Point cloud acquisition method, point cloud acquisition device, and point cloud acquisition program

Abstract

The present invention provides a point cloud acquisition method, a point cloud acquisition device, and a point cloud acquisition program capable of suitably acquiring the shape of the planar portion of a structure. [Solution] The point cloud acquisition method comprises the steps of: S103 to generate a plurality of three-dimensional Dornay triangles T2 with each of the three laser reflection points as its vertex; S104 to calculate a plane containing each of the plurality of three-dimensional Dornay triangles T2 based on the equation of a plane; S105 to calculate the normal vector V of each of the plurality of planes and to find the frequently occurring normal vector Vf of the plurality of normal vectors V such that the frequency of angle appearance peaks at Cp; and S107 to S109 to acquire a plane reflection point cloud by detecting a reference point cloud G1 containing a plurality of laser reflection points P that constitute the three-dimensional Dornay triangles T2 included in the reference plane Rp which is the plane defined by the frequently occurring normal vector Vf.

Description

[Technical Field]

[0001] This invention relates to a point cloud acquisition method, a point cloud acquisition device, and a point cloud acquisition program. [Background technology]

[0002] Patent Document 1 describes a method for object detection using a laser point cloud. In this method, a computer device receives laser data representing the vehicle's environment, and the laser data includes multiple data points associated with one or more objects in the environment. The method also generates a two-dimensional depth image based on the multiple data points, which includes multiple pixels indicating the respective positions of one or more objects in the environment relative to the vehicle. The generated two-dimensional depth image is modified to assign values ​​to one or more predetermined pixels of the multiple pixels that map one or more parts of one or more objects in the environment where laser data is missing. Modifying the generated two-dimensional depth image to assign values ​​to one or more predetermined pixels is based on the neighboring pixels of the multiple pixels located proximal to one or more predetermined pixels of the multiple pixels in the two-dimensional depth image. Furthermore, based on the modified two-dimensional depth image, the method determines multiple normal vectors for one or more sets of pixels in the multiple pixels corresponding to each surface of one or more objects in the environment, and provides object recognition information indicating one or more objects in the environment to one or more systems of the vehicle based on the multiple normal vectors of the one or more sets of pixels. [Prior art documents] [Patent Documents]

[0003] [Patent Document 1] Japanese Patent Publication No. 2017-152049 [Overview of the Initiative] [Problems that the invention aims to solve]

[0004] The method described in Patent Document 1 above attempts to detect objects around a vehicle using a laser reflection point cloud. In contrast, there is currently a demand for a technology to acquire the external shape of artificial structures such as buildings. Artificial structures may have flat surfaces, such as roofs and walls. Therefore, there is a demand for a technology that can suitably acquire the shape of such flat surfaces using a laser reflection point cloud.

[0005] Therefore, the present invention aims to provide a point cloud acquisition method, a point cloud acquisition device, and a point cloud acquisition program that can suitably acquire the shape of the planar portion of a structure. [Means for solving the problem]

[0006] The point cloud acquisition method according to the present invention is [1] "a point cloud acquisition method for acquiring a plane reflection point cloud corresponding to a plane portion of a structure from a laser reflection point cloud of a target region including a plurality of laser reflection points having three-dimensional coordinates consisting of X coordinates, Y coordinates, and Z coordinates, comprising: a triangle generation step of generating a plurality of three-dimensional Dornay triangles with each of the three laser reflection points as its vertex; a plane calculation step after the triangle generation step of calculating a plane including each of the plurality of three-dimensional Dornay triangles based on the equation of a plane; a normal vector calculation step after the plane calculation step of calculating the normal vectors of each of the plurality of planes and finding the most frequent normal vector that has the peak frequency of appearance among the plurality of normal vectors; and a point cloud acquisition step after the vector calculation step of acquiring the plane reflection point cloud by detecting a reference point cloud including a plurality of laser reflection points constituting the three-dimensional Dornay triangles included in a reference plane which is the plane defined by the most frequent normal vector."

