Indoor three-dimensional shape generation device, indoor three-dimensional shape generation method, and indoor three-dimensional shape generation program

Abstract

To simply and highly accurately generate a three-dimensional shape of an indoor space of a building.SOLUTION: An indoor three-dimensional shape generation device comprises: a laser light data acquisition unit that acquires laser light data including a reflection time and a reflection direction of reflected light of laser light irradiated toward an inner surface of an indoor space of a building; a point cloud data creation unit that creates point cloud data of the indoor space from the laser light data; a classification unit that classifies the point cloud data into ceiling candidate point cloud data, floor candidate point cloud data, and wall candidate point cloud data based on directions of normals at respective points included in the point cloud data; a floor and ceiling point cloud data detection unit that applies a floor and ceiling point cloud data detection algorithm to the floor candidate point cloud data and the ceiling candidate point cloud data and detects floor point cloud data and ceiling point cloud data of the indoor space; a wall point cloud data detection unit that applies a wall point cloud data detection algorithm to the wall candidate point cloud data and detects wall point cloud data of the indoor space; and a three-dimensional shape generation unit that generates a three-dimensional shape of the indoor space from the floor point cloud data, the ceiling point cloud data, and the wall point cloud data.SELECTED DRAWING: Figure 2A

Description

[Technical Field]

[0001] The present invention relates to an indoor three-dimensional shape generating device, an indoor three-dimensional shape generating method, and an indoor three-dimensional shape generating program. [Background technology]

[0002] In the above-mentioned technical field, Patent Document 1 discloses that a 3D model of an indoor space is created based on indoor point cloud data obtained by scanning the indoor space with a 3D laser scanner and parameters input by a user, such as the expected number of indoor rooms and the expected area of ​​the rooms (see, for example, Claim 1, paragraph

[0009] of the same document). [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Application Publication No. 2022-142994 Summary of the Invention [Problem to be solved by the invention]

[0004] However, the technology described in Patent Document 1 creates a 3D model based on parameters input by the user, making it easy to create a 3D model. However, if the input parameters are not accurate, the accuracy of the generated 3D model is low. [Means for solving the problem]

[0005] In order to achieve the above object, the indoor 3D shape generation device according to the present invention comprises: a laser light data acquisition unit that acquires laser light data including a reflection time and a reflection direction of a laser light reflected from an inner surface of a room of a building; a point cloud data creation unit that creates point cloud data of the room from the laser light data; a classification unit that classifies the point cloud data into ceiling candidate point cloud data, floor candidate point cloud data, and wall candidate point cloud data according to the direction of a normal at each point included in the point cloud data; a floor and ceiling point cloud data detection unit that applies a floor and ceiling point cloud data detection algorithm to the floor candidate point cloud data and the ceiling candidate point cloud data to detect floor point cloud data and ceiling point cloud data in the room; a wall point cloud data detection unit that applies a wall point cloud data detection algorithm to the wall candidate point cloud data to detect wall point cloud data in the room; a three-dimensional shape generation unit that generates a three-dimensional shape of the room from the floor point cloud data, the ceiling point cloud data, and the wall point cloud data; Equipped with:

[0006] In order to achieve the above object, a method for generating a three-dimensional indoor shape according to the present invention comprises: a laser light data acquisition step of acquiring laser light data including a reflection position, a reflection time, and a reflection direction of a reflected light of a laser light irradiated into a room of a building; a classification step of classifying the laser light data into ceiling candidate point cloud data, floor candidate point cloud data, and wall candidate point cloud data according to the direction of a normal at a reflection position of the laser light; a floor and ceiling detection step of applying a floor and ceiling detection algorithm to the floor candidate point cloud data and the ceiling candidate point cloud data to detect the floor and ceiling in the room; a wall detection step of applying a wall detection algorithm to the wall candidate point cloud data to detect walls in the room; a three-dimensional shape generation step of generating a three-dimensional shape of the room from the floor and ceiling detected by the floor and ceiling detection unit and the walls detected by the wall detection unit; Includes:

[0007] Furthermore, in order to achieve the above object, the indoor three-dimensional shape generation program according to the present invention comprises: a laser light data acquisition step of acquiring laser light data including a reflection position, a reflection time, and a reflection direction of a reflected light of a laser light irradiated into a room of a building; a classification step of classifying the laser light data into ceiling candidate point cloud data, floor candidate point cloud data, and wall candidate point cloud data according to the direction of a normal at a reflection position of the laser light; a floor and ceiling detection step of applying a floor and ceiling detection algorithm to the floor candidate point cloud data and the ceiling candidate point cloud data to detect the floor and ceiling in the room; a wall detection step of applying a wall detection algorithm to the wall candidate point cloud data to detect walls in the room; a three-dimensional shape generation step of generating a three-dimensional shape of the room from the floor and ceiling detected by the floor and ceiling detection unit and the walls detected by the wall detection unit; to be executed by the computer. [Effects of the Invention]

[0008] According to the present invention, the three-dimensional shape of the interior of a building can be created easily and with high accuracy. [Brief explanation of the drawings]

[0009] [Figure 1A] This is a diagram for explaining a method of generating a 3D model by irradiating a room with laser light and simply converting the data of the reflected light into a 3D model. [Figure 1B] FIG. 1B is a plan view of the 3D model shown in FIG. 1A. [Figure 1C] 1 is a diagram for explaining an overview of indoor three-dimensional shape generation by an indoor three-dimensional shape generation device according to a first embodiment of the present invention. FIG. [Figure 1D] FIG. 1D is a plan view of the three-dimensional shape shown in FIG. 1C. [Figure 2A] 1 is a block diagram illustrating the configuration of an indoor three-dimensional shape generation device according to a first embodiment of the present invention. [Figure 2B] 1 is a diagram showing an example of point cloud data created by the indoor three-dimensional shape generation device according to the first embodiment of the present invention. FIG. [Figure 2C] 1 is a diagram for explaining a method for classifying point cloud data by the indoor 3D shape generation device according to the first embodiment of the present invention. FIG. [Figure 2D]2A to 2C are diagrams showing an example of floor point cloud data and ceiling point cloud data detected by the indoor three-dimensional shape generation device according to the first embodiment of the present invention. [Figure 2E] FIG. 1 is a diagram for explaining the floor and ceiling point cloud data detection algorithm used by the indoor 3D shape generation device according to the first embodiment of the present invention, and is a perspective view schematically showing a portion of floor candidate point cloud data. [Figure 2F] FIG. 2F is a side view schematically showing the floor candidate point cloud data shown in FIG. 2E as viewed from the horizontal direction. [Figure 2G] 2A to 2D are diagrams illustrating a method for detecting two-dimensional wall point cloud data from wall candidate point cloud data by the indoor three-dimensional shape generating device according to the first embodiment of the present invention. [Figure 2H] FIG. 2 is a diagram for explaining a line detection algorithm used by the indoor three-dimensional shape generation device according to the first embodiment of the present invention, and is a plan view schematically showing a part of two-dimensional wall candidate point cloud data. [Figure 3] 3 is a diagram for explaining an example of an outlier data detection table included in the indoor 3D shape generation device according to the first embodiment of the present invention. FIG. [Figure 4] 1 is a diagram illustrating the hardware configuration of an indoor three-dimensional shape generation device according to a first embodiment of the present invention. [Figure 5] 4 is a flowchart illustrating a processing procedure of the indoor three-dimensional shape generation device according to the first embodiment of the present invention. [Figure 6] FIG. 10 is a diagram for explaining an outline of indoor three-dimensional shape generation by an indoor three-dimensional shape generation device according to a second embodiment of the present invention. [Figure 7A] FIG. 6 is a block diagram illustrating the configuration of an indoor three-dimensional shape generation device according to a second embodiment of the present invention. [Figure 7B] 10A and 10B are diagrams showing an example of two-dimensional wall candidate point cloud data and two-dimensional wall point cloud data detected by the indoor three-dimensional shape generation device according to the second embodiment of the present invention. [Figure 7C] 7C is an enlarged plan view schematically showing the vicinity of the boundary between point cloud data corresponding to a wall of a living room and point cloud data corresponding to a wall of an entrance hall in the two-dimensional wall candidate point cloud data shown in FIG. 7B. FIG. [Figure 7D] FIG. 10 is a diagram showing an example of two-dimensional wall point cloud data that reflects the uneven shape of a wall, generated by the indoor three-dimensional shape generation device according to the second embodiment of the present invention. [Figure 7E] FIG. 10 is a diagram for explaining another example of a method for generating two-dimensional wall point cloud data by the indoor three-dimensional shape generation device according to the second embodiment of the present invention. [Figure 7F] FIG. 10 is a diagram for explaining yet another example of a method for generating two-dimensional wall point cloud data by the indoor three-dimensional shape generation device according to the second embodiment of the present invention. [Figure 7G] FIG. 10 is a diagram showing an example of floor candidate point cloud data detected by the indoor 3D shape generation device according to the second embodiment of the present invention. [Figure 7H] 7B is an enlarged side view schematically showing point cloud data corresponding to a floor of a room and point cloud data corresponding to the floor in FIG. 7G. FIG. [Figure 7I] FIG. 10 is a diagram showing an example of floor point cloud data that reflects an uneven shape, generated by the indoor three-dimensional shape generation device according to the second embodiment of the present invention. [Figure 7J] FIG. 10 is a diagram showing an example of ceiling candidate point cloud data detected by the indoor 3D shape generation device according to the second embodiment of the present invention. [Figure 7K] FIG. 7D is an enlarged side view schematically showing point cloud data corresponding to portions of the ceiling other than the beams and point cloud data corresponding to the beams in FIG. 7J. [Figure 7L] FIG. 10 is a diagram showing an example of ceiling point cloud data that reflects a concave-convex shape, generated by the indoor three-dimensional shape generation device according to the second embodiment of the present invention. [Figure 8] FIG. 10 is a diagram for explaining an example of a concave-convex shape estimation table included in the indoor three-dimensional shape generation device according to the second embodiment of the present invention. [Figure 9] FIG. 10 is a diagram illustrating the hardware configuration of an indoor three-dimensional shape generation device according to a second embodiment of the present invention. [Figure 10] 10 is a flowchart illustrating a processing procedure of an indoor three-dimensional shape generation device according to a second embodiment of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0010] Hereinafter, embodiments of the present invention will be described in detail by way of example with reference to the drawings. However, the configurations, numerical values, processing flows, functional elements, etc. described in the following embodiments are merely examples, and are open to modification and alteration, and are not intended to limit the technical scope of the present invention to the following description.

