JP7230288B1 - learning model - Google Patents

Abstract

A learning model, a generation method thereof, an information processing device, an information method, and a program for automatically and highly accurately calculating house evaluation-related data based on house drawings and house information are provided.
A learning model 62 inputs input data including building drawings and building information, and teacher data including processed building drawings obtained by processing the building drawings and processed building information obtained by processing the building information. a layer, an output layer that outputs evaluation basic building drawings and evaluation reference building information based on training data, an intermediate layer in which parameters are learned based on building drawings, building information, processed building drawings, and processed building information; Prepare. In the method of generating the learning model, when building drawings and building information including at least one of building components and building equipment are input to the input layer, the middle layer performs calculations to generate the same data as the processed building drawings and processed building information. The computer is operated to output the evaluation basic building drawing and the evaluation basic building information from the output layer.
[Selection drawing] Fig. 1

Description

TECHNICAL FIELD The present invention relates to a house evaluation, and more particularly to a learning model for calculating data necessary for performing house evaluation calculations based on house drawings, a learning model generation method, an information processing apparatus, an information processing method, and a computer program. .

The evaluation method of the house, which is the basis of the fixed property tax, is based on the "fixed asset (house) evaluation standard" and adopts the method of evaluation based on the rebuilding price.

The Fixed Assets (Houses) Evaluation Criteria is a public notice issued by the Minister of Internal Affairs and Communications based on the Local Tax Act to standardize evaluation methods nationwide and to allow each municipality to make proper and balanced evaluations. Specifically, it is a standard that stipulates the price evaluation standards, evaluation implementation methods, and procedures for land, houses, and depreciable assets. In addition, the reconstruction price method is the construction cost (reconstruction cost score) required if the same house that was the subject of the evaluation was to be newly built in the same place at the time of evaluation. Then, the depreciation due to the state of wear and tear caused by the passage of years after the construction of the house (aging demerit point correction rate) is taken into account to determine the value (assessed value) of the house.
The aging demerit point correction rate represents depreciation due to the state of wear and tear caused by the passage of years after the construction of the house.

Next, the method of calculating the appraisal value of a house is to add a rebuilding cost score for each house, multiply it by a demerit point correction factor according to the aging (wear and tear) situation, and further consider the price level, design management cost, etc. The appraisal value of the target house is to be obtained by multiplying the price per unit.

In the house survey, regarding the addition of rebuilding cost scores, for example, each house is classified by part based on fixed asset evaluation standards (external finishes such as roofs and outer walls, internal finishes such as floors, walls, ceilings, etc., building equipment, etc.) Adopt the calculation method by For this reason, a house survey will be conducted and evaluated individually for newly constructed or expanded buildings. That is, an investigator goes to the site, fills in the floor plan of the house and the type of finish for each room on the spot, and performs calculation processing of the house evaluation based on the record. In the actual calculation process, the number of points determined for each finish and the amount of use of that finish are used to obtain the number of points for each finish, which is aggregated for each part, and finally the rebuilding cost for each part The scores are totaled to evaluate each building. Many of the work processes for performing such house evaluations are performed manually, which is extremely inefficient.
Conventionally, the inventions described in, for example, Patent Literature 1 and Patent Literature 2 shown below are known as house evaluation calculation processing apparatuses that input a floor plan of a house and perform house evaluation calculations.

In the invention described in Patent Document 1, rating items and their additional conditions, house floor plan data, and attribute data such as rating items and quantities for each part of the house are input by a reading device, and room areas for each room name are input. To provide a rating addition device for house evaluation calculation and a rating addition method that enable efficient data input by minimizing repeated rating item input operations for each house by inputting area and rating items. there is

The invention described in Patent Document 2 selects a plurality of types of rating pattern files that store different rating items and additional conditions for rating items such as finish data for each part of a house and additional conditions, To provide a score addition device for house evaluation calculation and a score addition method capable of efficiently inputting data by minimizing the operation of inputting a score item of a house which is repeated for each house.

On the other hand, in contrast to the invention described in Patent Document 1 and the invention described in Patent Document 2, the information processing apparatus of the present invention uses the building drawing and building information, which are the original drawing and the current information, to score items for each part of the house. Attribute data such as additional conditions and quantities, house floor plan data, etc. are read, and a machine learning model is used to calculate the area of the room area and scoring items for each room name. It does not need to be created in advance, and dramatically reduces the time and labor involved in the repeated rating item input operation for each house. They are different.

JP-A-05-89142 Japanese Patent No. 3177489

In this way, although the conventional house evaluation method enables somewhat efficient data input, it still lacks efficient data input for repeated input work for multiple houses. There was a problem that the method was not provided.

The present invention has been made in view of such circumstances, and the evaluation data for the evaluation items necessary for the house evaluation is automatically obtained with high accuracy from the house floor plan, the finishing table, etc. An object of the present invention is to provide an obtainable learning model, a learning model generation method, an information processing apparatus, an information processing method, and a computer program.

A learning model according to one aspect of the present invention inputs input data including building drawings and building information, and teacher data including processed building drawings obtained by processing the building drawings and processed building information obtained by processing the building information. an output layer that outputs evaluation basic building drawings and evaluation standard building information based on training data; An intermediate layer whose parameters are learned so as to reduce the loss between the evaluation basic building drawing and the evaluation basic building information, which are the output results from the layer, and the processed building drawing and the processing building information, which are the correct values, and the building drawing includes: At least one of a building component and a building facility is provided, the building information is detailed information about the building component and the building facility, and the processed building drawing labels and colors the building component and the building facility, respectively. At least one of the processed building component and the processed building equipment that have been processed is provided, the processed building information is detailed information about the processed building component and the processed building equipment, and the building drawing and the building information are input to the input layer In this case, the computer is operated to output from the output layer the same evaluation basic building drawing and evaluation basic building information as the processed building drawing and the processed building information, respectively, through calculation by the intermediate layer.

A method of generating a learning model according to an aspect of the present invention comprises a building information reading step of reading at least a building drawing and building information, which is detailed information about the building; and a processing step of adding information about the processing to the building information, and at least one of the processed building drawing and the processed building information processed based on the building drawing and the building information by the processing step. A teacher data acquisition step of acquiring a building drawing and processed building information, which is detailed information about processed building components and processed building equipment, as teacher data; and a learning model generation step of generating a learning model outputting the same evaluation basic building drawing and evaluation basic building information as the processed building drawing and the processed building information, respectively, based on the data.

