Building information acquisition system

Abstract

To solve a conventional problem, or, to provide a building information acquisition system capable of reducing effort required to generate building data as compared with conventional techniques.SOLUTION: A building information acquisition system which acquires building information related to a building plan view on the basis of on raster data of the building plan view comprises region extraction means and orientation specification means. When the building plan view is input into a first learned model, an "orientation symbol area (area in which an orientation symbol is displayed)" is extracted. When the orientation symbol area is input into a second learned model, a "drawing orientation (orientation of the building plan view)" is identified.SELECTED DRAWING: Figure 3

Description

[Technical Field]

[0001] The present invention relates to a technology for acquiring numerical information from floor plans of raster data, and more specifically, to a building information acquisition system that can acquire desired numerical information from information contained in floor plans by utilizing a trained model generated by machine learning. [Background technology]

[0002] Local governments and other organizations manage fixed asset information. In carrying out this work, they sometimes use digital data maps (hereinafter referred to as "digital topographic maps"). Digital topographic maps contain digital data of buildings (fixed assets) such as houses and warehouses, and by displaying the digital topographic maps on a display, it is possible to identify buildings, etc., which is useful for the management of fixed asset information.

[0003] However, digital data may not be generated for newly built houses or buildings that have not been identified, meaning that digital topographic maps may not include digital data for all buildings (hereinafter simply referred to as "building data").

[0004] To address this issue, local governments and other organizations are working to generate new building data to add it to digital topographic maps. The input information for generating building data is often a floor plan showing the outline of the land and building (hereafter referred to as a "building floor plan"). Since these building floor plans are often in paper form, building data is generated by converting the building floor plan into raster data. More specifically, the raster data building floor plan (hereafter simply referred to as a "raster building floor plan") is displayed on a display or other device, and the operator traces the outline of the building to generate the building data.

[0005] To incorporate the building data generated by tracing into the digital topographic map, its size and orientation must be adjusted, so the operator enlarges or reduces the building data and also changes its orientation according to the scale and orientation shown on the building plan.

[0006] As described above, generating building data based on a building floor plan requires a lot of processing, and generating missing building data requires searching for a building floor plan that corresponds to the location. Therefore, generating new building data requires a considerable amount of time and effort from the operator.

[0007] Therefore, efforts have been made to reduce the workload of operators who generate building data. For example, Patent Document 1 proposes a technology that converts "electronic cadastral map data (equivalent to raster building floor plans)" into a transparent image, overlays this on an "electronic block map (equivalent to a digital topographic map)" to generate a composite map, and allows the operator to zoom in / out, rotate, and move the building data. [Prior art documents] [Patent documents]

[0008] [Patent Document 1] Japanese Patent Publication No. 2021-61039 Summary of the Invention [Problem to be solved by the invention]

[0009] The technology disclosed in Patent Document 1 relieves the operator from the process of tracing the outline of a building, which reduces the effort required to generate building data. However, the process of enlarging / reducing, rotating, and moving the building data remains, and the effort required for these processes cannot be reduced.

[0010] An object of the present invention is to solve the conventional problems, that is, to provide a building information acquisition system that can reduce the effort required to generate building data compared to conventional techniques. [Means for solving the problem]

[0011] The present invention utilizes a trained model created through machine learning, and focuses on the fact that by simultaneously learning key elements in a building floor plan, such as the building shape and direction symbols, address, and scale, these key elements can be detected from the building floor plan all at once. This invention is based on an unprecedented idea.

[0012] The building information acquisition system of the present invention acquires "building information" related to a building floor plan based on raster data "building floor plans," and includes an area extraction means and an orientation identification means. The building floor plan includes a "northographic symbol" indicating the orientation, a "building outline" of the building when viewed from above, a "building address" indicating the location of the building, and a "planning scale" indicating the scale of the building floor plan. The area extraction means extracts a "northographic symbol area (an area where a northographic symbol is displayed)" from the building floor plan when the building floor plan is input into a first trained model. The orientation identification means identifies the "planning orientation (orientation on the building floor plan)" when the northographic symbol area is input into a second trained model. The first trained model is generated by machine learning a building floor plan with an orientation symbol area attached as an information tag, and the second trained model is generated by machine learning a northographic symbol area with an orientation on the floor plan attached as an information tag. The plan orientation is then acquired as building information.