[0011] [First embodiment] An indoor three-dimensional shape generating device 100 according to a first embodiment of the present invention will be described with reference to FIGS. 1A to 5. FIG.

[0012] <Prerequisite technology> In recent years, the Ministry of Land, Infrastructure, Transport and Tourism has been spearheading the use of BIM (Building Information Modeling) models for buildings. A BIM model is a three-dimensional (3D) model of a building constructed on a computer, with information on the shape of components and equipment and their attributes added. A 3D BIM model includes two-dimensional design drawings such as floor plans, elevations, and cross sections. Furthermore, by analyzing BIM models with various analytical software, various simulations can be performed for shadows, ventilation, air conditioning, lighting, and more. In this way, BIM models are used as multifaceted base data for the design and / or construction of buildings.

[0013] BIM models can also be used to efficiently maintain buildings. For example, they can be used to predict the degree of deterioration of a building over time and calculate the costs required for repairs. However, if the use of BIM models is not considered during the planning stage of a building and a BIM model is not created, the BIM model will not exist and cannot be used for building maintenance. When a BIM model does not exist, it is often created based on undigitized paper-based design drawings or measurement data obtained by measuring the dimensions of each part of the building. However, creating a BIM model from undigitized paper-based design drawings and measurement data takes a lot of time and money. Furthermore, BIM models created from undigitized paper-based design drawings and measurement data are often incomplete for use in building maintenance. For example, if a building is a commercial facility, measurements are taken while the facility is in use, making it difficult to obtain accurate measurement data.

[0014] As a technique for 3D modeling, specifically, CAD (Computer Aided Design) data, of the interior shape of an existing building, as shown in FIG. 1A , a three-dimensional shape 130 of the interior 110 is generated by 3D-converting data of reflected light when a laser beam is irradiated onto the interior 110. FIG. 1A shows an example of an interior, i.e., the interior 110, which includes a pillar 114 and a shelf 115. In the method shown in FIG. 1A , the data of reflected light of the laser beam is directly converted into 3D data to generate the three-dimensional shape 130 of the interior 110. Therefore, as shown in FIG. 1B , the generated three-dimensional shape 130 reflects the shapes of obstacles such as the pillar 114 and the shelf 115. In other words, the three-dimensional shape 130 generated by the method shown in FIG. 1A has dimensions different from the actual interior 110 and is therefore inaccurate. Therefore, the three-dimensional shape 130 generated by the method shown in FIG. 1A is not suitable for use in building maintenance and management, and is therefore not suitable for practical use.

[0015] <Technology of the indoor 3D shape generation device according to this embodiment> 1C is a diagram for explaining an overview of indoor three-dimensional shape generation by the indoor three-dimensional shape generation device 100 according to this embodiment. The indoor three-dimensional shape generation device 100 is used to generate an indoor three-dimensional shape 150, i.e., a BIM model.

[0016] The indoor 3D shape generation device 100 can easily generate a BIM model with high accuracy even if the room 110 contains obstacles such as pillars 114 and shelves 115. As shown in FIG. 1C , the indoor 3D shape generation device 100 generates a BIM model, i.e., a 3D shape 150, based on laser light data acquired by a laser sensor 120. Specifically, the laser sensor 120 first irradiates the inner surface of the room 110 with laser light and obtains laser light data, which is data on the reflected light. Next, the indoor 3D shape generation device 100 generates the 3D shape 150 by applying a specific algorithm to the laser light data. As a result, the 3D shape 150 is generated while eliminating the influence of the pillars 114 and shelves 115, as shown in FIG. 1D . In other words, the 3D shape 150 generated by the indoor 3D shape generation device 100 does not reflect the shapes of the pillars 114 and shelves 115, and the 3D shape 150 accurately reflects the actual dimensions of the room 110. Therefore, the indoor three-dimensional shape generating device 100 can generate a highly accurate three-dimensional shape 150 that is suitable for practical use.

[0017] Furthermore, as described above, the indoor three-dimensional shape generation device 100 can generate the three-dimensional shape 150 after eliminating the influence of the pillars 114 and shelves 115, so there is no need to move the pillars 114 and shelves 115 to expose the entire surface of the wall 113 before irradiating the inner surface of the room 110 with laser light. Furthermore, because the indoor three-dimensional shape generation device 100 generates the three-dimensional shape 150 based on laser light data, there is no need to generate the three-dimensional shape from paper-based design drawings that have not been converted into digital form, or to measure the dimensions of each part of the building and generate the three-dimensional shape from the obtained measurement data.

[0018] The laser light irradiated onto the inner surface of the room 110 is irradiated from a laser sensor 120 disposed in the room 110. The laser sensor 120 irradiates the inner surface of the room 110 of the building with laser light and acquires laser light data, which is data on the reflected light. When acquiring the laser light data, the laser light is scanned to evenly irradiate all surfaces of the inner surface of the room 110, i.e., the floor 111, the ceiling 112, and the walls 113. The method for evenly irradiating the laser light is not particularly limited. For example, the laser sensor 120 may be attached to a rotatable base so that the laser sensor 120 can be directed in any direction. Furthermore, in order to irradiate all inner surfaces of the room 110 with laser light and prevent blind spots, the laser light may be scanned by an operator holding the laser sensor 120 and pointing the laser sensor 120 at each inner surface of the room 110 while moving.

[0019] 2A, the configuration of the indoor 3D shape generation device 100 will be described. The indoor 3D shape generation device 100 includes a laser light data acquisition unit 201, a point cloud data creation unit 202, a classification unit 203, a floor / ceiling point cloud data detection unit 204, a wall point cloud data detection unit 205, and a 3D shape generation unit 206.

[0020] [Laser light data acquisition unit 201] The laser light data acquisition unit 201 acquires laser light data including the reflection time and reflection direction of reflected light of laser light irradiated onto the inner surface of a building room. Laser light is irradiated from the laser sensor 120. The reflection time of reflected light of laser light is the time it takes for the irradiated laser light to reflect off the target and return. The reflection direction of reflected light of laser light is the direction from which the reflected light arrives.

[0021] For example, a LiDAR (Light Detection and Ranging) sensor or the like can be used as the laser sensor 120. The laser sensor 120 may be a camera of a smartphone, a tablet, or the like with a LiDAR sensor attached thereto.

[0022] The laser sensor 120 and the indoor three-dimensional shape generation device 100 are connected by wire or wirelessly, and the laser light data acquired by the laser sensor 120 is transmitted to the indoor three-dimensional shape generation device 100 by wired or wireless communication. Instead of transmitting the laser light data by wired or wireless communication, the laser light data acquired by the laser sensor 120 may be stored in a storage medium such as a hard disk drive (HDD), a solid state drive (SSD), or a universal serial bus (USB) memory, and the indoor three-dimensional shape generation device 100 may acquire the laser light data from the storage medium. The number of laser sensors 120 used in this embodiment may be one or more.

[0023] [Point cloud data creation unit 202] The point cloud data creation unit 202 creates point cloud data 211 of the interior of the room from the laser light data, as shown in FIG. 2B. The point cloud data 211 is position data of each point on the interior surface of the room in a three-dimensional coordinate system with the position of the laser sensor 120 as the origin. In other words, the point cloud data is position data of reflection points of the laser light irradiated by the laser sensor 120. The interval between points in the point cloud data is, for example, 0 mm to 50 mm. The point cloud data may be thinned out to enable high-speed handling of the point cloud data. Note that in FIG. 2B, for convenience, boundaries of each surface in the point cloud data 211 are illustrated to make them easier to see.

[0024] [Classification Department 203] The classification unit 203 classifies the point cloud data into ceiling candidate point cloud data, floor candidate point cloud data, and wall candidate point cloud data based on the direction of the normal at each point included in the point cloud data. As shown in FIG. 2C, normals 223a-223f at each point are perpendicular lines at the points 221a-221f to be classified with respect to planes 222a-222f that are assumed to be formed from the points 221a-221f to be classified and their surrounding points. The normals have vectors pointing toward a certain point inside the room 110 as the origin. The certain point inside the room 110 may be, for example, the position of the laser sensor 120. The classification unit 203 classifies point 221a, whose normal line 223a is oriented vertically upward, as floor candidate point cloud data, point 221b, whose normal line 223b is oriented vertically downward, as ceiling candidate point cloud data, and points 221c-221f, whose normal lines 223c-223f are oriented horizontally, as wall candidate point cloud data. Note that in FIG. 2C, the up-down direction on the paper is the vertical direction, the direction from the bottom to the top of the paper is the vertically upward direction, and the direction from the top to the bottom of the paper is the vertically downward direction. Also, in FIG. 2C, the left-right direction on the paper is the horizontal direction.