An information processing apparatus according to an aspect of the present invention includes a reading unit that reads a building drawing including at least one of building components and building equipment, building information that is detailed information about the building components and building equipment, and a building drawing. A processing section that labels and colors the building components and building equipment provided by the building, and adds information on the processing to the building information, and the processing that is processed by the processing section based on the building drawing and building information. a teacher data acquisition unit that acquires, as teacher data, a processed building drawing including at least one of a building component and a processed building facility, and processed building information that is detailed information about the processed building component and the processed building facility; When input data including building information is input, an evaluation basic building drawing and evaluation basic building information that are the same as the processed building drawing and the processed building information are output based on the training data, and the processed building constituent elements included in the processed building drawing are output. Input the building drawing and building information into the learning model that outputs the evaluation basic building component elements and the evaluation basic building equipment possessed by the evaluation basic building drawing that is the same as the evaluation basic building drawing and the processed building equipment, respectively, and the evaluation basic building drawing and the evaluation basic building information , an evaluation basic information acquisition unit that acquires the evaluation basic building components and the evaluation basic building equipment included in the evaluation basic building drawing, and extracts information about the outer wall, the inner wall, and the wall thickness from the evaluation basic building drawing and the evaluation basic building information, A boundary identification unit that identifies the outline of the building configured by the information as a boundary line, and identifies detailed information about the room, including room type and room size, based on the boundary line, evaluation basic building drawing, and evaluation basic building information. a correction unit that adjusts and moves the position of the boundary line so that the boundary line identified by the boundary identification unit is configured along the rectangle of the building module; A wall identification part that identifies detailed information about the wall, including the type and size of the wall, based on the boundary line that has been obtained, the evaluation basic building drawing, and the evaluation basic building information, and the evaluation basic building drawing and the evaluation basic building information and an output unit for outputting evaluation data necessary for house evaluation based on the detailed information of the room acquired by the room specifying unit and the detailed information of the wall acquired by the wall specifying unit.

An information processing method according to an aspect of the present invention includes a reading step of reading a building drawing including at least one of building components and building equipment, and building information, which is detailed information about the building components and building equipment; A processing step of labeling and coloring the building constituent elements and building equipment provided in the building, adding information on the processing to the building information, and processing respectively processed based on the building drawing and the building information by the processing step. a teacher data acquiring step of acquiring, as teacher data, a processed building drawing including at least one of a building component and a processed building facility, and processed building information, which is detailed information about the processed building component and the processed building facility; When input data including building information is input, an evaluation basic building drawing and evaluation basic building information that are the same as the processed building drawing and the processed building information are output based on the teacher data, and the processed building drawing included in the processed building drawing is output. Building drawings and building information are input into a learning model that outputs evaluation basic building components and evaluation basic building equipment provided in the same evaluation basic building drawing as constituent elements and processed building equipment, and evaluation basic building drawings and evaluation basics An evaluation basic information acquisition step of acquiring building information, evaluation basic building components and evaluation basic building equipment provided in the evaluation basic building drawing, and obtaining information on outer walls, inner walls, and wall thickness from the evaluation basic building drawing and the evaluation basic building information. Boundary identification step of extracting and identifying building contours configured by the information as boundary lines and room details including room type and room size based on the boundary lines, the evaluation basic building drawing and the evaluation basic building information a room identification step of identifying information; a correction step of adjusting and moving the position of the boundary line identified by the boundary identification step so as to be configured along the rectangle of the building module; A wall identification step of identifying detailed wall information, including wall types and wall sizes, based on the boundary lines corrected by the step and the evaluation foundation building drawing and evaluation foundation building information; an output step of outputting evaluation data necessary for house evaluation based on the basic building information, the detailed information of the rooms obtained by the room identification step, and the detailed information of the walls obtained by the wall identification step.

A computer program according to an aspect of the present invention comprises a reading step of reading a building drawing including at least one of building components and building equipment, and building information, which is detailed information about the building components and building equipment; A processed building that is processed based on the building drawing and the building information by the processing step of labeling and coloring the building constituent elements and building equipment, and adding information about the processing to the building information, and the processing step. a teacher data acquisition step of acquiring, as teacher data, a processed building drawing including at least one of a component and a processed building facility, and processed building information, which is detailed information about the processed building component and the processed building facility; a building drawing and a building When input data including information is input, an evaluation basic building drawing and evaluation basic building information that are the same as the processed building drawing and the processed building information are output based on the teacher data, and the processed building configuration included in the processed building drawing is output. Input the building drawing and building information into the learning model that outputs the evaluation basic building component elements and the evaluation basic building equipment provided in the same evaluation basic building drawing as the element and the processed building equipment, respectively, and evaluate the evaluation basic building drawing and the evaluation basic building An evaluation basic information acquisition step of acquiring information, evaluation basic building components and evaluation basic building equipment provided in the evaluation basic building drawing, and extracting information about outer walls, inner walls, and wall thickness from the evaluation basic building drawing and the evaluation basic building information. and a boundary identification step of identifying the outline of the building configured by the information as a boundary line, and detailed room information including the room type and room size based on the boundary line, the evaluation basic building drawing and the evaluation basic building information a room identification step of identifying a room, a correction step of correcting the boundary identified by the boundary identification step by adjusting the position of the boundary and moving it so as to be configured along the rectangle of the building module, a correction step A wall identification step of identifying detailed wall information, including wall types and wall sizes, based on the boundary lines corrected by the evaluation foundation building drawing and the evaluation foundation building information, and the evaluation foundation building drawing and the evaluation foundation and an output step of outputting evaluation data necessary for house evaluation based on the building information, the detailed information of the room obtained by the room identification step, and the detailed information of the wall obtained by the wall identification step.

According to the present invention, in house evaluation, it is possible to continuously execute the drawing analysis process, the score addition process, and the score number calculation process, and the score item input operation, the score item correction operation, etc., which are repeated for each house. It is possible to greatly reduce the amount of work in
As a result, it is possible to greatly reduce the work time required for calculating the number of scores of the house, including the score correction, and thus realize more efficient processing.