[0007] The point group acquisition device according to the present invention is a point group acquisition device for obtaining a plane reflection point group corresponding to a flat part of a structure from a laser reflection point group of a target area including a plurality of laser reflection points having three-dimensional coordinates composed of an X coordinate, a Y coordinate, and a Z coordinate, the point group acquisition device including: a triangle generation unit that generates a plurality of three-dimensional Delaunay triangles each having one of the three laser reflection points as a vertex; a plane calculation unit that calculates a plane including each of the plurality of three-dimensional Delaunay triangles based on the equation of a plane; a normal vector calculation unit that calculates a normal vector of each of the plurality of planes and obtains a frequently occurring normal vector having a peak in the appearance frequency among the plurality of normal vectors; and a point group acquisition unit that obtains the plane reflection point group by detecting a reference point group including the plurality of laser reflection points that constitute the three-dimensional Delaunay triangles included in the reference plane that is the plane defined by the frequently occurring normal vector.

[0008] The point group acquisition program according to the present invention is a point group acquisition program for causing a computer to function as a point group acquisition device for obtaining a plane reflection point group corresponding to a flat part of a structure from a laser reflection point group of a target area including a plurality of laser reflection points having three-dimensional coordinates composed of an X coordinate, a Y coordinate, and a Z coordinate, the program causing the computer to realize: a triangle generation function that generates a plurality of three-dimensional Delaunay triangles each having one of the three laser reflection points as a vertex; a plane calculation function that calculates a plane including each of the plurality of three-dimensional Delaunay triangles based on the equation of a plane; a normal vector calculation function that calculates a normal vector of each of the plurality of planes and obtains a frequently occurring normal vector having a peak in the appearance frequency among the plurality of normal vectors; and a point group acquisition function that obtains the plane reflection point group by detecting a reference point group including the plurality of laser reflection points that constitute the three-dimensional Delaunay triangles included in the reference plane that is the plane defined by the frequently occurring normal vector.

[0009] These methods, apparatuses, and programs generate three-dimensional Dornay triangles from a group of laser reflection points in a target region, each having three-dimensional coordinates consisting of X, Y, and Z coordinates. A plane containing each of these three-dimensional Dornay triangles is then calculated based on the equation of a plane. The normal vectors for each of these planes are then calculated. Generally, on the flat surfaces of artificial structures, such as roofs and walls, multiple planes corresponding to numerous three-dimensional Dornay triangles face the same direction, whereas on the surfaces of natural objects such as trees, the direction of these planes tends to be random. In other words, on the flat surfaces of structures, the angles of the normal vectors of these planes tend to be the same. Therefore, the most frequent normal vector, where the angle of occurrence peaks among multiple normal vectors, is determined. Then, by detecting a group of reference points containing multiple laser reflection points that constitute the three-dimensional Dornay triangles included in the reference plane defined by the most frequent normal vector, a group of plane reflection points corresponding to the flat surfaces of a structure is obtained. Therefore, by utilizing this planar reflection point cloud, it becomes possible to suitably obtain the shape of the planar portion of a structure. As a result, it becomes possible to estimate the external shape of a structure or to automatically generate an external shape drawing of a structure.

[0010] The point cloud acquisition method according to the present invention may also be [2] "the point cloud acquisition method according to [1] above, comprising a projection step of acquiring a two-dimensional laser reflection point cloud by projecting the laser reflection point cloud onto an XY plane including the X coordinate and the Y coordinate before the triangle generation step, wherein the triangle generation step includes a first generation step of generating a plurality of two-dimensional Dornay triangles whose vertices are each of the three laser reflection points included in the two-dimensional laser reflection point cloud, and a second generation step of generating a plurality of three-dimensional Dornay triangles by assigning the Z coordinate to each vertex of the plurality of two-dimensional Dornay triangles after the first generation step." In this case, the process can be more easily understood by interposing the two-dimensional laser reflection point cloud and the two-dimensional Dornay triangles when generating the three-dimensional Dornay triangles.

