a computer (personal computer) communication host, a computer system connected on the Internet, and the like. It is also possible to distribute the functions to each device on the network and configure the computer 1900 in the entire network.

The present disclosure is not limited to the above embodiment, and can be appropriately modified without departing from the spirit. It is clear from the description of the claims that such modified forms may also be included in the technical scope of the present disclosure.

The execution order of each process in the apparatus and method shown in the claims, the specification, and the drawings is not specified as "before", "prior to", etc., and the order of execution of the previous processes is not specified. It can be achieved in any order unless the output is used in later processing. Even if the scope of claims, the description, and the operation flow in the drawings are explained using "first", "next", etc. for convenience, it means that it is essential to carry out in this order. is not it.

The present disclosure is not limited to the above embodiment, and can be appropriately modified without departing from the spirit.

1 Composite shape, 2, 5, 7 partial shape, 3, 4a, 4b, 4c, 6, 8 graphic elements, 9 internal space, 10, 20, 30 devices, 50 1st to 3rd matrix diagrams, 52 4th to 6th matrix diagram, 100 storage unit, 110 acquisition unit, 120, 320 conversion processing unit, 122 representative determination unit, 124,324 mapping processing unit, 140,340 estimation unit, 160 shape generation unit, 180,380 prediction model generation unit , 182,382 learning processing unit, 184,384 prediction model determination unit, 190 learning database, 250 correction processing unit, 330 duplication preprocessing unit, 350 duplication processing unit, 1000 control unit, 1010 CPU, 1020 ROM, 1030 RAM, 1100 Display device, 1200 storage device, 1210 data storage unit, 1220 program storage unit, 1300 storage medium drive device, 1400 communication control device, 1500 input / output I / F, 1900 computer

Claims (1)

A device that generates a target partial shape from a composite shape having multiple graphic elements.
A representative determination unit that determines representative position information representing the position of the graphic element for each of the plurality of graphic elements based on the graphic classification information corresponding to the plurality of graphic elements.
Using the representative position information of the plurality of graphic elements, the representative position of the graphic element corresponding to the target partial shape by the prediction model generated by machine learning based on the composite shape for learning and the target partial shape for learning. An estimation unit that estimates information and
A device including a shape generating unit that generates the target partial shape based on the estimated representative position information of the graphic element and the graphic classification information of the graphic element.

Images