In addition, since it is possible to partially modify the rating items and correction items that have been added, it is possible to flexibly respond to changes, additions, and deletions of scores, which has the effect of improving work efficiency. be.

It should be noted that the present invention frees the user from long hours of repetitive simple work that requires a great deal of concentration, reduces fatigue, and frees the user from mistakes associated with fatigue. Since it will lead to a decrease in litigation by taxpayers caused by fluctuations in economic growth, the Sustainable Development Goals (SDGs) led by the United Nations, Goal 3 "Good Health and Well-Being" and Goal 8 "Decent Work and Economic Growth" It will also be possible to contribute to Goal 9, “Industry, innovation and infrastructure”.

FIG. 4 is an explanatory diagram relating to learning model generation processing according to the first embodiment of the present invention; FIG. 4 is an explanatory diagram relating to learning model generation processing according to the first embodiment of the present invention; FIG. 4 is an explanatory diagram relating to learning model generation processing according to the first embodiment of the present invention; FIG. 2 is a schematic diagram relating to processing of building drawings and building information according to Embodiment 1 of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS It is a schematic diagram regarding the building drawing which concerns on Embodiment 1 of this invention. FIG. 2 is a schematic diagram related to building information according to Embodiment 1 of the present invention; FIG. 2 is a schematic diagram related to building information according to Embodiment 1 of the present invention; BRIEF DESCRIPTION OF THE DRAWINGS It is a schematic diagram regarding the top view of the building drawing which concerns on Embodiment 1 of this invention. It is a schematic diagram regarding the processing building drawing which concerns on Embodiment 1 of this invention. It is a schematic diagram regarding the processing building drawing which concerns on Embodiment 1 of this invention. FIG. 2 is an explanatory diagram relating to a processed building drawing according to Embodiment 1 of the present invention; FIG. 4 is a flow chart diagram relating to processing of building drawings and building information according to Embodiment 1 of the present invention. FIG. 4 is a flow chart diagram relating to learning model generation processing according to the first embodiment of the present invention. 1 is a block diagram showing the configuration of an information processing apparatus according to Embodiment 1 of the present invention; FIG. 1 is a block diagram showing the configuration of an information processing apparatus according to Embodiment 1 of the present invention; FIG. 1 is a block diagram showing the configuration of an information processing apparatus according to Embodiment 1 of the present invention; FIG. It is a flowchart figure which shows the processing procedure of the house evaluation data output by the information processing apparatus which concerns on Embodiment 1 of this invention. It is a flowchart figure relevant to the processing procedure of the house evaluation data output by the information processing apparatus which concerns on Embodiment 1 of this invention. It is a flowchart figure relevant to the processing procedure of the house evaluation data output by the information processing apparatus which concerns on Embodiment 1 of this invention. 1 is a schematic diagram of an outline of a drawing of a building according to Embodiment 1 of the present invention; FIG. FIG. 2 is a schematic diagram obtained by converting an outline of a drawing of a building according to Embodiment 1 of the present invention into a drawing of line segments; FIG. 2 is a schematic diagram in which a vector quantity is added to a line segment drawing of a building drawing according to Embodiment 1 of the present invention; It is explanatory drawing which extracted the outer wall and the inner wall in the line segment drawing of the building drawing which concerns on Embodiment 1 of this invention. It is explanatory drawing which extracted the outer wall and the inner wall in the line segment drawing of the building drawing which concerns on Embodiment 1 of this invention. It is explanatory drawing which specified the division in the line segment drawing of the building drawing which concerns on Embodiment 1 of this invention. It is explanatory drawing which specified the boundary line in the line segment drawing of the building drawing which concerns on Embodiment 1 of this invention. It is explanatory drawing which specified the boundary line in the line segment drawing of the building drawing which concerns on Embodiment 1 of this invention. It is explanatory drawing which specified the boundary line in the line segment drawing of the building drawing which concerns on Embodiment 1 of this invention. FIG. 3 is an explanatory diagram showing a defect of a boundary line in a building drawing according to Embodiment 1 of the present invention; It is explanatory drawing which eliminated the defect of the boundary line in the building drawing which concerns on Embodiment 1 of this invention. FIG. 2 is a schematic diagram of a drawing configured with boundary lines in the building drawing according to Embodiment 1 of the present invention. It is explanatory drawing which added the vector amount to the drawing comprised with the boundary line in the drawing of the building which concerns on Embodiment 1 of this invention. FIG. 3 is an explanatory diagram illustrating a rule for adding vector quantities to a drawing composed of boundary lines in the building drawing according to the first embodiment of the present invention; FIG. 2 is a schematic diagram in which a vector quantity is added to a drawing composed of boundary lines in the drawing of a building according to Embodiment 1 of the present invention. FIG. 4 is an explanatory diagram identifying rooms in a drawing formed by boundary lines in the building drawing according to Embodiment 1 of the present invention. FIG. 2 is a schematic diagram identifying rooms in a drawing formed by boundary lines in the building drawing according to Embodiment 1 of the present invention. It is explanatory drawing which correct|amended the drawing comprised with the boundary line in the building drawing which concerns on Embodiment 1 of this invention. It is explanatory drawing which correct|amended the drawing comprised with the boundary line in the building drawing which concerns on Embodiment 1 of this invention. FIG. 2 is a schematic diagram of a corrected drawing composed of boundary lines in the building drawing according to Embodiment 1 of the present invention. FIG. 2 is an explanatory diagram identifying walls of a drawing formed by boundary lines in the building drawing according to Embodiment 1 of the present invention.

BEST MODE FOR CARRYING OUT THE INVENTION Hereinafter, embodiments of the present invention will be described with reference to the drawings. However, the components described in the following embodiments are merely examples, and are not intended to limit the technical scope of the present invention only to them. Note that the terms “figure” and “drawing” in this specification may both indicate a concept or may indicate specific data depending on the context. Where "drawings" and "drawings" are data, they are images of any format.