[0013] The building information acquisition system of the present invention may further include a geometric information acquisition means, an actual building shape setting means, a building coordinate identification means, and a building layout setting means. In this case, when a building floor plan is input into the first trained model, the area extraction means extracts an "outline area (area where the building outline is displayed)," an "address area (area where the building address is displayed)," and a "scale area (area where the drawing scale is displayed)" from the building floor plan. In this case, the first trained model is generated by machine learning a building floor plan to which the outline area, address area, and scale area are attached as information tags. The geometric information acquisition means acquires "building geometric information" that constitutes the building outline based on the building outline included in the outline area, and the actual building shape setting means sets an "actual building shape" that represents the actual shape of the building outline based on the drawing scale and building geometric information included in the scale area. The building coordinate identification means is a means for identifying "building coordinates" consisting of planar coordinates based on the building address included in the address area, and the building layout setting means is a means for setting "actual building layout", which is the actual planar layout of the building outline lines, by giving an orientation to the actual building shape based on the drawing orientation identified by the orientation identification means and by assigning the building coordinates identified by the building coordinate identification means to a part of the actual building shape.

[0014] The building information acquisition system of the present invention may further include a map display control means. This map display control means is a means for displaying a "digital topographic map (digital data representing the topography)" on the display means, and for displaying the building information converted into digital data superimposed on the digital topographic map on the display means.

[0015] The building information acquisition system of the present invention may further include a first trained model generation means and a second trained model generation means, wherein the first trained model generation means is a means for generating a first trained model by machine learning a building floor plan to which a direction symbol area is attached as an information tag, and the second trained model generation means is a means for generating a second trained model by machine learning a direction symbol area to which a drawing direction is attached as an information tag. [Effects of the Invention]

[0016] The building information acquisition system of the present invention has the following advantages. (1) When generating building data, the amount of processing required by an operator can be reduced, such as tracing the building's outline, or enlarging / reducing, rotating, and moving the building data. (2) By reducing the amount of processing required by operators, mistakes such as human errors are reduced and results can be obtained in a short time. (3) Since the first trained model detects the main elements of a building floor plan at once, detection accuracy is improved and computational costs are reduced. [Brief explanation of the drawings]

[0017] [Figure 1] A paper building plan showing north arrows, building outlines, building address, and drawing scale. [Figure 2] 1 is a block diagram showing the main configuration of a building information acquisition system according to the present invention. [Figure 3] A model diagram that schematically shows the "drawing orientation," "outline area," "address area," and "scale area" included in a raster building floor plan. [Figure 4] A model diagram showing various shapes and patterns of compass symbols. [Figure 5] FIG. 10 is a model diagram illustrating a situation in which building geometric information is enlarged by an actual building shape setting means. [Figure 6] FIG. 10 is a model diagram showing a state in which the actual building shape is rotated by the building layout setting means. [Figure 7] 10 is a model diagram showing a state in which building coordinates are assigned to the centroid of an actual building shape by the building layout setting means. [Figure 8] 1 is a flowchart showing the main processing flow of the building information acquisition system of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0018] An example of an embodiment of the building information acquisition system of the present invention will be described with reference to the drawings. The present invention is capable of acquiring information (hereinafter referred to as "building information") including the outline and layout of a building (particularly its location and orientation) from various information contained in the "building floor plan" shown in Figure 1, and generating digital data (i.e., building data) based on the building information. As mentioned above, this building floor plan is often created on paper, and as shown in Figure 1, it includes at least a "compass symbol" indicating the orientation in the plan (hereinafter referred to as "plan orientation"), a "building outline" that represents the outline of the building when viewed from above, a "building address" that indicates the location of the building, and a "plan scale" that is the scale of the building floor plan.

[0019] 2 is a block diagram showing the main components of the building information acquisition system 100 of the present invention. As shown in this figure, the building information acquisition system 100 of the present invention is configured to include an area extraction means 101 and an orientation identification means 102, and can also be configured to include a geometric information acquisition means 103, an actual building shape setting means 104, a building coordinate identification means 105, a building layout setting means 106, a drawing display control means 107, a first trained model generation means 108, a second trained model generation means 109, a first trained model storage means 110, a second trained model storage means 111, a raster data storage means 112, etc.

[0020] Each of the means constituting the building information acquisition system 100 (particularly, the area extraction means 101 to the second trained model generation means 109) can be manufactured as a dedicated device, or a general-purpose computer device can be used. That is, the computer device executes arithmetic processing according to a predetermined program, thereby performing processing specific to each means. This computer device includes a processor such as a CPU, memories such as ROM and RAM, and some also include input means such as a mouse and keyboard, and a display, and can be configured, for example, as a personal computer (PC) or a server.