[0011] The point cloud acquisition method according to the present invention may be "[3] an offset calculation step of calculating offsets of each of a plurality of the planes based on the equation of the plane after the plane calculation step and before the point cloud acquisition step, and in the point cloud acquisition step, detecting the reference point cloud including a plurality of the laser reflection points constituting the 3D Dornay triangle included in the reference plane which is the plane defined by the frequent normal vector and the offset, the point cloud acquisition method described in [1] or [2] above". In this case, it becomes possible to separate different planes defined by similar frequent normal vectors from each other.

[0012] The point cloud acquisition method according to the present invention may be "[4] the point cloud acquisition step includes a noise removal step of removing, as noise, the isolated laser reflection points and / or the laser reflection points close to the ground surface among the laser reflection points included in the reference point cloud, the point cloud acquisition method described in any one of [1] to [3] above". In this case, it becomes possible to more accurately acquire the shape of the planar portion of the structure.

[0013] The point cloud acquisition method according to the present invention may be "[5] the point cloud acquisition step includes a proximity reflection point detection step of detecting, as proximity reflection points, the laser reflection points located on the same plane as the reference plane and adjacent to the 3D Dornay triangle included in the reference plane, and including the proximity reflection points in the reference point cloud, the point cloud acquisition method described in any one of [1] to [4] above". In this case, it becomes possible to more accurately acquire the shape of the edge portion of the shape of the planar portion of the structure.

Advantages of the Invention

[0014] According to the present invention, it is possible to provide a point cloud acquisition method, a point cloud acquisition device, and a point cloud acquisition program that can suitably acquire the shape of the planar portion of a structure using a laser reflection point cloud.

Brief Description of the Drawings

[0015] [Figure 1]Figure 1 shows the functional configuration of the point cloud acquisition device according to this embodiment. [Figure 2] Figure 2 is a flowchart showing the point cloud acquisition method according to this embodiment. [Figure 3] This diagram illustrates one step in the point cloud acquisition method shown in Figure 2. [Figure 4] This diagram illustrates one step in the point cloud acquisition method shown in Figure 2. [Figure 5] This diagram illustrates one step in the point cloud acquisition method shown in Figure 2. [Figure 6] Figure 6 is a graph plotting multiple normal vectors calculated in step S105. [Figure 7] Figure 7 is a diagram illustrating one step in the point cloud acquisition method shown in Figure 2, and is an example of a histogram of normal vectors generated with an angle as an argument. [Figure 8] Figure 8 is an example of a histogram showing the distribution of offsets. [Figure 9] Figure 9 is a diagram illustrating one step in the point cloud acquisition method shown in Figure 2. [Figure 10] Figure 10 is a diagram illustrating one step in the point cloud acquisition method shown in Figure 2. [Figure 11] Figure 11 is a diagram illustrating one step in the point cloud acquisition method shown in Figure 2. [Figure 12] Figure 12 is a diagram illustrating one step in the point cloud acquisition method shown in Figure 2. [Figure 13] Figure 13 shows the point cloud acquisition program according to this embodiment. [Modes for carrying out the invention]

[0016] An embodiment will be described below with reference to the drawings. In the descriptions of each drawing, the same or corresponding elements will be denoted by the same reference numeral, and redundant explanations may be omitted.

[0017] Figure 1 shows the functional configuration of the point cloud acquisition device according to this embodiment. The point cloud acquisition device 1 shown in Figure 1 is physically configured as a computer system including a CPU, a main memory consisting of RAM and ROM, an auxiliary storage device consisting of a hard disk and memory, a communication control device, etc. Each functional unit of the point cloud acquisition device 1 can be realized by loading predetermined computer software onto the hardware such as the CPU and main memory of the computer system, thereby operating the communication control device etc. under the control of the CPU, and reading and writing data to the main memory and auxiliary storage device.