For convenience of explanation, the same reference numerals given to each drawing indicate the same or equivalent parts unless otherwise specified, and also intend elements having similar structures and/or functions. Detailed description may be omitted.
Further, overlapping explanations are appropriately simplified or omitted.
In addition, in this embodiment, the symbol "-" may be used to represent a numerical range, and the numerical values before and after "-" are included in the numerical range.
(Embodiment 1)
[Learning model]

1 to 4 are explanatory diagrams relating to the process of generating the learning model 62 shown in FIG. FIG. 1 conceptually shows a process of performing machine learning and generating a learning model 62 . As shown in FIG. 1, the neural network that generates the learning model 62 has an input layer that receives information, an intermediate layer that receives information from the input layer, performs various calculations, and weights between the input layer and the intermediate layer. It has an output layer whose values are multiplied and processed with an activation function. By matching the results obtained in the output layer with the training data and correcting and adjusting errors from the output layer to the input layer, even a complex neural network with many intermediate layers can learn more appropriately. becomes possible.

As shown in FIG. 2, the input layer includes building drawings and building information included in input data, and processed building drawings and processed building information obtained by processing the building drawings and building information, respectively. It has a plurality of neurons that receive inputs such as information about each pixel included and information about the processed building, and passes the input information about the pixels to the intermediate layer.
The intermediate layer has a plurality of neurons for extracting information about the image of the processed building drawing and information about the processed building information, and passes the extracted information to the output layer.

Although the number of intermediate layers is three in FIG. 1, the present invention is not limited to this. For example, when the learning model 62 is a CNN, the intermediate layer alternately has a convolution layer that convolves the pixel values of each pixel input from the input layer and a pooling layer that maps the pixel values convolved in the convolution layer. It has a connected configuration, and while compressing the pixel information of the processed building drawing, etc., finally extracts the feature amount of the image. In FIG. 1, descriptions of the convolution layer and the pooling layer are omitted.
As shown in FIG. 3, the output layer outputs the evaluation basic building drawing and the evaluation basic building information based on the image feature amount and the like output from the intermediate layer.

The learning model 62 is learned so that the processed building drawing and the processed building information obtained by processing the building drawing and the building information match the evaluation basic building drawing and the evaluation basic building information, and when the building drawing and the building information are input, It is learned to output the evaluation basic building drawing and the evaluation basic building information in which the building drawing and the building information are respectively processed and the processed building drawing and the processed building information match the evaluation basic building drawing and the evaluation basic building information.

Here, the processed building drawing is a drawing in which labels and colors are added to each of a plurality of building components and building facilities included in the building drawing. Specifically, as shown in FIG. 4, in architectural drawing data, label information and color information are added to pixels belonging to each building component and each building facility, and the added information (processing additional information ) is added to the building information to obtain the processed building information.

Building drawings include layout, floor plan, elevation, electrical/heating wiring diagram, development, foundation plan, foundation cross-section, structure plan, hardware plan, rectangular gauge, electrical equipment construction special specifications, Includes special specifications for mechanical equipment construction, outdoor water supply and drainage system drawings, etc. FIG. 5 shows a plan view and an elevation view according to Embodiment 1 of the present invention.

A floor plan is a drawing that shows the shape of the building and its positional relationship with the site. It is a horizontal cross section at a height of about 1 m from the floor surface of each floor. It is a drawing for knowing the internal structure. Projections from the top of the object, floor plan, floor height, area, wall structure, opening orientation of openings, fittings, built-in furniture, planar dimensions such as equipment placement, etc. are displayed. Dimensions are drawn. The floor plan is the most basic drawing in the design documents, and is often used as a heading or index for other drawings. It is a figure drawn at an appropriate scale by looking down from directly above on a cross section of a building or civil engineering structure cut horizontally at a certain height. The plan view described in the first embodiment of the present invention is a drawing showing the shape of each floor and describing the floor area and the quadrature method.

Elevation is a drawing showing the overall appearance of a building, and unless there is a need to specify otherwise, it is normal to create a three-dimensional drawing of the four sides, north, south, east and west. is. The shapes of exterior walls, doors, windows, and entrances, as well as the reference ground surface, are illustrated, and side slanted lines and road slanted lines are also checked on this drawing. A projection onto a vertical plane, including a front view of an object viewed from the front or a side view viewed from the side. The scale is about 1/50 to 1/100. Elevation drawings have a strong character as exterior design drawings of buildings, and are often referred to as 'figure drawings' because they often omit details such as detailed dimensions. The layout of each drawing is written from left to right in counterclockwise order, but recently there are many cases where 3D images are created using CG and color perspective.

Further, building constituent elements refer to elements that constitute a building, such as outer walls, inner walls, fittings, floors, and the like. Building equipment refers to various equipment installed in buildings, and refers to systems and equipment that consist of machines, piping wiring, appliances, etc. that allow you to select and use air, water, electricity, gas, etc.

The building information is detailed information about each building component and each building facility, and includes the name, layout, size, etc. of each building component and each building facility. The building information includes a finishing table as shown in FIG. 6, which summarizes the finishing of various parts of the building, and a list of facilities such as large items such as foundations, structures, roofs, and outer walls, as well as kitchens and washbasins. , specifications and the like as shown in FIG. The finishing table is a drawing showing finishing materials such as floors, walls and ceilings of each room. Finishing tables and specifications are composed of tables and characters. In addition, the finish table includes an exterior finish table that indicates the finish of the "outside of the house" such as the roof and outer walls, and an interior finish that indicates the finish of the "inside (interior) of the house" such as the walls, floors, and ceilings of each room. Tables, etc. are included.

Next, in the processed building drawing shown in FIG. 9, label information such as building component names and building equipment names is added to pixels of segments of each building component and each building equipment included in the building drawing shown in FIG. and color information associated with them are added. Information such as building component names, etc., building equipment names, etc., and information on colors associated with them are, for example, exterior wall = 1 (ocher), interior wall = 2 (turquoise), living room = 3 (yellow green). is shown as The building components and building equipment are fabricated and obtained as fabricated building components and fabricated building equipment.
Fabricated building information includes detailed information about fabricated building components and fabricated building equipment.

In the processed building drawing, the building drawing is colored by processing, and in the processed building information, label and color information are added to the building information by processing. The added information such as labels and colors is used as processing additional information.
10 and 11 show the processed building drawing shown in FIG. 9 disassembled for each processed building component and for each processed building equipment.
[Processing drawing generation process]
FIG. 12 is a flow chart showing an example of a processing procedure of building drawing and building information processing by the control unit 10 provided in the information processing apparatus 1 shown in FIG. 14 .