[0025] [Floor and ceiling point cloud data detection unit 204] The floor and ceiling point cloud data detection unit 204 applies a floor and ceiling point cloud data detection algorithm to the floor candidate point cloud data 221a and the ceiling candidate point cloud data 221b, and detects floor point cloud data 231 and ceiling point cloud data 232 of the room 110, as shown in FIG. 2D. The floor and ceiling point cloud data detection algorithm is an algorithm that estimates a plane from point cloud data. The floor point cloud data 231 is point cloud data of a plane estimated from the floor candidate point cloud data 221a. The ceiling point cloud data 232 is point cloud data of a plane estimated from the ceiling candidate point cloud data 221b. In FIG. 2D, the contours of the floor point cloud data 231 and the ceiling point cloud data 232 are illustrated for convenience in order to make the contours of the floor point cloud data 231 and the ceiling point cloud data 232 easier to see.

[0026] The floor and ceiling point cloud data detection algorithm is an algorithm that detects floor point cloud data 231 based on the degree of match between a plane passing through three points arbitrarily selected from the floor candidate point cloud data 221a and the floor candidate point cloud data 221a. The floor and ceiling point cloud data detection algorithm is an algorithm that detects ceiling point cloud data 232 based on the degree of match between a plane passing through three points arbitrarily selected from the ceiling candidate point cloud data 221b and the ceiling candidate point cloud data 221b. Random Sample Consensus (RANSAC) is used as the floor and ceiling point cloud data detection algorithm. RANSAC is an algorithm that estimates a model representing a data group from a data group consisting of multiple data. RANSAC can estimate a model of a data group while minimizing the influence of outlier data that has extreme values ​​compared to other data.

[0027] The floor candidate point cloud data 221a or the ceiling candidate point cloud data 221b may contain outlier data. For example, when a plane is estimated using the least squares method from the floor candidate point cloud data 221a or the ceiling candidate point cloud data 221b that contains outlier data, the estimated plane may differ from the plane that should have been estimated due to the influence of the outlier data. By using RANSAC as the floor / ceiling point cloud data detection algorithm, the influence of the outlier data can be minimized, and a plane that is close to the floor 111 or ceiling 112 of the actual room 110 can be estimated.

[0028] The floor and ceiling point cloud data detection algorithm will be described with reference to Figures 2E and 2F, taking as an example a case where floor point cloud data 231 is detected from floor candidate point cloud data 221a by RANSAC. The same description applies to a case where ceiling point cloud data 232 is detected from ceiling candidate point cloud data 221b. Figure 2E is a perspective view schematically showing a portion of the floor candidate point cloud data 221a. Figure 2F is a side view schematically showing the floor candidate point cloud data 221a shown in Figure 2E as viewed from the horizontal direction.

[0029] <1-1: Floor and ceiling point cloud data detection algorithm (RANSAC)> (1-1-1) First, as shown in Fig. 2E, three arbitrary points 241a to 241c are selected from the floor candidate point cloud data 221a, and then a plane 242 passing through the selected three arbitrary points 241a to 241c is generated. (1-1-2) Next, as shown in FIG. 2F, outlier data 251 is removed from the floor candidate point cloud data 221a when the generated plane 242 is used as a reference. Various known methods can be used to remove outlier data. For example, points 251 for which the length of a perpendicular line dropped onto the plane 242 is equal to or greater than a predetermined value D are removed as outlier data. Note that the predetermined value D is a value determined by the operator in consideration of the accuracy (e.g., the degree of variation) of the floor candidate point cloud data 221a. (1-1-3) Then, the degree of match between the remaining point cloud data 252 and the plane 242 is evaluated. Various known methods can be used as a method for evaluating the degree of match. For example, MSE (Mean Squared Error) may be used as an index of the degree of match. Alternatively, the number of point cloud data 252 remaining after excluding the outlier data 251 from the floor candidate point cloud data 221a may be used as an index of the degree of match. Alternatively, the sum of the lengths of the perpendicular lines dropped onto the plane 242 for each of the remaining point cloud data 252 may be used as an index of the degree of match. (1-1-4) Repeat steps (1-1-1) to (1-1-3) a predetermined number of times. The predetermined number of times can be set arbitrarily. (1-1-5) Then, the floor candidate point cloud data 221a included in the plane with the highest degree of coincidence among the generated planes 242 is set as the floor point cloud data 231.

[0030] As the floor and ceiling point cloud data detection algorithm, in addition to RANSAC, the following algorithms such as <1-2> can also be used.

[0031] <1-2> (1-2-1) First, any three points 241a to 241c are selected from the floor candidate point cloud data 221a, and then a plane 242 passing through the selected three points 241a to 241c is generated. (1-2-2) Next, for each point of the floor candidate point cloud data 221a, it is determined whether or not it is outlier data 251 when the generated plane 242 is used as a reference. As a method for determining whether or not it is outlier data, the same method as in (1-1-2) of <1-1> can be used. (1-2-3) Repeat steps (1-2-1) and (1-2-2) a predetermined number of times. The predetermined number of times can be set arbitrarily. (1-2-4) Then, for each point in the floor candidate point cloud data 221a, it is determined whether the number of times that the point has been determined to be outlier data is equal to or greater than a certain number. (1-2-5) Points that have been determined to be outlier data a certain number of times or more are removed from the floor candidate point cloud data 221a, and the remaining point cloud data is set as floor point cloud data 231.

[0032] Above, the floor and ceiling point cloud data detection algorithm has been explained using the example of detecting floor point cloud data 231 from floor candidate point cloud data 221a, but the same algorithms as <1-1> and <1-2> can also be used when detecting ceiling point cloud data 232 from ceiling candidate point cloud data 221b.

[0033] The floor surface of the room 110 can be generated from the floor point cloud data 231 detected by the floor and ceiling point cloud data detection unit 204. For example, the plane with the highest degree of match in (1-1-5) of <1-1> may be used as the floor surface of the room 110. Also, a plane generated by applying the least squares method to the floor point cloud data 231 obtained by <1-2> may be used as the floor surface of the room 110.

[0034] Just as the floor surface of the room 110 can be generated from the floor point cloud data 231 detected by the floor and ceiling point cloud data detection unit 204, the ceiling surface of the room 110 can also be generated from the ceiling point cloud data 232 detected by the floor and ceiling point cloud data detection unit 204. For example, the plane with the highest degree of match in (1-1-5) of <1-1> may be used as the ceiling surface of the room 110. Also, a plane generated by applying the least squares method to the ceiling point cloud data 232 obtained by <1-2> may be used as the ceiling surface of the room 110.

[0035] [Wall point cloud data detection unit 205] The wall point cloud data detection unit 205 applies a wall point cloud data detection algorithm to the wall candidate point cloud data 221c to 221f to detect wall point cloud data inside the room. The wall point cloud data detection algorithm is an algorithm for estimating a plane from point cloud data. The wall point cloud data is point cloud data of a plane estimated from the wall candidate point cloud data 221c to 221f.

[0036] The wall point cloud data detection unit 205 extracts two-dimensional wall point cloud data 263 by applying a line detection algorithm to two-dimensional wall candidate point cloud data 262 obtained by orthogonally projecting the wall candidate point cloud data 221c to 221f onto a floor surface generated by the floor point cloud data 231 as a wall point cloud data detection algorithm. The wall point cloud data detection unit 205 also duplicates the extracted two-dimensional wall point cloud data 263 so that the same wall contour line continues from the floor surface to the ceiling surface generated by the ceiling point cloud data 232, thereby obtaining wall point cloud data. Here, the wall contour line means the outer periphery when the two-dimensional wall point cloud data 263 or the duplicated data is viewed in a planar view in the vertical direction. The wall point cloud data detection unit 205 will be described in detail below.

[0037] 2G, the wall point cloud data detection unit 205 first orthogonally projects the wall candidate point cloud data 221c to 221f onto the floor surface generated by the floor point cloud data 231 to obtain two-dimensional wall candidate point cloud data 262. In FIG. 2G, for the sake of convenience, the contours of the wall candidate point cloud data 221c to 221f and the two-dimensional wall candidate point cloud data 262 are illustrated to make the contours of the wall candidate point cloud data 221c to 221f and the two-dimensional wall candidate point cloud data 262 easier to see.

[0038] Next, as shown in FIG. 2G, a line detection algorithm is applied to the two-dimensional wall candidate point cloud data 262 to extract two-dimensional wall point cloud data 263. The line detection algorithm is applied to each of the point cloud data in the two-dimensional wall candidate point cloud data 262 that have the same normal direction. Specifically, the line detection algorithm is applied to each of the point cloud data whose normal direction is the X1 direction, the point cloud data whose normal direction is the X2 direction, the point cloud data whose normal direction is the X3 direction, and the point cloud data whose normal direction is the X4 direction. The two-dimensional wall point cloud data 263 is point cloud data of a line estimated from the two-dimensional wall candidate point cloud data 262. In FIG. 2G, the outline of the two-dimensional wall point cloud data 263 is illustrated for convenience in order to make the two-dimensional wall point cloud data 263 easier to see.