[0021] The first trained model storage means 110, the second trained model storage means 111, and the raster data storage means 112 can be configured as a storage device of a general-purpose computer (for example, a personal computer) or as a database server. When configured as a database server, they can be placed on a local network (LAN: Local Area Network) or as a cloud server that stores data via the Internet.

[0022] Below, each of the main elements constituting the building information acquisition system 100 of the present invention will be described in detail.

[0023] (Raster data storage means) The raster data storage means 112 stores raster data of a building floor plan (i.e., a raster building floor plan) such as that shown in Fig. 1. This raster building floor plan can be generated, for example, by scanning a paper-based building floor plan.

[0024] (area extraction means) The area extraction means 101 is a means for extracting various areas included in a building floor plan, and for example, extracts a "directional symbol area" in which a directional symbol is displayed as shown in Fig. 3. The area extraction means 101 can also extract an "outline area" in which a building outline is displayed, an "address area" in which a building address is displayed, and a "scale area" in which a drawing scale is displayed.

[0025] The area extraction means 101 includes a trained model (hereinafter referred to as the "first trained model"). This first trained model is generated by machine learning using a raster building floor plan with various information tags attached as training data, and these information tags include a "directional symbol area," a "outline area," an "address area," and a "scale area." Note that the machine learning used to generate the first trained model can be deep learning such as a convolutional neural network (CNN), as well as various conventional machine learning techniques. This first trained model is generated by the first trained model generation means 108, and the generated first trained model is stored in the first trained model storage means 110 (FIG. 2).

[0026] The "north sign area," "outline area," "address area," and "scale area" are displayed at different locations on each drawing, and their shapes and display types (characters, figures, symbols, etc.) also differ. For example, it is known that various shapes and patterns are used for north signs, as shown in Figure 4. Therefore, by repeatedly learning the positions on the drawing where various areas are depicted, as well as their shapes and display types, the first trained model can identify the type of area (north sign area, outline area, address area, scale area). In other words, the first trained model identifies the north sign area, outline area, address area, and scale area from the input raster building floor plan.

[0027] (direction identification means) The orientation identification means 102 is a means for identifying the "plan direction" indicated by the "orientation symbol" included in the orientation symbol area. The orientation identification means 102 includes a trained model (hereinafter referred to as the "second trained model"), and this second trained model is generated by machine learning using the orientation symbol area with the "plan direction" tagged as an information tag as training data. As with the first trained model, the machine learning for generating the second trained model can employ deep learning such as CNN, as well as various conventional machine learning techniques. This second trained model is generated by the second trained model generation means 109, and the generated second trained model is stored in the second trained model storage means 111 (Figure 2).

[0028] Naturally, the direction indicated by a direction symbol (i.e., the drawing orientation) differs for each drawing. Therefore, the second trained model can identify the drawing orientation by repeatedly learning the shape of the direction symbol contained in the direction symbol area and the drawing orientation indicated by that direction symbol. In other words, the second trained model identifies the drawing orientation from the input direction symbol area. The drawing orientation identified by the orientation identification means 102 is then recorded as digital data in text format.

[0029] (geometric information acquisition means) The geometric information acquisition means 103 is a means for acquiring "building geometric information" that constitutes the building outline based on the building outline included in the outline area. This building geometric information includes the graphic elements included in the building outline, i.e., components such as line segments, curves, and surfaces, as well as shape elements such as line segment lengths, curve radii, and included angles between line segments. The components (line segments and surfaces) acquired by the geometric information acquisition means 103 are converted into digital data such as polylines and polygons, and the shape elements (line segment lengths and included angles) are converted into digital data in text format. Note that the geometric information acquisition means 103 can use various conventional analysis technologies, such as image recognition technology, to acquire building geometric information as digital data from the building outline as raster data.

[0030] (Means for setting actual building shape) The real building shape setting means 104 is a means for setting an "real building shape" that represents the shape of an actual building based on the building geometric information. Since the building geometric information is composed only of dimensions on a drawing, it is naturally different from the size of an actual building. Therefore, the real building shape setting means 104 sets the real building shape by enlarging (or reducing) the building geometric information as shown in Figure 5.

[0031] The "drawing scale" is used to enlarge (or reduce) the building geometric information. Therefore, the real building shape setting means 104 identifies the drawing scale from the numbers included in the scale area extracted by the area extraction means 101. When the real building shape setting means 104 reads the drawing scale from the scale area, it can use various conventional analysis techniques, including OCR (Optical Character Recognition / Reader). The drawing scale read by the real building shape setting means 104 is recorded as digital data in text format. The real building shape setting means 104 then sets the real building shape by enlarging (or reducing) the building geometric information according to this drawing scale. Note that the value obtained by multiplying the shape elements of the building geometric information (such as line length and area) by the drawing scale can be recorded as digital data in text format, or the enlarged (or reduced) figure shown in Figure 5 can be recorded as digital data in polygon format.