For example, the control unit 10 included in the information processing apparatus 1 shown in FIG. 14 reads building drawings and building information (S01). Next, the control unit 10 performs a processing process of adding label information and the like and color information and the like to the building drawing and the building information (S02). Next, the control unit 10 acquires the processed building drawing and the processed building information obtained by processing the building drawing and the building information (S03).
[Learning model generation method]
FIG. 13 is a flowchart showing an example of the processing procedure of the learning model 62 generation processing by the control unit 10 included in the information processing apparatus 1 .
The control unit 10 shown in FIG. 14 acquires the processed building drawing and the processed building information obtained by processing the building drawing and the building information, respectively, as teacher data (S11).

Next, when the teacher data is input, the control unit 10 generates a learning model 62 (learned model) that outputs an evaluation basic building drawing and evaluation basic building information that match the processed building drawing and the processed building information, respectively. (S12).

Specifically, the control unit 10 inputs a building drawing and building information included in the input data, and a processed building drawing and processed building information obtained by processing the building drawing and building information, which are teaching data, to the input layer of the neural network. , an evaluation basic building drawing and evaluation basic building information corresponding to the building drawing and building information, respectively, are output from the output layer. The control unit 10 compares the evaluation basic building drawing and the evaluation basic building information with the correct values of the teacher data (label information and color information for each pixel, etc.), and outputs the evaluation basic building drawing and the evaluation basic building that are output from the output layer. The parameters (weights, etc.) used for arithmetic processing in the intermediate layer are optimized so that the information approaches the correct value. The parameters are, for example, weights (coupling coefficients) between neurons, coefficients of activation functions used in each neuron, and the like. Although the parameter optimization method is not particularly limited, for example, the information processing device 1 can optimize various parameters using the error backpropagation method.

The learning model 62 can also be optimized and the accuracy improved by increasing the types and numbers of processed building drawings and processed building information and increasing the number of times of learning. The control unit 10 stores the generated learning model 62 in the auxiliary storage unit 60, and ends the series of processes.

Note that the learning algorithm used to generate the learning model 62 according to the first embodiment of the present invention will be described as semantic segmentation, but it is not limited to semantic segmentation. Machine), Bayesian network, regression tree, etc., can also be applied to learning algorithms built with other learning algorithms.

Also, the image processing method according to the first embodiment of the present invention uses discrimination by a trained model using semantic segmentation, but the present invention is not limited to this, and other image processing methods may be used. For example, it is possible to discriminate between a processed building drawing and an evaluation basic building drawing using a learned model modeled by gradient boosting, random forest, or the like. In this case, the control unit 10 can divide an image into meshes, obtain the average and standard deviation of RGB in each mesh, and use the obtained values as explanatory variables. That is, the control unit 10, as a discriminating means, converts an output value obtained by inputting an explanatory variable obtained from each of a large number of learning images into a learned model whose parameters are updated so as to match the annotation of the learning image. , the explanatory variables obtained from the image are input for discrimination.
[Semantic segmentation]
Semantic segmentation can be used for the machine learning model according to Embodiment 1 of the present invention.

Here, semantic segmentation is a deep learning algorithm that associates labels and categories with each pixel (pixel) in an image. used to recognize collections of

As a method used in semantic segmentation, for example, by considering a fully connected layer of a general CNN to perform 1 x 1 convolution covering the entire region, the same result can be obtained even if the fully connected layer is replaced with a convolution layer. FCN (Fully Convolutional Networks) that can be obtained, SegNet which is an Encoder-Decoder type FCN, Encoder-Decoder type U characterized by "skip connection" that skips and joins low-dimensional features to high-dimensional features -Net, Encoder-Decoder type, using Spatial Pyramid Pooling, enabling highly accurate semantic segmentation by obtaining rich surrounding context PSPNet, v1 uses Atrous convolution (Dilated convolution) for convolution layer In v2, Atrous Spatial Pyramid Pooling (ASPP), inspired by Spatial Pyramid Pooling, was added, and in v3, the "improved ASPP" that improved the ASPP added in v2 was used, and v1 and v2 were used. Post-processing by CRF was abolished, v2's "multi-scale processing of images" was changed to "multi-scale processing of features" to improve calculation efficiency, Batch Normalization was introduced, and v3+ further improved v3, Building a faster and more powerful network by adding a simple and efficient decoder module, introducing the Xception model in the backbone, and applying depthwise separable convolution to both Atrous Spatial Pyramid Pooling and the decoder module. There is DeepLab (v1-v3+) etc.
[Information processing device]
FIG. 14 is a block diagram showing a schematic configuration of the information processing device 1 according to Embodiment 1 of the present invention.

The information processing apparatus 1 is an information processing apparatus capable of various types of information processing and transmission/reception of information, and is a computer. Computers can be divided into personal computers (personal computers), workstations, server computers (servers), mainframes, supercomputers (supercomputers), ultracomputers, minicomputers (minicomputers), and office computers, depending on the purpose of use, specifications, specs, performance, etc. (office computer), pocket computer (pocket computer), microcomputer (microcomputer), personal digital assistant (PDA) and sequencer (PLC: programmable logic controller), etc. All of these are applicable to information processing equipment. It is possible. Also, a cloud computer communicably connected via a network N such as the Internet can execute the processing of the information processing device 1 .
The information processing apparatus 1 according to the first embodiment of the present invention is assumed to be a personal computer, and will be described below.

The information processing apparatus 1 according to the first embodiment of the present invention reads a building drawing and building information, and processes the processed building drawing and processed building drawing, which is an image or the like subjected to image processing (processing) based on the read building drawing and building information. When building information is input to a learning model, the learning model is trained to output evaluation basic building drawings and evaluation basic building information that match the processed building drawings and processed building information, respectively, and when the building drawings and building information are input, The evaluation basic building drawing and the evaluation basic building information, which are basic information for detecting, identifying, specifying, and converting information necessary for house evaluation into data, are output using a learning model 62 that has been learned by machine learning, and the building drawing Then, from the building information and the processing additional information, a process of calculating the score items necessary for the house evaluation and the number of scores for the score items is performed. The learning model 62 is assumed to be used as a program module that is part of artificial intelligence software.
In the first embodiment of the present invention, a case will be described below in which information (house evaluation data) necessary for house evaluation is obtained when building drawings and building information are read.
As shown in FIG. 14, the information processing apparatus 1 includes a control section 10, a main storage section 50, a communication section 30, an input section 40, an output section 20, an auxiliary storage section 60 and a display section .