[0018] As shown in Figure 1, the point cloud acquisition device 1 functionally comprises a projection unit 2, a triangle generation unit 3, a plane calculation unit 4, a normal vector calculation unit 5, an offset calculation unit 6, and a point cloud acquisition unit 7. The triangle generation unit 3 includes a first generation unit 31 and a second generation unit 32. The point cloud acquisition unit 7g includes a noise reduction unit 71 and a nearby reflection point detection unit 72. The operation of each part of the point cloud acquisition device 1 will be described in detail in the point cloud acquisition method described below.

[0019] Figure 2 is a flowchart showing the point cloud acquisition method according to this embodiment. The point cloud acquisition method shown in Figure 2 is implemented, for example, in the point cloud acquisition device 1 shown in Figure 1. The point cloud acquisition device 1 and point cloud acquisition method (and the point cloud acquisition program described later) according to this embodiment are for acquiring a planar reflection point cloud (point cloud G3) corresponding to the planar portion of a structure from a laser reflection point cloud G0 (see Figure 3: CC BY "Shizuoka Prefecture (2020)") of a target area OA that includes a plurality of laser reflection points P having three-dimensional coordinates consisting of X, Y, and Z coordinates (see Figure 12).

[0020] The laser reflection point cloud G0 shown in Figure 3 can be acquired by laser measurement from above the target area OA using a (manned or unmanned) aircraft. The structure is an artificial object, such as a building, and has a flat surface, such as a roof. The laser reflection point cloud G0 has been acquired in advance. The target area OA can be any area, including, for example, urban areas, forests, or rivers.

[0021] In the point group acquisition method according to this embodiment, first, the projection unit 2 projects the three-dimensional laser reflection point group G0 onto the XY plane including the X-axis and the Y-axis, thereby obtaining a two-dimensional laser reflection point group G0a as shown in FIG. 4 (step S101: projection step). Note that the lower diagram in FIG. 4 shows the two-dimensional laser reflection point group G0a in the target area OA, and the upper diagram in FIG. 4 shows an enlarged view of the area R in the lower diagram.

[0022] Subsequently, the first generation unit 31 of the triangle generation unit 3 generates a plurality of two-dimensional Delaunay triangles T1 having as vertices each of the three laser reflection points P included in the two-dimensional laser reflection point group G0a obtained in step S101 (step S102: first generation step, triangle generation step). As an example, the first generation unit 31 generates a plurality of two-dimensional Delaunay triangles T1 by dividing the two-dimensional laser reflection point group G0a in the target area OA into Delaunay triangles.

[0023] Subsequently, as shown in FIG. 5, the second generation unit 32 of the triangle generation unit 3 assigns (returns) a Z coordinate to each vertex (laser reflection point P) of the plurality of two-dimensional Delaunay triangles T1 generated in step S102, thereby generating a plurality of three-dimensional Delaunay triangles T2 (step S103: second generation step, triangle generation step). In this way, the triangle generation unit 3 generates a plurality of three-dimensional Delaunay triangles T2 having as vertices each of the three laser reflection points P.

[0024] Subsequently, the plane calculation unit 4 calculates a plane including each of the plurality of three-dimensional Delaunay triangles T2 generated in step S03 based on the equation of the plane (ax + by + cz + d = 0) (step S104: plane calculation step). As an example, for a certain three-dimensional Delaunay triangle T2, the plane calculation unit 4 uses the three vertices (laser reflection points P), namely point A(A x ,A y ,A z ), point B(B x ,B y ,B z ), and point C(C x ,C y ,C zThe plane passing through ) is defined by the vector AB(B x -A x ,B y -A y ,B z -A z ) and vector AC(C x -A x ,C y -A y ,C z -A z By the cross product with ), a = (B y -A y )(C z -A z )-(C y -A y )(B z -A z ), b=(B z -A z )(C x -A x )-(C z -A z )(B x -A x ), c=(B x -A x )(C y -A y )-(C x -A x )(B y -A y ), d=-(aA x +bA y +cA z It can be calculated as follows:

[0025] Next, the normal vector calculation unit 5 calculates the normal vectors for each of the multiple planes calculated in step S104, and also finds the most frequent normal vector where the angle appears most frequently among the multiple normal vectors (step S105: normal vector calculation step). In step S105, the normal vector calculation unit 5 first calculates the normal vector (a,b,c) using the coefficients a,b,c of the plane equation calculated in step S104. In step S105, the normal vector calculation unit 5 calculates normal vectors for multiple planes corresponding to multiple 3D Dornay triangles T2.