[0032] (Means for identifying building coordinates) The building coordinate identification means 105 is a means for identifying the planar coordinates of a building. Coordinates are often not attached to building floor plans. Therefore, the actual building shape set up in the explanation up to this point is not assigned coordinates, meaning it is not possible to identify where on the earth this actual building shape should be displayed. Therefore, the building coordinate identification means 105 reads out the building address from the letters and numbers contained in the address area extracted by the area extraction means 101, converts the building address into specified planar coordinates, and sets them as "building coordinates."

[0033] When the building coordinate identification means 105 reads out the building address from letters and numbers, it can use various conventional analysis techniques such as OCR, just like the actual building shape setting means 104. The building address read out by the building coordinate identification means 105 is recorded as digital data in text format.

[0034] Furthermore, when the building coordinate identification means 105 converts the building address as text data into predetermined plane coordinates, it can use various conventional coordinate conversion techniques, such as obtaining plane coordinates by looking up the building address in a table (such as an address dictionary) that is created in advance and consists of addresses and plane coordinates.The plane coordinates used for building coordinates can be set in a so-called absolute coordinate system such as the Japanese Geodetic System or the World Geodetic System, or they can be set in an arbitrary coordinate system that is limited to a specific range.The building coordinates set by the building coordinate identification means 105 are recorded as digital data in text format.

[0035] (Building placement setting means) The building layout setting means 106 sets the "actual building layout," which is the planar layout of an actual building. Specifically, as shown in FIG. 6, the actual building shape is rotated according to the drawing orientation determined by the orientation determination means 102, and as shown in FIG. 7, building coordinates determined by the building coordinate determination means 105 are assigned to a portion of the actual building shape. Note that in FIG. 7, building coordinates are assigned to the centroid of the actual building shape, but building coordinates can be assigned to any position, such as the corners of the building. This allows the actual building shape to be oriented in the actual building direction and have the actual building's planar position, meaning that it can be placed in the correct position and orientation on Earth. The information required to place the actual building shape in a specified position and orientation, i.e., the combination of the drawing orientation and building coordinates, is the "actual building layout," and this actual building layout is recorded as digital data in text format. Digital data including the actual building shape and actual building layout is then recorded as "building data."

[0036] (Drawing display control means) The map display control means 107 displays a digital topographic map on a display or other display means, and also superimposes building data on the digital topographic map. This digital topographic map has a coordinate system set, so building data can be placed in the correct position and orientation on the digital topographic map. Of course, when superimposing building data, it is advisable to reduce (or enlarge) the actual building shape to match the scale of the digital topographic map before displaying it. For example, if a local government uses a geographic information system, it can display building data on the digital topographic map as its base map.

[0037] As explained above, building data is generated without the operator having to trace, zoom in / out, rotate, or move the building data. In other words, according to the present invention, the operator only needs to modify the building data, which significantly reduces the time and effort required.

[0038] (Processing flow) The main processing of the building information acquisition system 100 of the present invention will be described in detail below with reference to Fig. 8. Fig. 8 is a flow chart showing the flow of the main processing of the building information acquisition system 100 of the present invention. In Fig. 8, the central column shows the processing to be performed, the left column shows what is necessary for that processing, and the right column shows what results from that processing.

[0039] To acquire building information using the building information acquisition system 100 of the present invention, as shown in Fig. 8, first, a paper-based building floor plan is converted into a raster building floor plan by scanning or the like (Step 201 in Fig. 8), and the raster building floor plan is stored in the raster data storage means 112. Next, the area extraction means 101 reads the raster building floor plan from the raster data storage means 112, and inputs the raster building floor plan into the first trained model to extract the "orientation symbol area," "outline area," "address area," and "scale area" (Step 202 in Fig. 8).

[0040] When the area extraction means 101 extracts the direction symbol area, the direction symbol area is input to the second trained model, whereby the direction identification means 102 identifies the "drawing direction" (Step 203 in FIG. 8). In addition, the geometric information acquisition means 103 acquires "building geometric information" from the outline area extracted by the area extraction means 101 (Step 204 in FIG. 8).