The control unit 10 has an arithmetic processing unit such as one or more CPU (Central Processing Unit), MPU (Micro-Processing Unit), GPU (Graphics Processing Unit), etc. A program 63 stored in the auxiliary storage unit 60 is read and executed to perform various information processing, control processing, and the like related to the information processing apparatus 1 . Each functional unit in FIG. 14 is executed by the control unit 10 operating based on the program 63 .

The control unit 10 includes, as functional units, a reading unit 11, a processing unit 12, a teacher data acquisition unit 13, an evaluation basic information acquisition unit 14, a boundary identification unit 15, a room identification unit 16, a correction unit 17, a wall identification unit 18, an evaluation An information acquisition unit 19 and an output unit 20 are provided.

The reading unit 11 reads a building drawing including at least one of building components and building equipment, and building information, which is detailed information about the building components and building equipment.

The processing unit 12 performs processing (image processing) for adding labels and colors to the read building drawings and building information, and more specifically, to the building constituent elements and building equipment included in the building drawings. to the building information.

The teacher data acquiring unit 13 obtains a processed building drawing including at least one of the processed building component elements and the processed building equipment processed by the processing unit 12 based on the building drawing and the building information, and the processed building component elements and the processed building equipment. The processed building information, which is detailed information about the building, is acquired as teacher data.

When input data including a building drawing and building information is input, the evaluation basic information acquisition unit 14 outputs the same evaluation basic building drawing and evaluation basic building information as the processed building drawing and the processed building information, respectively, based on the teacher data. Furthermore, the building drawing and building information are input to a learning model 62 that outputs the evaluation basic building component elements and the evaluation basic building equipment included in the same evaluation basic building drawing as the processed building component and the processed building equipment included in the processed building drawing. Then, the evaluation basic building drawing, the evaluation basic building information, and the evaluation basic building components and the evaluation basic building facilities included in the evaluation basic building drawing are acquired.

The boundary specifying unit 15 extracts information about outer walls, inner walls, and wall thicknesses from the evaluation basic building drawing and the evaluation basic building information, and specifies the outline of the building formed by the information as a boundary line.
The room identification unit 16 identifies detailed information about the room, including room type and room size, based on the boundary line, the evaluation basic building drawing, and the evaluation basic building information.

The correction unit 17 adjusts and moves the position of the boundary line specified by the boundary specifying unit 15 so that the boundary line is configured along the rectangle of the building module.

The wall specifying unit 18 specifies detailed information about the wall, including the type and size of the wall, based on the boundary line corrected by the correcting unit 17, the evaluation basic building drawing, and the evaluation basic building information.

The evaluation information acquisition unit 19 acquires the evaluation building drawing and the evaluation building based on the evaluation basic building drawing and the evaluation basic building information, the detailed information of the room acquired by the room specifying unit 16, and the detailed information of the wall acquired by the wall specifying unit 18. Get information.

The output unit 20 outputs evaluation data necessary for house evaluation based on the evaluation basic building drawing and the evaluation basic building information, the detailed information of the room acquired by the room specifying unit, and the detailed information of the wall acquired by the wall specifying unit. . Here, the evaluation data is, for example, information necessary for calculating the rebuilding cost score, and includes the standard score for the score item, correction coefficient, calculation unit, and the like.

Here, the score item is an item for assigning a standard score according to classifications such as types and products of materials generally used for each part, and aspects of construction. The standard score is calculated based on the construction cost for the "standard amount" according to the classification of the score item. The standard amount refers to the construction amount required per unit for each part of the house when constructing a new house standardized for each structure and use in the "reconstruction cost evaluation standard table".

In addition, the standard score is based on the cost equivalent to the construction cost calculated based on the price level in the special ward area at a predetermined time based on the base year according to the classification of the score item, and one yen of the cost is expressed as one point. ing.
The correction items and correction coefficients are corrected by the correction coefficients when the construction amount of the house to be evaluated is different from the standard construction amount for each part.

The calculation unit is shown per unit (m, m ² , m ³ , t, piece, etc.) for the convenience of assigning scores. Therefore, the rebuilding cost score for each part is calculated by multiplying the standard score by a predetermined calculation unit.

The output unit 20 also outputs the evaluation building drawing and the evaluation building information acquired by the evaluation information acquisition unit 19 . The evaluation building drawing and the evaluation building information include all data in which the building drawing and building information are digitized, and the operator corrects the correction coefficients and the calculation results of the learning model 62 as necessary when fluctuations occur. including.

The main storage unit 50 is a temporary storage area such as SRAM (Static Random Access Memory), DRAM (Dynamic Random Access Memory), flash memory, etc., and temporarily stores data necessary for the control unit 10 to perform arithmetic processing. Remember. The communication unit 30 is a communication module for performing processing related to communication, and transmits/receives building drawings, building information, and the like to/from the outside. The input unit 40 is a keyboard for inputting characters, symbols, etc., a mouse for specifying coordinates of a pointer and sending a signal for controlling the position. The display unit 70 is a display or the like for displaying characters and graphics, and displays the read architectural drawing and the like, and the calculated score items and score required for house evaluation. Further, the display unit 70 displays setting input information, setting change information, setting confirmation information, and the like for various functional units constituting the information processing apparatus 1 .

The auxiliary storage unit 60 is a large-capacity memory, hard disk, or the like, and stores programs 63 and other data necessary for the control unit 10 to execute processing. The auxiliary storage unit 60 also stores a drawing/information DB 61 and a learning model 62 . The drawing/information DB 61 includes data relating to building drawings and building information, processed building drawings and processed building information, evaluation basic building drawings and evaluation basic building information, line segment building drawings and line segment building information, evaluation building drawings and evaluation building information, etc. is a database that stores

The program 63 stored in the auxiliary storage unit 60 may be provided from a recording medium on which the program 63 is readable. The recording medium is, for example, a USB (Universal Serial Bus) memory, an SD (Secure Digital) card, a micro SD card, a portable memory such as a compact flash (registered trademark). The program 63 recorded on the recording medium is read from the recording medium using a reading device (not shown) and stored in the auxiliary storage unit 60 . Moreover, when the information processing device 1 includes a communication unit capable of communicating with an external communication device, the program 63 stored in the auxiliary storage unit 60 may be provided by communication via the communication unit.