[0039] The line detection algorithm is an algorithm that extracts two-dimensional wall point cloud data 263 based on the degree of coincidence between a line passing through two points arbitrarily selected from the two-dimensional wall candidate point cloud data 262 and the two-dimensional wall candidate point cloud data 262. RANSAC is used as the line detection algorithm. By using RANSAC as the line detection algorithm, it is possible to eliminate the influence of outlier data as much as possible and to estimate a line that is close to the line generated by orthogonally projecting the wall 113 of the actual room 110 onto the floor 111.

[0040] Taking the case where RANSAC is used as the line detection algorithm as an example, the line detection algorithm will be described with reference to Fig. 2H, which is a plan view schematically showing a part of the two-dimensional wall candidate point cloud data 262.

[0041] <2-1: Line detection algorithm (RANSAC)> (2-1-1) First, as shown in Fig. 2H, any two points 262a and 262b are selected from the two-dimensional wall candidate point cloud data 262. Next, a straight line 271 passing through the selected two points 262a and 262b is generated. (2-1-2) Next, outlier data 272 are removed from the two-dimensional wall candidate point cloud data 262 when the calculated straight line 271 is used as a reference. Various known methods can be used to remove outlier data. For example, points 272 where the length of the perpendicular line to the straight line 271 is equal to or greater than a predetermined value H are removed as outlier data. Note that the predetermined value H is a value determined by the operator in consideration of the accuracy (e.g., the degree of variation) of the two-dimensional wall candidate point cloud data 262. (2-1-3) Then, the degree of coincidence between the remaining point cloud data 273 and the straight line 271 is evaluated. The method of evaluating the degree of coincidence can be the same as that in (1-1-3) of <1-1>. For example, the index of the degree of coincidence can be the MSE, the number of point cloud data 273 remaining after excluding the outlier data 272 from the two-dimensional wall candidate point cloud data 262, the sum of the lengths of the perpendicular lines dropped to the straight line 271 for each of the remaining point cloud data 273, or the like. (2-1-4) Repeat steps (2-1-1) to (2-1-3) a predetermined number of times. The predetermined number of times can be set arbitrarily. (2-1-5) Then, the two-dimensional wall candidate point cloud data 262 included in the straight line with the highest degree of coincidence among the generated straight lines 271 is set as the two-dimensional wall point cloud data 263.

[0042] As the line detection algorithm, in addition to RANSAC, the following algorithm <2-2> can also be used.

[0043] <2-2> (2-2-1) First, any two points 262a and 262b are selected from the two-dimensional wall candidate point cloud data 262. Next, a straight line 271 passing through the selected two points 262a and 262b is generated. (2-2-2) Next, for each point of the two-dimensional wall candidate point cloud data 262, it is determined whether or not it is outlier data 272 when the generated straight line 271 is used as a reference. As a method for determining whether or not it is outlier data, the same method as in (2-1-2) of <2-1> can be used. (2-2-3) Repeat steps (2-2-1) and (2-2-2) a predetermined number of times. The predetermined number of times can be set arbitrarily. (2-2-4) Then, for each point of the two-dimensional wall candidate point cloud data 262, it is determined whether the number of times that the point has been determined to be outlier data is equal to or greater than a certain number. (2-2-5) Points that have been determined to be outlier data a certain number of times or more are removed from the two-dimensional wall candidate point cloud data 262, and the remaining point cloud data is set as two-dimensional wall point cloud data.

[0044] In this embodiment, the laser beam data acquisition unit 201 acquires laser beam data in a state in which the pillars 114 and shelves 115 are present in the room 110. Therefore, the shapes of the pillars 114 and shelves 115 are reflected in the two-dimensional wall candidate point cloud data 262. In the two-dimensional wall candidate point cloud data 262, the point cloud data corresponding to the pillars 114 and shelves 115 are located at positions away from other point cloud data. Therefore, when a line detection algorithm is applied to the two-dimensional wall candidate point cloud data 262, the point cloud data corresponding to the pillars 114 and shelves 115 become outlier data. As a result, the two-dimensional wall point cloud data 263 extracted by the line detection algorithm excludes the pillars 114 and shelves 115, and accurately reflects the actual position of the wall 113.

[0045] The wall point cloud data detection unit 205 may use the two-dimensional wall point cloud data 263 detected by the straight line detection algorithm as the two-dimensional wall point cloud data 263 as is, or may further apply a specific algorithm to the two-dimensional wall point cloud data 263 detected by the straight line detection algorithm and use the detected point cloud data as the two-dimensional wall point cloud data 263. For example, the point cloud data obtained in (2-1-5) of <2-1> may be used as is as the two-dimensional wall point cloud data 263. Furthermore, point cloud data detected by applying the least squares method to the two-dimensional wall point cloud data 263 obtained in <2-2> may be used as the two-dimensional wall point cloud data 263.

[0046] After extracting the two-dimensional wall point cloud data 263, the wall point cloud data detection unit 205 generates wall point cloud data by duplicating the two-dimensional wall point cloud data 263 so that the same wall contour line continues from the floor surface generated by the floor point cloud data 231 to the ceiling surface generated by the ceiling point cloud data 232. As described above, the two-dimensional wall point cloud data 263 accurately reflects the positions of the actual walls 113, and therefore the wall point cloud data generated based on the two-dimensional wall point cloud data 263 also accurately reflects the positions of the actual walls 113.

[0047] [3D shape generation unit 206] The three-dimensional shape generation unit 206 generates a three-dimensional shape 150 of the room 110 from the floor point cloud data 231, the ceiling point cloud data 232, and the wall point cloud data. Specifically, the three-dimensional shape 150 of the room 110 is defined as a shape surrounded by a floor surface generated by the floor point cloud data, a ceiling surface generated by the ceiling point cloud data 232, and a wall surface generated by the wall point cloud data.

[0048] FIG. 3 is a diagram illustrating an example of an outlier data detection table 301 included in the indoor 3D shape generation device 100. The outlier data detection table 301 is a table that stores predetermined values ​​312 in association with detection targets 311 of a floor and ceiling point cloud data detection algorithm or a line detection algorithm. The detection targets 311 are point cloud data detected by the floor and ceiling point cloud data detection algorithm or the line detection algorithm. The predetermined value 312 is a threshold value for the length of a perpendicular line dropped to the reference plane 242 or line 271, which is used by the floor and ceiling point cloud data detection algorithm or the line detection algorithm when determining outlier data. In the embodiment, "D", "D", and "H" are set as the predetermined values ​​312 for the floor point cloud data, the ceiling point cloud data, and the 2D wall point cloud data, respectively. The predetermined value 312 may be a different value for each detection target 311, or may be the same value. The floor / ceiling point cloud data detection unit 204 and the wall point cloud data detection unit 205 each refer to the outlier data detection table 301 to estimate a plane or a straight line.

[0049] The hardware configuration of the indoor 3D shape generation device 100 will be described with reference to FIG. 4. The CPU (Central Processing Unit) 410 is a processor for arithmetic and control, and executes programs to realize the various functional components of the indoor 3D shape generation device 100 shown in FIG. 2A. The CPU 410 may have multiple processors and execute different programs, modules, tasks, threads, etc. in parallel. The ROM (Read Only Memory) 420 stores fixed data such as initial data and programs, as well as other programs. The network interface 430 communicates with other devices via a network. The CPU 410 is not limited to a single CPU, and may include multiple CPUs or a GPU (Graphics Processing Unit) for image processing. The network interface 430 preferably has a CPU independent of the CPU 410 and writes and reads transmission / reception data 446 to and from an area of ​​the RAM (Random Access Memory) 440. It is also preferable to provide a DMAC (Direct Memory Access Controller) (not shown) for transferring data between the RAM 440 and storage 450. Furthermore, the CPU 410 processes the data after recognizing that the data has been received or transferred to the RAM 440. The CPU 410 also prepares the processing results in the RAM 440, and leaves the subsequent transmission or transfer to the network interface 430 or DMAC.

[0050] The RAM 440 is a random access memory used by the CPU 410 as a work area for temporary storage. The RAM 440 has a storage area reserved for storing data necessary for implementing this embodiment. The laser light data 441 is data of reflected light of laser light irradiated onto the inner surface of the room 110 of the building, acquired by the laser light data acquisition unit 201. The point cloud data 442 is point cloud data of the room created from the laser light data by the point cloud data creation unit 202. The floor, ceiling, and wall candidate point cloud data 443 is the ceiling candidate point cloud data 221b, floor candidate point cloud data 221a, and wall candidate point cloud data 221c to 221f classified by the classification unit 203. The floor, ceiling, and wall point cloud data 444 is the floor point cloud data 231 and ceiling point cloud data 232 detected by the floor and ceiling point cloud data detection unit 204, and the wall point cloud data detected by the wall point cloud data detection unit 205. The three-dimensional shape data 445 is the three-dimensional shape 150 of the room 110 generated by the three-dimensional shape generating unit 206.

[0051] The transmitted / received data 446 is data transmitted and received via the network interface 430. The RAM 440 also has an application execution area 447 for executing various application modules.

[0052] The storage 450 stores a database, various parameters, or the following data or programs required to implement this embodiment. The storage 450 stores an outlier data detection table 301. The outlier data detection table 301 is a table that manages the relationship between the detection target 311 and the predetermined value 312 shown in FIG. 3.