[0041] When the building geometric information is acquired by the geometric information acquisition means 103, the actual building shape setting means 104 identifies the "drawing scale" from the scale area extracted by the area extraction means 101 (Step 205 in Fig. 8), and sets the "actual building shape" by enlarging (or reducing) the building geometric information according to the drawing scale (Step 206 in Fig. 8). In addition, the building coordinate identification means 105 reads out the "building address" from the address area extracted by the area extraction means 101 (Step 207 in Fig. 8), and converts the building address into planar coordinates and sets them as "building coordinates" (Step 208 in Fig. 8).

[0042] Once the actual building shape, drawing orientation, and building coordinates are obtained, the building layout setting means 106 sets the "actual building layout," which is the planar layout of the actual buildings (Step 209 in Fig. 8). Then, once the digital data including the actual building shape and actual building layout is recorded as "building data," the drawing display control means 107 superimposes and displays the building data on a digital topographical map, which is a base map of the GIS (Step 210 in Fig. 8). [Industrial Applicability]

[0043] The building information acquisition system of the present invention can be used not only for the management of fixed asset information by local governments, etc., but also for various other tasks that utilize various information contained in paper building floor plans as other digital data. Considering that the present invention can contribute to resolving the labor shortage problem that many local governments face, it can be said that the present invention is not only applicable industrially but is also expected to make a significant contribution to society. [Explanation of symbols]

[0044] 100 Building information acquisition system of the present invention 101 (Building Information Acquisition System) Area Extraction Method 102 (Building Information Acquisition System) Direction Identification Means 103 Geometric information acquisition means (of building information acquisition system) 104 (Building information acquisition system) actual building shape setting means 105 (Building information acquisition system) building coordinate identification means 106 (Building information acquisition system) building layout setting means 107 (Building Information Acquisition System) Drawing Display Control Means 108 (Building information acquisition system) first trained model generation means 109 (Building information acquisition system) second trained model generation means 110 (building information acquisition system) first trained model storage means 111 (Building information acquisition system) second trained model storage means 112 Raster data storage means (of building information acquisition system)

Claims

1. A system for acquiring "building information" related to a building floor plan based on the "building floor plan" of raster data, The building floor plan includes a "directional symbol" indicating the direction, a "building outline" when the building is viewed from above, a "building address" indicating the location of the building, and a "drawing scale" which is the scale of the building floor plan, an area extraction means for extracting a "directional symbol area" in which the direction symbol is displayed from the building floor plan when the building floor plan is input to the first trained model; and an orientation specification means for specifying a "drawing orientation" that is an orientation in the building floor plan when the orientation symbol area is input to a second trained model, The first trained model is generated by machine learning the building floor plan to which the direction symbol area is attached as an information tag, The second trained model is generated by machine learning the direction symbol region to which the drawing direction as an information tag is attached, The drawing orientation is acquired as the building information. A building information acquisition system.

2. When the building floor plan is input to the first trained model, the area extraction means extracts an "outline area" in which the building outline is displayed, an "address area" in which the building address is displayed, and a "scale area" in which the drawing scale is displayed from the building floor plan; the first learned model is generated by machine learning the building floor plan to which the outline region, the address region, and the scale region are attached as information tags; Geometric information acquisition means for acquiring "building geometric information" constituting the building outline based on the building outline included in the outline area; an actual building shape setting means for setting an "actual building shape" representing the actual shape of the building outline based on the drawing scale included in the scale area and the building geometric information; a building coordinate specifying means for specifying "building coordinates" consisting of planar coordinates based on the building address included in the address area; a building layout setting means for setting an orientation for the actual building shape based on the drawing orientation specified by the orientation specifying means, and for setting an "actual building layout" which is an actual planar layout of the building outline by assigning the building coordinates specified by the building coordinate specifying means to a part of the actual building shape; The building geometric information includes line segments constituting the building outline, lengths of the line segments on the building floor plan, and angles between the line segments; The actual building shape and the actual building layout are acquired as the building information.

2. The building information acquisition system according to claim 1.

3. The system further comprises a drawing display control means for displaying a "digital topographical map" which is digital data representing the topography on a display means, and for displaying the building information converted into digital data on the display means in a state where the building information is superimposed on the digital topographical map.

3. The building information acquisition system according to claim 2.

4. a first trained model generation means for generating the first trained model by machine learning the building floor plan to which the direction symbol area is attached as an information tag; Further provided is a second trained model generation means for generating the second trained model by machine learning the orientation symbol area to which the drawing orientation as an information tag is attached, 2. The building information acquisition system according to claim 1.

Images