Note that an external storage device connected to the information processing apparatus 1 can be applied to the auxiliary storage unit 60 . A multicomputer including a plurality of computers can be applied to the information processing apparatus 1, and a virtual machine that is virtually constructed by software can also be applied.

The control unit 10 can also add, delete, and edit data and programs in the drawing/information DB 61 , the learning model 62 , and the program 63 provided in the auxiliary storage unit 60 .

Next, the boundary specifying unit 15, based on the evaluation basic building drawing and the evaluation basic building information, has a wall contour extraction unit 151 that extracts the structure of the wall as a wall contour, and converts the contour into a drawing composed of line segments. Then, a line segment conversion unit 152 acquires the line segment building drawing and the line segment building information, and applies the line segment building drawing to an orthogonal coordinate system so that the vector directions of the line segments of a plurality of continuous outer walls are the same. A wall extracting unit 153 that gives vector information to the line segment of the outer wall of the wall and extracts the inner wall as a line segment from the vector information of the line segment corresponding to the outer wall and the wall thickness included in the evaluation basic building information. gives vector information in the opposite direction to the line segment of the outer wall to the line segment of the inner wall extracted by . and a boundary determination unit 154 that identifies a boundary line that is a minute and determines a configuration drawing configured by the boundary line.

Based on the boundary line, the room identification unit 16 identifies the area enclosed by the boundary line as a section of the space separated inside the building, and the evaluation basic building component and the evaluation basic building equipment. By applying the label and color information of the evaluation basic building drawing and evaluation basic building information provided to each pixel of the corresponding block, and adopting the most common type of label and color information in the block, and a room identification unit 162 for identifying the type of room.
[Information processing method]
FIG. 17 is a flow chart diagram showing an example of a processing procedure for outputting data required for house evaluation by the information processing apparatus 1 according to Embodiment 1 of the present invention.

First, the reading unit 11 reads a building drawing such as shown in FIGS. building information is read (S21).

Next, the processing unit 12 performs processing (image processing) for adding labels and colors to the read building drawings and building information, and more specifically, to the building constituent elements and building equipment included in the building drawings. A process of adding information on processing to the building information is executed (S22).

Next, the teacher data acquisition unit 13 obtains a processed building drawing including at least one of the processed building component elements and the processed building equipment processed by the processing unit 12 based on the building drawing and the building information, and the processed building component elements and Processed building information as shown in FIG. 9, which is detailed information about the processed building equipment, is acquired as teaching data (S23).

The evaluation basic information acquisition unit 14 inputs the input data including the building drawing and the building information to the learning model 62, thereby obtaining the evaluation basic building drawing and the evaluation basic building information, and the evaluation basic building components and the evaluation basic building components included in the evaluation basic building drawing. The evaluation basic building equipment is acquired (S24). Here, when input data including building drawings and building information is input, the learning model 62 outputs evaluation basic building drawings and evaluation basic building information identical to the processed building drawings and processed building information, respectively, based on the teacher data. Furthermore, it is a learned model in machine learning that outputs an evaluation basic building component element and an evaluation basic building equipment included in the same evaluation basic building drawing as the processed building component element and the processed building equipment included in the processed building drawing.

The boundary specifying unit 15 extracts information about outer walls, inner walls, and wall thickness from the basic building drawing for evaluation and the basic building information for evaluation. A boundary defined to be a line is identified (S25).

Specifically, in S25, the wall contour extraction unit 151 extracts the configuration of the wall included in the evaluation basic building drawing as a wall contour as shown in FIG. 20 based on the evaluation basic building drawing and the evaluation basic building information ( S251).

The line segment conversion unit 152 uses a technique such as the method of least squares on the figure shown in FIG. Building information is acquired (S252).

As shown in FIG. 22, the wall extracting unit 153 applies the line segment building drawing to an orthogonal coordinate system, and extracts a plurality of line segments of the outer wall so that the vector directions of the continuous line segments of the outer wall are the same. Vector information is given, and the inner wall is extracted as a line segment from the vector information of the line segment corresponding to the outer wall and the wall thickness included in the evaluation basic building information (S253). As shown in FIG. 23, the inner wall according to Embodiment 1 of the present invention is determined to be within a predetermined wall thickness range from the center line of the wall thickness, and the line segment within that range is identified as the inner wall. The specified inner wall is given a vector in the opposite direction to the vector given to the outer wall. As shown in FIG. 24, the line segment connecting to the inner wall is also determined to be the inner wall, and the newly discovered inner wall is given a vector in the same direction as the vector given to the already discovered inner wall. Similar logic is used to detect all internal walls for the x and y directions, as shown in FIG.

As shown in FIG. 26, the boundary determining unit 154 gives the line segment of the inner wall extracted by the wall extracting unit 153 vector information in the opposite direction to the line segment of the outer wall, and extracts the vector information of the outer wall, the vector information of the inner wall, and the vector information of the inner wall. From the wall thickness, a boundary line, which is a line segment in the middle of the wall thickness sandwiched between the outer wall and the inner wall, is specified, and a configuration drawing configured by the boundary line is determined (S254). First, regarding the detection of the horizontal boundary line, the vectors in the opposite direction respectively given to the outer wall and the inner wall are set, and the intermediate line between them is determined as the boundary line. As shown in FIG. 26, for example, when continuity is lost in the vectors given to the inner wall, if the vectors have the same direction and the error is within the range of (wall thickness/2) as shown in FIG. A continuous inner wall vector is determined by determining the same inner wall and approximating the discontinuous inner wall vectors by a mathematical method. A similar logic is applied to determine vertical boundaries. As a result, as shown in FIG. 28, for all outer and inner walls, a boundary line approximately equal to the center line of the wall thickness is determined. Furthermore, in the boundary line configuration shown in FIG. 28, if the mutually orthogonal boundary lines are projected or separated as shown in FIG. 29, the projected distance or separation distance is less than the wall thickness. If so, as shown in FIG. 30, the boundary lines are corrected so that the protrusion and separation of the boundary lines are eliminated. By performing correction processing for protrusion and separation of boundary lines for all boundary lines, a boundary line drawing composed of boundary lines corresponding to the building drawing can be obtained as shown in FIG.