[0053] The storage 450 further stores a laser light data acquisition module 451, a point cloud data creation module 452, a classification module 453, a floor / ceiling point cloud data detection module 454, a wall point cloud data detection module 455, and a three-dimensional shape generation module 456. The laser light data acquisition module 451 is a module that acquires laser light data of reflected light from a laser light irradiated onto the inner surface of the room 110 of a building. The point cloud data creation module 452 is a module that creates point cloud data of the room 110 from the laser light data. The classification module 453 is a module that classifies the point cloud data of the room 110 into ceiling candidate point cloud data 221b, floor candidate point cloud data 221a, and wall candidate point cloud data 221c to 221f. The floor / ceiling point cloud data detection module 454 is a module that detects ceiling point cloud data 232 and floor point cloud data 231 from the ceiling candidate point cloud data 221b and floor candidate point cloud data 221a. The wall point cloud data detection module 455 is a module that detects wall point cloud data from the wall candidate point cloud data 221c to 221f. The 3D shape generation module 456 is a module that generates the 3D shape 150 from the ceiling point cloud data 232, the floor point cloud data 231, and the wall point cloud data. These modules 451 to 456 are read into the application execution area 447 of the RAM 440 by the CPU 410 and executed. The control program 457 is a program for controlling the entire indoor 3D shape generation device 100.

[0054] The input / output interface 460 interfaces input / output data with input / output devices. A display unit 461 and an operation unit 462 are connected to the input / output interface 460. A storage medium 464 may also be connected to the input / output interface 460. A speaker 463 serving as an audio output unit, a microphone (not shown) serving as an audio input unit, or a GPS position determination unit may also be connected. Note that the RAM 440 and storage 450 shown in FIG. 4 do not include programs or data related to the general-purpose functions of the indoor 3D shape generation device 100 or other feasible functions.

[0055] Next, the processing procedure of the indoor 3D shape generation device 100 will be described with reference to the flowchart shown in Fig. 5. This flowchart is executed by the CPU 410 in Fig. 4 using the RAM 440, and realizes each functional configuration of the indoor 3D shape generation device 100 in Fig. 2A.

[0056] In step S501, the laser light data acquisition unit 201 acquires laser light data. In step S503, the point cloud data creation unit 202 creates point cloud data of the room from the laser light data. In step S505, the classification unit 203 classifies the point cloud data into ceiling candidate point cloud data 221b, floor candidate point cloud data 221a, and wall candidate point cloud data 221c to 221f according to the normal direction at each point included in the point cloud data. In step S507, the floor and ceiling point cloud data detection unit 204 applies a floor and ceiling point cloud data detection algorithm to the floor candidate point cloud data 221a and the ceiling candidate point cloud data 221b, and detects floor point cloud data 231 and ceiling point cloud data 232 of the room. In step S509, the wall point cloud data detection unit 205 applies a wall point cloud data detection algorithm to the wall candidate point cloud data 221c to 221f, and detects wall point cloud data of the room. In step 511, the three-dimensional shape generating unit 206 generates the three-dimensional shape 150 of the room 110 from the floor point cloud data 231, the ceiling point cloud data 232, and the wall point cloud data.

[0057] According to this embodiment, a worker such as an operator of the indoor three-dimensional shape generation device 100 can generate the three-dimensional shape 150 of the room 110 without inputting parameters to the indoor three-dimensional shape generation device 100. Therefore, according to this embodiment, the three-dimensional shape 150 of the room 110 of a building can be easily created.

[0058] Furthermore, in this embodiment, since the operator does not need to input parameters, the situation where inaccurate parameters are input does not occur. Therefore, according to this embodiment, the three-dimensional shape 150 of the interior 110 of the building can be created with high accuracy.

[0059] Furthermore, in this embodiment, as described above, the wall point cloud data detection unit 205 can detect wall point cloud data that accurately reflects the shape of the actual wall 113. As a result, the finally generated three-dimensional shape 150 also accurately reflects the shape of the actual room 110.

[0060] In this embodiment, the point cloud data 211 is created from the laser light data acquired by the laser sensor 120, and the three-dimensional shape 150 is generated based on the point cloud data 211. However, the point cloud data 211 used to generate the three-dimensional shape 150 may be point cloud data from which point cloud data corresponding to obstacles such as the pillars 114 and shelves 115 have been removed in advance. For example, a visible image of the room 110 may be captured by a camera or the like, and a worker such as an operator may compare the visible image with the point cloud data 211 and remove from the point cloud data 211 the point cloud data corresponding to the obstacles in the room 110.

[0061] [Second embodiment] Next, an indoor three-dimensional shape generation device 600 according to a second embodiment of the present invention will be described with reference to Figs. 6 to 10. The indoor three-dimensional shape generation device 600 is a device that generates a three-dimensional shape 650 of a room 610 having an uneven shape. As in the first embodiment, the indoor three-dimensional shape generation device 600 generates the three-dimensional shape 650 based on point cloud data obtained from laser light data acquired by the laser sensor 120. The indoor three-dimensional shape generation device 600 generates the three-dimensional shape 650 that reflects the uneven shape of the room 610 by adopting point cloud data corresponding to the uneven shape of the room 610 as the point cloud data used to generate the three-dimensional shape 650.

[0062] The interior room 610 shown in FIG. 6 has a living room 611 and a porch 612 that projects from the living room 611, and has a shape of two connected rectangular parallelepipeds. The porch 612 has a frame 611a, and the earthen floor 111b of the porch 612 is one step lower than the floor 111a of the living room 611. Furthermore, the ceiling 112 of the living room 611 is provided with beams 613 that project from the ceiling 112. The interior room 610 may have a dirt floor. The ceiling 112 may be a raised or lowered ceiling.

[0063] Next, the configuration of the indoor 3D shape generation device 600 according to this embodiment will be described with reference to Fig. 7A. The indoor 3D shape generation device 600 according to this embodiment differs from the first embodiment in that it includes a wall unevenness estimation unit 701, a floor unevenness estimation unit 702, and a ceiling unevenness estimation unit 703. The other configurations and operations are the same as those in the first embodiment, so the same configurations and operations are denoted by the same reference numerals and detailed descriptions thereof will be omitted.

[0064] As shown in Figures 7B to 7D, the wall unevenness shape estimation unit 701 estimates the unevenness shape of the wall of the room 610 based on wall unevenness part point cloud data from the two-dimensional wall candidate point cloud data 721 whose distance from the two-dimensional wall point cloud data 722 is equal to or greater than a predetermined distance T1.

[0065] 7B, as in the first embodiment, the two-dimensional wall candidate point cloud data 721 is created from laser light data, which is data on reflected light of laser light irradiated on the inner surface of the room 610. Specifically, the two-dimensional wall candidate point cloud data 721 is created by orthogonally projecting, onto the floor surface, point cloud data whose normal direction is horizontal, among point cloud data which is position data of reflection points of laser light.

[0066] The two-dimensional wall candidate point cloud data 721 includes point cloud data 721a corresponding to the wall of the living room 611 and point cloud data 721b corresponding to the wall of the entrance 612. However, if a line detection algorithm is applied directly to the two-dimensional wall candidate point cloud data 721, the point cloud data 721b corresponding to the wall of the entrance 612 will be outlier data, as shown in Fig. 7B, and the detected two-dimensional wall point cloud data 722 will not reflect the shape of the wall of the entrance 612.

[0067] On the other hand, when the point cloud data 721b is wall unevenness point cloud data whose distance from the two-dimensional wall point cloud data 722 is equal to or greater than the predetermined distance T1, the wall unevenness shape estimation unit 701 adopts the point cloud data 721b as point cloud data to be used for creating two-dimensional wall point cloud data, and estimates the uneven shape of the wall of the room 610. A method by which the wall unevenness shape estimation unit 701 estimates the uneven shape of the wall of the room 610 will be described with reference to Fig. 7C.

[0068] First, point cloud data having the same normal direction as point cloud data 721b, specifically point cloud data having a normal direction in the X1 direction, is selected from the two-dimensional wall point cloud data 722. In FIG. 7C, point cloud data having the same normal direction as point cloud data 721b are indicated by black circles, and point cloud data having a normal direction different from that of point cloud data 721b are indicated by white circles. Then, a straight line 724 represented by the selected two-dimensional wall point cloud data 722 is generated. The straight line 724 may be, for example, the straight line with the highest degree of match in (2-1-5) of the above-mentioned <2-1>. Alternatively, the straight line 724 may be a straight line detected by applying the least squares method to the two-dimensional wall point cloud data 722 obtained in <2-2>.

[0069] Next, perpendicular lines (L1, L2, L3, ···) are drawn from each point (P1, P2, P3, ···) of the point group data 721b to the straight line 724. Then, it is determined whether the length of the perpendicular line is equal to or greater than a predetermined distance T1. If the length of the perpendicular line is equal to or greater than the predetermined distance T1, the point is determined to be wall unevenness part point group data and is adopted as the point group data used for creating the two-dimensional wall point group data. In FIG. 7C, the length of the perpendicular line L1 is less than the predetermined distance T1 (L1 < T1), the length of the perpendicular line L2 is equal to the predetermined distance T1 (L2 = T1), and the length of the perpendicular line L3 is greater than the predetermined distance T1 (L3 > T1). Therefore, the points P2 and P3 are wall unevenness part point group data and are adopted as the point group data used for creating the two-dimensional wall point group data. The point P1 is not wall unevenness part point group data and is not adopted as the point group data used for creating the two-dimensional wall point group data. Note that the predetermined distance T1 can be set arbitrarily. For example, the length of 10% of the maximum length between the walls 113, 113 facing each other across the wall 113 whose uneven shape is to be estimated may be set as the predetermined distance T1.