[0026] Figure 6 is a graph plotting multiple normal vectors calculated in step S105. Figure 6 shows the distribution of the normal vectors V. In step S105, the normal vector calculation unit 5 then generates a histogram as shown in Figure 7, using the angle as an argument. In each of the planar sections, the angles of the normal vectors V of the multiple planes corresponding to the multiple 3D Dournay triangles T2 are the same. Therefore, the histogram detects the peak Cp of the frequency of occurrence of the angle of the normal vector V originating from each of the planar sections. Thus, the normal vector calculation unit 5 can determine the normal vector V that has this peak Cp as the frequently occurring normal vector Vf.

[0027] Here, as shown in Figure 6, if there are frequently occurring normal vectors Va originating from one plane and frequently occurring normal vectors Vb originating from another plane, and their angles are the same, then in a histogram like the one shown in Figure 7, they are detected as a single group of peaks Cp, degenerating from each other (i.e., different planes are not separated). Therefore, in the next step, the offset calculation unit 6 calculates the offset of the plane calculated in step S104 based on the equation of the plane (step S106: offset calculation step). In step S104, the offset calculation unit 6 calculates the offset of each plane as a constant d in the equation of the plane. As a result, as shown in the histogram in Figure 8, the frequently occurring normal vectors Va and Vb originating from each plane are detected separately from each other. Note that in Figure 8, the horizontal axis represents the offset value and the vertical axis represents the frequency.

[0028] Next, as shown in Figure 9, the point cloud acquisition unit 7 detects a reference point cloud G1 that includes multiple laser reflection points P that constitute a three-dimensional Dornay triangle T2, which is included in the reference plane Rp (see Figure 11), which is a plane defined by the frequently occurring normal vector Vf (step S107: point cloud acquisition step). As an example, in step S106, the point cloud acquisition unit 7 can detect the reference point cloud G1 for one of the frequently occurring normal vectors Vf (e.g., frequently occurring normal vector Va) among the multiple frequently occurring normal vectors Vf that have been detected separately from each other by calculating the offset d.

[0029] The reference point cloud G1 shown in Figure 9 may contain laser reflection points P that are unrelated to a single planar area, such as a specific roof of a structure, as noise. Therefore, in the subsequent step, as shown in Figure 10, the noise reduction unit 71 of the point cloud acquisition unit 7 removes isolated laser reflection points P and / or laser reflection points P that are closer to the ground (for example, compared to other laser reflection points P) from the reference point cloud G1 as noise (step S108: noise reduction step). As a result, step S108 generates the point cloud G2 after noise reduction.

[0030] The point cloud G2 generated in this way may define a slightly narrower range compared to the actual planar portion. Therefore, in the subsequent step, as shown in Figure 11, laser reflection points P located on the same plane as the reference plane Rp and adjacent to the three-dimensional Dornay triangle T2 included in the reference plane Rp are detected as nearby reflection points Pn, and these nearby reflection points Pn are included in the point cloud G2 (step S109: nearby reflection point detection step). As a result, in step S109, as shown in Figure 12, a point cloud G3 is generated in which the reference point cloud G1 has undergone noise reduction processing and processing to include nearby reflection points Pn. Point cloud G3 is an example of a planar reflection point cloud corresponding to the planar portion of a structure.

[0031] In other words, the point cloud acquisition unit 7 acquires a plane reflection point cloud (point cloud G3) by detecting a reference point cloud G1 that includes multiple laser reflection points P that constitute a three-dimensional Dornay triangle T2 included in the reference plane Rp, which is a plane defined by a frequently occurring normal vector Vf.