Next, the room identification unit 16 identifies detailed information about the room, including the room type and room size, based on the boundary line, the evaluation basic building drawing, and the evaluation basic building information (S26).

Specifically, in S26, the section identification unit 161 identifies the area enclosed by the boundary line as a section of the separated space inside the building based on the boundary line (S261). First, as shown in FIG. 32, vectors assigned to all the inner walls are searched for with reference to the rightward vector assigned to the side surface corresponding to the inner wall in the drawing of the boundary line. As shown in FIG. 33(a), in the search for the inner wall vector, if the boundary line diverges, downward search is selected so that the vector is clockwise. As shown in FIG. 33(b), we choose to search left so that the vectors are clockwise. As shown in FIG. 33(c), similar to FIGS. 33(a) and (b), the upward search is chosen so that the vectors are clockwise. As a result, the search process is completed by connecting to the reference vector that started the search, and the closed partition is specified. Similarly, as shown in FIG. 34, the vectors assigned to the locations corresponding to the inner walls of all boundary drawings are searched to identify the partition.

The room identification unit 162 applies the label and color information contained in the evaluation basic building drawing including the evaluation basic building component and the evaluation basic building equipment and the evaluation basic building information to each pixel of the corresponding block, and the most in the block. The type of room for the lot is identified (S262) by employing information about a large variety of labels and colors. That is, for example, the label and color information of the evaluation basic building drawing including the evaluation basic building components and evaluation basic building equipment shown in FIG. , and the most identification information (label and color information) is adopted as the identification information that the section has. Information about the room is identified from the identification information adopted for the section. FIG. 36 shows a diagram in which the rooms in the boundary drawing are identified by the room identification unit 162 .

Subsequently, the correction unit 17 adjusts and moves the boundary lines identified by the boundary identification unit 15 so that they are configured along the rectangle of the building module (S27). The correction unit 17 moves and corrects the boundary line of the boundary line drawing shown in FIG. 36 so as to match the architectural module. Here, building modules refer to the basic dimensions and units that serve as design standards in architecture, such as the distance between pillars in a building, the dimensions of tatami mats, etc. It is a unit. Many houses in Japan use the "shaku module" with the unit of "910 mm = 3 shaku". FIG. 37 shows an example of performing correction to move the boundary line of the boundary line drawing. It is usually moved to the latitude and longitude of the building module closest to the boundary. If there is a special dimension such as a closet, an auxiliary dimension indicated by a dashed line shown in FIG. 38 is added and corrected by moving to the auxiliary dimension. FIG. 39 shows a diagram obtained by correcting the boundary drawing shown in FIG.

Next, the wall identifying unit 18 identifies detailed information about the wall, including the type and size of the wall, based on the boundary line corrected by the correcting unit 17, the evaluation basic building drawing, and the evaluation basic building information. (S28). 39 based on the evaluation basic building drawing including the evaluation basic building components and the evaluation basic building equipment shown in FIG. 40(a) and the evaluation basic building information. Identify the type, size, etc. of the wall of the

Based on the evaluation basic building drawing and the evaluation basic building information, the detailed information of the room acquired by the room specifying unit 16, and the detailed information of the wall acquired by the wall specifying unit 18, the evaluation information acquisition unit 19 obtains the information shown in FIG. Obtain an evaluation building drawing and evaluation building information corresponding to the perimeter drawing as shown.

Subsequently, the output unit 20 outputs evaluation data necessary for house evaluation based on the evaluation basic building drawing and the evaluation basic building information, the detailed information on the rooms acquired by the room specifying unit, and the detailed information on the walls acquired by the wall specifying unit. to output The output unit 20 also outputs the evaluation building drawing and the evaluation building information acquired by the evaluation information acquisition unit 19 (S29).

The embodiments disclosed this time are illustrative in all respects and should be considered not restrictive. The scope of the present invention is indicated by the scope of the claims rather than the meaning described above, and is intended to include all modifications within the scope and meaning equivalent to the scope of the claims.

In addition, in the processing or operation described above, as long as there is no contradiction in the processing or operation, such as using data that cannot be used in that step in a certain step, the processing or operation is performed. You can change it freely. Moreover, each embodiment described above is an illustration for explaining the present invention, and the present invention is not limited to these examples. The present invention can be embodied in various forms without departing from the gist thereof.

1 Information processing device 10 Control unit 11 Reading unit 12 Processing unit 13 Teacher data acquisition unit 14 Evaluation basic information acquisition unit 15 Boundary identification unit 16 Room identification unit 17 Correction unit 18 Wall identification unit 19 Evaluation information acquisition unit 20 Output unit 30 Communication unit 40 Input unit 50 Main storage unit 60 Auxiliary storage unit 61 Drawing/information DB
62 learning model 63 program 70 display unit 151 wall contour extraction unit 152 line segment conversion unit 153 wall extraction unit 154 boundary determination unit 161 section identification unit 162 room identification unit N network

Claims (2)

an input layer for inputting input data comprising building drawings and building information, and teacher data comprising processed building drawings obtained by processing the building drawings and processed building information obtained by processing the building information;
an output layer for outputting an evaluation basic building drawing and evaluation basic building information based on the training data;
When the building drawing, the building information, the processed building drawing, and the processed building information are input to the input layer, the evaluation basic building drawing and the evaluation basic building information, which are the output results from the output layer, and correct values and an intermediate layer containing parameters that are learned so as to reduce the loss between the processed building drawing and the processed building information,
the building drawing comprises at least one of building components and building equipment;
the building information is detailed information about the building components and the building equipment;
The processed building drawing comprises at least one of processed building components and processed building equipment that are labeled and colored for the building components and the building equipment, respectively;
The processed building information is detailed information about the processed building components and the processed building equipment,
When the building drawing and the building information are input to the input layer, the evaluation basic building drawing and A learning model characterized by causing a computer to function so as to output the evaluation basic building information from the output layer. The parameters used for arithmetic processing in the intermediate layer compare the evaluation basic building drawing and the evaluation basic building information with the correct values of the teacher data, and compare the evaluation basic building drawing and the evaluation basic building information with the correct values. is optimized to approach
2. The learning model of claim 1, wherein the information to be compared is label information and color information for each pixel .

Images