[0070] Also, the wall unevenness part point group data may be points within a certain range of the distance from the straight line 724. Specifically, a point where the length of the perpendicular line is equal to or greater than the predetermined distance T1 and equal to or less than the predetermined distance S1 (T1 ≤ L ≤ S1) may be determined to be wall unevenness part point group data. The predetermined distance S1 can be set arbitrarily. The predetermined distance S1 may be, for example, the length of 20% of the maximum length between the walls 113, 113 facing each other across the wall 113 whose uneven shape is to be estimated.

[0071] Then, for the point group data whose normal direction is the X2 direction and the point group data whose normal direction is the X3 direction, the straight line and the uneven shape are estimated in the same manner as the point group data whose normal direction is the X1 direction.

[0072] Based on the point group data 721b determined to be wall unevenness part point group data, the wall unevenness part estimating unit 701 detects two-dimensional wall point group data 725 reflecting the uneven shape of the wall 113, that is, the shape of the wall of the entrance 612, as shown in FIG. 7D.

[0073] It is not necessary to determine whether all points in the two-dimensional wall candidate point cloud data 721 are wall irregularity point cloud data. For example, as shown in FIG. 7E, it may be estimated that wall irregularity point cloud data 727 exists at a certain width on a line connecting an end 726a of point cloud data 726 corresponding to a wall 728a on the entrance side of the entrance 612 and an end 722a of the two-dimensional wall point cloud data 722 detected by a line detection algorithm. The wall irregularity point cloud data 727 is point cloud data corresponding to a wall 728b located on the left side when entering the entrance 612. Estimating the wall irregularity point cloud data 727 in this manner reduces the number of times that it is determined whether a point is wall irregularity point cloud data. Therefore, the number of calculations performed by the wall irregularity shape estimation unit 701 can be reduced, thereby reducing the load on the wall irregularity shape estimation unit 701.

[0074] Furthermore, depending on the positional relationship between the shape of the room 610 and the laser sensor 120, there may be areas that are shaded by the wall 113 of the room 610 and are not irradiated with the laser light. For example, if the laser sensor 120 is disposed in the center of the living room 611, the wall 728b will be shaded by the wall 113 of the living room 611 and will not be irradiated with the laser light. Therefore, point cloud data corresponding to the wall 728b cannot be acquired. Even in such a case, by estimating the wall unevenness point cloud data 727 as described above, it is possible to estimate point cloud data corresponding to the area corresponding to the wall 728b, i.e., the area for which point cloud data could not be acquired.

[0075] Note that, when an uneven shape exists midway along a wall as shown in Fig. 7F, the uneven shape of the wall may be estimated by taking into account the distance between points in the point cloud data in a direction parallel to the wall. Specifically, the wall uneven shape estimation unit 701 may estimate the uneven shape of the wall by taking into account the distance between adjacent points in the two-dimensional wall point cloud data 722 in a direction parallel to a straight line 724. This will be specifically described below with reference to Fig. 7F.

[0076] The wall unevenness shape estimation unit 701 calculates the distance (K1, K2, K3) between each point (E1, E2, E3, E4, ...) of the point cloud data in the two-dimensional wall point cloud data 722, whose normal direction is the X1 direction, and the adjacent point in the direction parallel to the line 724. Next, the calculated distance (K1, K2, K3) between the two points is calculated as a predetermined distance K Th Determine whether the distance is greater than the predetermined distance K. Th can be set arbitrarily. Th may be, for example, the interval (pitch) between points in the point cloud data acquired by the laser sensor 120.

[0077] The wall unevenness shape estimation unit 701 determines whether the distance between the two points is a predetermined distance K Th If the distance between the two points is equal to or less than a predetermined distance K, the two points are determined to be point cloud data corresponding to the same continuous wall. Th 7F, the distance K1 between the point E1 and the point E2 is greater than the predetermined distance K Th The distance K2 between the points E2 and E3 is the predetermined distance K Th The distance K3 between the points E3 and E4 is greater than the predetermined distance K Th 7F is taken as the up-down direction, point cloud data 722a above point E2 and point cloud data 722b below point E3 of the two-dimensional wall point cloud data 722 are determined to be point cloud data corresponding to separate, discontinuous walls. Therefore, it is estimated that the wall represented by point cloud data 722a and the wall represented by point cloud data 722b are spaced apart, and that there is an uneven shape between them.

[0078] Next, as described above, the wall unevenness shape estimation unit 701 determines whether the point cloud data 721b, which is outlier data with respect to the line 724, is wall unevenness shape point cloud data based on the predetermined distance T1. As a result, three walls represented by the lines 724a, 724b, and 724c are estimated from the point cloud data whose normal direction is the X1 direction.

[0079] Thereafter, for the point cloud data whose normal line direction is the X2 direction and the point cloud data whose normal line direction is the X3 direction, estimation of the lines and the uneven shape is performed in the same manner as for the point cloud data whose normal line direction is the X1 direction. As a result, a line 724d is estimated from the point cloud data whose normal line direction is the X2 direction, and a line 724e is estimated from the point cloud data whose normal line direction is the X3 direction. Then, the uneven shape of the wall is estimated by connecting the lines 724a, 724b, 724c, 724d, and 724e.

[0080] As shown in Figures 7G to 7I, the floor unevenness shape estimation unit 702 estimates the unevenness shape of the floor 111 of the room 610 based on floor unevenness part point cloud data from the floor candidate point cloud data 741 whose distance from the floor point cloud data 742 is equal to or greater than a predetermined distance T2.

[0081] 7G, similarly to the first embodiment, the floor candidate point cloud data 741 is created from laser light data, which is data on reflected light of laser light irradiated on the inner surface of the room 610. Specifically, among the point cloud data, which is position data of the reflection points of the laser light, point cloud data whose normal direction is vertically upward is the floor candidate point cloud data 741.

[0082] The floor candidate point cloud data 741 includes point cloud data 741a corresponding to the floor 111a of the living room 611 and point cloud data 741b corresponding to the tiled floor 111b of the entrance 612. However, if a floor and ceiling point cloud data detection algorithm is applied directly to the floor candidate point cloud data 741, the point cloud data 741b corresponding to the tiled floor 111b of the entrance 612 will be outlier data, and the detected floor point cloud data 742 will not reflect the shape of the tiled floor 111b of the entrance 612.

[0083] On the other hand, when the point cloud data 741b is floor unevenness point cloud data whose distance from the floor point cloud data 742 is equal to or greater than a predetermined distance T2, the floor unevenness shape estimation unit 702 adopts the point cloud data 741b as the point cloud data to be used for creating the floor point cloud data, and estimates the unevenness shape of the floor of the interior 610. A method by which the floor unevenness shape estimation unit 702 estimates the unevenness shape of the floor of the interior 610 will be described while referring to FIG. 7H.

[0084] First, a plane 744 represented by the floor point cloud data 742 is generated. The plane 744 may be, for example, the plane with the highest degree of coincidence in (1-1-5) of <1-1> described above. Further, the plane 744 may be a plane detected by applying the least squares method to the floor point cloud data 742 obtained by <1-2>.

[0085] Next, the plane 744 and the point cloud data 741b are orthogonally projected onto the wall 113. In FIG. 7H, the orthogonal projection is performed toward the back side of the paper surface. Then, perpendicular lines (M1, M2, M3, ···) are drawn from each point (Q1, Q2, Q3, ···) of the point cloud data 741b that has been orthogonally projected to the straight line generated by orthogonally projecting the plane 744. Thereafter, it is determined whether or not the length of the perpendicular line is equal to or greater than a predetermined distance T2. When the length of the perpendicular line is equal to or greater than the predetermined distance T2, the point is determined to be floor unevenness point cloud data and is adopted as the point cloud data to be used for creating the floor point cloud data. In FIG. 7H, the length of the perpendicular line M1 is less than the predetermined distance T2 (M1 < T2), the length of the perpendicular line M2 is equal to the predetermined distance T2 (M2 = T2), and the length of the perpendicular line M3 is greater than the predetermined distance T2 (M3 > T2). Therefore, the points Q2 and Q3 are floor unevenness point cloud data and are adopted as the point cloud data to be used for creating the floor point cloud data. The point Q1 is not floor unevenness point cloud data and is not adopted as the point cloud data to be used for creating the floor point cloud data. The predetermined distance T2 can be arbitrarily set. For example, the height of the entrance 612, that is, the length of 10% of the distance from the concrete floor 111b of the entrance 612 to the ceiling 112 may be set as the predetermined distance T2.

[0086] Furthermore, the floor unevenness point cloud data may be points whose distance to a straight line generated by orthogonally projecting the plane 744 is within a certain range. Specifically, points whose perpendicular length is equal to or greater than a predetermined distance T2 and equal to or less than a predetermined distance S2 (T2≦M≦S2) may be determined to be floor unevenness point cloud data. The predetermined distance S2 can be set arbitrarily. For example, the predetermined distance S2 may be set to the height of the entrance 612, that is, 20% of the distance from the tatami mat 111b of the entrance 612 to the ceiling 112.

[0087] Based on the point cloud data 741b determined to be floor unevenness point cloud data, the floor unevenness shape estimation unit 702 detects floor point cloud data 745 that reflects the uneven shape of the floor 111, i.e., the shape of the floor of the entrance 612, as shown in Figure 7I.

[0088] As described above, the point cloud data 211 on which the floor point cloud data 745 is based is preferably acquired by a worker or the like moving around while holding the laser sensor 120 and scanning the laser light toward each inner surface of the room 110. However, depending on the placement position of the laser sensor 120 and the shape of the room 110, it may be difficult to irradiate all points on each inner surface with the laser light, and there may be some points where the point cloud data 211 cannot be acquired. For example, if the laser sensor 120 is placed in the center of the living room 611, the frame 611a may be shaded by the floor 111a and may not be irradiated with the laser light. In this case, the floor point cloud data 745 may not include point cloud data corresponding to the frame 611a. However, the point cloud data corresponding to the frame 611a can be estimated based on the floor point cloud data 745. Specifically, it can be estimated that point cloud data corresponding to the frame 611a exists at a certain width on a line connecting an end 746a of point cloud data 746 that is relatively low in height position and an end 747a of point cloud data 747 that is relatively high in height position among the floor point cloud data 745. The point cloud data 746 that is relatively low in height position is point cloud data that corresponds to the tatami mat 111b of the entrance 612. The point cloud data 747 that is relatively high in height position is point cloud data that corresponds to the floor 111a of the living room 611.

[0089] Furthermore, even if point cloud data corresponding to the frame 611a can be acquired, the normal direction of the point cloud data is horizontal, so if classification is based only on the normal direction, the point cloud data will not be included in the floor candidate point cloud data 741. As a result, the point cloud data corresponding to the frame 611a will not be included in the floor point cloud data 745. In this case as well, the point cloud data corresponding to the frame 611a can be estimated based on the floor point cloud data 745, as described above.

[0090] The floor unevenness estimation unit 702 may estimate the unevenness of the floor by taking into account the distance between adjacent points in the floor point cloud data 742 in a direction parallel to a straight line generated by orthogonally projecting the plane 744. As a method for estimating the unevenness of the floor by taking into account the distance, a method similar to the method used by the wall unevenness estimation unit 701 to estimate the unevenness of the wall by taking into account the distance between adjacent points in the two-dimensional wall point cloud data 722 in a direction parallel to the straight line 724 can be used.

[0091] As shown in Figures 7J to 7L, the ceiling unevenness shape estimation unit 703 estimates the unevenness shape of the ceiling 112 in the room 610 based on the ceiling unevenness part point cloud data among the ceiling candidate point cloud data 731 whose distance from the ceiling point cloud data 732 is equal to or greater than a predetermined distance.

[0092] 7J, similarly to the first embodiment, the ceiling candidate point cloud data 731 is created from laser light data, which is data on reflected light of laser light irradiated on the inner surface of the room 610. Specifically, among the point cloud data, which is position data of reflection points of the laser light, point cloud data whose normal direction is vertically downward is the ceiling candidate point cloud data 731.

[0093] The ceiling candidate point cloud data 731 includes point cloud data 731a corresponding to parts of the ceiling 112 other than the beams 613, and point cloud data 731b corresponding to the beams 613. However, if the floor / ceiling point cloud data detection algorithm is applied directly to the ceiling candidate point cloud data 731, the point cloud data 731b corresponding to the beams 613 will be outlier data, and the detected ceiling point cloud data 732 will not reflect the shape of the beams 613.

[0094] On the other hand, if the point cloud data 731b is ceiling unevenness point cloud data whose distance from the ceiling point cloud data 732 is equal to or greater than the predetermined distance T3, the ceiling unevenness shape estimation unit 703 adopts the point cloud data 731b as point cloud data to be used in creating ceiling point cloud data, and estimates the uneven shape of the ceiling of the room 610. The method by which the ceiling unevenness shape estimation unit 703 estimates the uneven shape of the ceiling of the room 610 will be described with reference to Fig. 7K.

[0095] First, a plane 734 represented by the ceiling point cloud data 732 is generated. The plane 734 may be, for example, the plane with the highest degree of match in (1-1-5) of <1-1> above. Alternatively, the plane 734 may be a plane detected by applying the least squares method to the ceiling point cloud data 732 obtained by <1-2>.

[0096] Next, project the plane 734 and the point cloud data 731b orthogonally onto the wall 113. In Fig. 7K, the projection is made towards the back side of the paper plane. Then, draw perpendicular lines (N1, N2, N3, ···) from each point (R1, R2, R3, ···) of the projected point cloud data 731b to the straight line generated by projecting the plane 734. Thereafter, determine whether the length of the perpendicular line is greater than or equal to a predetermined distance T3. If the length of the perpendicular line is greater than or equal to the predetermined distance T3, determine that the point is ceiling unevenness point cloud data and adopt it as the point cloud data to be used for creating the ceiling point cloud data. In Fig. 7K, the length of the perpendicular line N1 is less than the predetermined distance T3 (N1 < T3), the length of the perpendicular line N2 is the same as the predetermined distance T3 (N2 = T3), and the length of the perpendicular line N3 is greater than the predetermined distance T3 (N3 > T3). Therefore, the points R2 and R3 are ceiling unevenness point cloud data and are adopted as the point cloud data to be used for creating the ceiling point cloud data. The point R1 is not ceiling unevenness point cloud data and is not adopted as the point cloud data to be used for creating the ceiling point cloud data. Note that the predetermined distance T3 can be set arbitrarily. For example, the height of the ceiling 112, that is, a length of 10% of the distance from the floor 111 to the ceiling 112 may be set as the predetermined distance T3.

[0097] Also, the ceiling unevenness point cloud data may be points within a certain range of the distance from the straight line generated by projecting the plane 734. Specifically, a point where the length of the perpendicular line is greater than or equal to the predetermined distance T3 and less than or equal to the predetermined distance S3 (T3 ≤ M ≤ S3) may be determined as the ceiling unevenness point cloud data. The predetermined distance S3 can be set arbitrarily. The predetermined distance S3 may be, for example, a length of 20% of the height of the ceiling 112, that is, the distance from the floor 111 to the ceiling 112.

[0098] Based on the point cloud data 731b determined to be the ceiling unevenness point cloud data, the ceiling unevenness estimating unit 703 detects the ceiling point cloud data 735 reflecting the uneven shape of the ceiling 112, that is, the shape of the beam 613, as shown in Fig. 7L.

[0099] The ceiling point cloud data 735 is detected based on the ceiling candidate point cloud data 731, whose normal direction is vertically downward. Therefore, the ceiling point cloud data 735 does not include point cloud data corresponding to the side surfaces 739a and 739b of the beam 613. However, it is possible to estimate the point cloud data corresponding to the side surfaces 739a and 739b of the beam 613 based on the ceiling point cloud data 735. Specifically, it is possible to estimate that the point cloud data corresponding to the side surfaces 739a and 739b of the beam 613 exist within a certain width on the shortest straight line connecting the ends 736a and 737a of the point cloud data 736 and 737, which are relatively high in height, and the ends 738a and 738b of the point cloud data 738, which is relatively low in height, in the ceiling point cloud data 735. The shortest straight line is a straight line connecting an end 736a of the point cloud data 736 and one end 738a of the point cloud data 738, and a straight line connecting an end 737a of the point cloud data 737 and the other end 738b of the point cloud data 738. The point cloud data 736 and 737 at relatively high height positions are point cloud data corresponding to portions of the ceiling 112 other than the beams 613. The point cloud data 738 at a relatively low height position is point cloud data corresponding to the lower surface (ceiling surface) 739c of the beam 613.

[0100] In this embodiment as well, the wall point cloud data detection unit 205 generates wall point cloud data by duplicating the two-dimensional wall point cloud data 725 so that the same wall contour line continues from the floor surface generated by the floor point cloud data 745 to the ceiling surface generated by the ceiling point cloud data 735. In this embodiment, the two-dimensional wall point cloud data 725 reflects the uneven shape of the wall 113 of the room 610. Therefore, the generated wall point cloud data reflects the uneven shape of the wall 113 of the room 610.

[0101] The ceiling unevenness estimation unit 703 may estimate the unevenness of the ceiling by taking into account the distance between adjacent points in the ceiling point cloud data 734 in a direction parallel to a straight line generated by orthogonally projecting the plane 734. As a method for estimating the unevenness of the ceiling by taking into account the distance, a method similar to the method used by the wall unevenness estimation unit 701 to estimate the unevenness of the wall by taking into account the distance between adjacent points in the two-dimensional wall point cloud data 722 in a direction parallel to the straight line 724 can be used.

[0102] The three-dimensional shape generating unit 206 of this embodiment generates a three-dimensional shape 650 of the room 610 from the floor point cloud data 745, the ceiling point cloud data 735, and the wall point cloud data. Specifically, the shape surrounded by the floor surface generated by the floor point cloud data, the ceiling surface generated by the ceiling point cloud data 232, and the wall surfaces generated by the wall point cloud data is defined as the three-dimensional shape 650 of the room 610. As described above, the wall point cloud data reflects the uneven shape of the walls 113 of the room 610. Furthermore, the floor point cloud data 745 reflects the uneven shape of the floor 111 of the room 610. Furthermore, the ceiling point cloud data 735 reflects the uneven shape of the ceiling 112 of the room 610. Therefore, the generated three-dimensional shape 650 reflects the uneven shape of the room 610.

[0103] FIG. 8 is a diagram illustrating an example of an unevenness estimation table 801 included in the indoor 3D shape generation device 600. The unevenness estimation table 801 is a table that stores a predetermined distance 812 in association with an estimation target 811 having an uneven shape. The estimation target 811 is the floor 111, the ceiling 112, or the wall 113 of the room 610. The predetermined distance 812 is a threshold value of the length of a perpendicular line used when the wall unevenness estimation unit 701, the floor unevenness estimation unit 702, or the ceiling unevenness estimation unit 703 determines the wall unevenness point cloud data, the floor unevenness point cloud data, or the ceiling unevenness point cloud data. In this embodiment, "T1," "T2," and "T3" are set as the predetermined distances 812 for the wall 113, the floor 111, and the ceiling 112, respectively. The predetermined distances 812 may be different values ​​for each estimation target 811, or may be the same value. The predetermined distance 812 may be the same value for each estimation target 811. The wall unevenness estimation unit 701, the floor unevenness estimation unit 702, and the ceiling unevenness estimation unit 703 each refer to the unevenness estimation table 801 to estimate the unevenness shape.

[0104] The hardware configuration of the indoor 3D shape generating device 600 will be described with reference to Fig. 9. The RAM 940 is a random access memory used by the CPU 410 as a work area for temporary storage. A storage area for storing data necessary for implementing this embodiment is secured in the RAM 940. The unevenness point cloud data 941 is wall unevenness point cloud data, floor unevenness point cloud data, or ceiling unevenness point cloud data estimated by the wall unevenness shape estimation unit 701, floor unevenness shape estimation unit 702, or ceiling unevenness shape estimation unit 703.

[0105] The storage 950 stores a database, various parameters, or the following data or programs required to implement this embodiment. The storage 950 stores an unevenness estimation table 801. The unevenness estimation table 801 is a table that manages the relationship between an estimation target 811 and a predetermined distance 812, as shown in FIG. 8.

[0106] The storage 950 further stores a wall unevenness estimation module 951, a floor unevenness estimation module 952, and a ceiling unevenness estimation module 953. The wall unevenness estimation module 951 is a module that estimates the unevenness of the wall 113. The floor unevenness estimation module 952 is a module that estimates the unevenness of the floor 111. The ceiling unevenness estimation module 953 is a module that estimates the unevenness of the ceiling 112. These modules 951 to 953 are read into the application execution area 447 of the RAM 940 by the CPU 410 and executed.

[0107] Next, the processing procedure of indoor three-dimensional shape generation device 600 will be described with reference to the flowchart shown in Fig. 10. This flowchart is executed by CPU 410 in Fig. 9 using RAM 940, and realizes each functional configuration of indoor three-dimensional shape generation device 600 in Fig. 7A.

[0108] In step S1001, the wall point cloud data detection unit 205 detects the two-dimensional wall candidate point cloud data 721. In step S1003, the wall unevenness shape estimation unit 701, the floor unevenness shape estimation unit 702, and the ceiling unevenness shape estimation unit 703 estimate the unevenness shapes of the wall 113, the floor 111, and the ceiling 112, respectively, and detect the two-dimensional wall point cloud data, floor point cloud data 745, and ceiling point cloud data 735 in which the unevenness shapes are reflected. In step S1005, the wall point cloud data detection unit 205 detects the wall point cloud data based on the two-dimensional wall point cloud data, floor point cloud data 745, and ceiling point cloud data 735 in which the unevenness shapes are reflected.

[0109] According to this embodiment, it is possible to generate a three-dimensional shape 650 that reflects the uneven shape of the room 610. Therefore, according to this embodiment, it is possible to generate a three-dimensional shape 650 with even higher accuracy.

[0110] Although the present invention has been described above with reference to the embodiments, the present invention is not limited to the above-described embodiments and can be modified as appropriate. Various modifications that can be understood by those skilled in the art can be made to the configuration and details of the present invention within the scope of the present invention. Furthermore, systems or devices that combine separate features included in each embodiment in any manner are also included in the scope of the present invention.

[0111] The present invention may also be applied to a system consisting of multiple devices or to a single device. Furthermore, the present invention may also be applied when an information processing program that realizes the functions of the embodiments is supplied to a system or device and executed by a built-in processor. Therefore, the technical scope of the present invention also includes a program installed on a computer to realize the functions of the present invention, a medium storing the program, a WWW (World Wide Web) server from which the program is downloaded, and a processor that executes the program. In particular, the technical scope of the present invention also includes a non-transitory computer-readable medium storing a program that causes a computer to execute at least the processing steps included in the above-described embodiments.

Claims

1. a laser light data acquisition unit that acquires laser light data including a reflection time and a reflection direction of a laser light reflected from an inner surface of a room of a building; a point cloud data creation unit that creates point cloud data of the room from the laser light data; a classification unit that classifies the point cloud data into ceiling candidate point cloud data, floor candidate point cloud data, and wall candidate point cloud data according to the direction of a normal at each point included in the point cloud data; a floor and ceiling point cloud data detection unit that applies a floor and ceiling point cloud data detection algorithm to the floor candidate point cloud data and the ceiling candidate point cloud data to detect floor point cloud data and ceiling point cloud data in the room; a wall point cloud data detection unit that applies a wall point cloud data detection algorithm to the wall candidate point cloud data to detect wall point cloud data in the room; a three-dimensional shape generation unit that generates a three-dimensional shape of the room from the floor point cloud data, the ceiling point cloud data, and the wall point cloud data; An indoor 3D shape generation device equipped with the device.

2. The wall point cloud data detection unit As the wall point cloud data detection algorithm, a line detection algorithm is applied to two-dimensional wall candidate point cloud data obtained by orthogonally projecting the wall candidate point cloud data onto a floor surface generated by the floor point cloud data to extract two-dimensional wall point cloud data; The indoor three-dimensional shape generation device according to claim 1, wherein the two-dimensional wall point cloud data is duplicated so that the same wall contour line continues from the floor surface to the ceiling surface generated by the ceiling point cloud data, thereby obtaining the wall point cloud data.

3. 3. The indoor three-dimensional shape generation device according to claim 2, wherein the line detection algorithm extracts the two-dimensional wall point cloud data based on a degree of coincidence between a line passing through two points arbitrarily selected from the two-dimensional wall candidate point cloud data and the two-dimensional wall candidate point cloud data.

4. The floor and ceiling point cloud data detection algorithm Detecting the floor point cloud data based on a degree of coincidence between a plane passing through three points arbitrarily selected from the floor candidate point cloud data and the floor candidate point cloud data; The indoor three-dimensional shape generation device according to claim 1, wherein the ceiling point cloud data is detected based on the degree of coincidence between a plane passing through three points arbitrarily selected from the ceiling candidate point cloud data and the ceiling candidate point cloud data.

5. a wall unevenness shape estimation unit that estimates an uneven shape of the wall inside the room based on wall unevenness part point cloud data that is a distance from the two-dimensional wall point cloud data that is equal to or greater than a predetermined distance from the two-dimensional wall candidate point cloud data, among the two-dimensional wall candidate point cloud data; a floor unevenness shape estimation unit that estimates an uneven shape of the indoor floor based on floor unevenness part point cloud data that is a distance from the floor point cloud data that is a predetermined distance or more from the floor candidate point cloud data; a ceiling irregularity shape estimation unit that estimates the irregularity shape of the ceiling in the room based on ceiling irregularity point cloud data that is a predetermined distance or more from the ceiling point cloud data among the ceiling candidate point cloud data, The indoor three-dimensional shape generating device according to claim 2 further comprising:

6. a laser light data acquisition step of acquiring laser light data including a reflection time and a reflection direction of reflected light of the laser light irradiated into the interior of the building; a point cloud data creation step of creating point cloud data of the room from the laser light data; a classification step of classifying the point cloud data into ceiling candidate point cloud data, floor candidate point cloud data, and wall candidate point cloud data according to the direction of a normal at each point included in the point cloud data; a floor and ceiling point cloud data detection step of applying a floor and ceiling point cloud data detection algorithm to the floor candidate point cloud data and the ceiling candidate point cloud data to detect floor point cloud data and ceiling point cloud data in the room; a wall point cloud data detection step of applying a wall point cloud data detection algorithm to the wall candidate point cloud data to detect wall point cloud data in the room; a three-dimensional shape generation step of generating a three-dimensional shape of the room from the floor point cloud data, the ceiling point cloud data, and the wall point cloud data; A method for generating a three-dimensional indoor shape, comprising:

7. a laser light data acquisition step of acquiring laser light data including a reflection time and a reflection direction of reflected light of the laser light irradiated into the interior of the building; a point cloud data creation step of creating point cloud data of the room from the laser light data; a classification step of classifying the point cloud data into ceiling candidate point cloud data, floor candidate point cloud data, and wall candidate point cloud data according to the direction of a normal at each point included in the point cloud data; a floor and ceiling point cloud data detection step of applying a floor and ceiling point cloud data detection algorithm to the floor candidate point cloud data and the ceiling candidate point cloud data to detect floor point cloud data and ceiling point cloud data in the room; a wall point cloud data detection step of applying a wall point cloud data detection algorithm to the wall candidate point cloud data to detect wall point cloud data in the room; a three-dimensional shape generation step of generating a three-dimensional shape of the room from the floor point cloud data, the ceiling point cloud data, and the wall point cloud data; A program for generating three-dimensional indoor shapes that causes a computer to execute the above.

Images