[0032] Subsequently, if necessary, by setting a line parallel to the line orthogonal to the corresponding normal vector V as an outline for the plane reflection point cloud, it becomes possible to more favorably obtain the shape of the plane portion.

[0033] Next, with reference to Figure 13, a point cloud acquisition program for making the computer function as a point cloud acquisition device 1 will be described. The point cloud acquisition program AP includes a main module m1, a projection module m2, a triangle generation module m3, a plane calculation module m4, a normal vector calculation module m5, an offset calculation module m6, and a point cloud acquisition module m7. The triangle generation module m3 includes a first generation module m31 and a second generation module m32. The point cloud acquisition module m7 includes a noise reduction module m71 and a nearby reflection point detection module m72.

[0034] The main module m10 is the part that provides overall control. The functions realized by executing the projection module m2, triangle generation module m3, plane calculation module m4, normal vector calculation module m5, offset calculation module m6, and point cloud acquisition module m7 (projection function, triangle generation function, plane calculation function, normal vector calculation function, offset calculation function, point cloud acquisition function) are the same as those of the projection unit 2, triangle generation unit 3, plane calculation unit 4, normal vector calculation unit 5, offset calculation unit 6, and point cloud acquisition unit 7 of the point cloud acquisition device 1 shown in Figure 1.

[0035] Furthermore, the functions realized by executing the first generation module m31 and the second generation module m32, as well as the noise reduction module m71 and the proximity reflection point detection module m72, are the same as those of the first generation unit 31 and the second generation unit 32, as well as the noise reduction unit 71 and the proximity reflection point detection unit 72 shown in Figure 1.

[0036] The point cloud acquisition program AP described above is provided by a storage medium M1 (e.g., main memory or auxiliary storage device) such as a magnetic disk, optical disk, or semiconductor memory. Alternatively, the point cloud acquisition program AP may be provided via a communication network as a computer data signal superimposed on a carrier wave.

[0037] As described above, in the point cloud acquisition method, point cloud acquisition device 1, and point cloud acquisition program AR according to this embodiment, a three-dimensional Dornay triangle T2 is generated from the laser reflection point cloud G0 of a target region OA which includes a plurality of laser reflection points P having three-dimensional coordinates consisting of X, Y, and Z coordinates, with each of the three laser reflection points P as its vertices. Furthermore, a plane containing each of the three-dimensional Dornay triangles T2 is calculated based on the equation of the plane. Then, the normal vector V of each of these plurality of planes is calculated.

[0038] Generally, on the flat surfaces of artificial structures, such as roofs and walls (in this embodiment, a roof as an example), many planes corresponding to numerous three-dimensional Dornay triangles T2 face the same direction, whereas on the surfaces of natural objects such as trees, the direction in which these planes face tends to be random. In other words, on the flat surfaces of structures, the angles of the normal vectors V of these planes tend to be the same.

[0039] Therefore, here, the most frequent normal vector Vf is determined, which is the angle among multiple normal vectors V whose frequency of appearance peaks at Cp. Then, by detecting the reference point group G1, which includes multiple laser reflection points P that constitute the 3D Dornay triangle T2 contained in the reference plane Rp, which is the plane defined by the most frequent normal vector Vf, a point group G3 (plane reflection point group) corresponding to the planar portion of the structure is obtained. Thus, by using this plane reflection point group, it becomes possible to suitably obtain the shape of the planar portion of the structure. As a result, it becomes possible to estimate the outline of the structure or to automatically generate an outline drawing of the structure.

[0040] Furthermore, the point cloud acquisition method according to this embodiment includes a projection step (step S101) in which a two-dimensional laser reflection point cloud G0a is acquired by projecting the laser reflection point cloud G0 onto an XY plane including the X and Y coordinates. The triangle generation step includes a first generation step (step S102) in which multiple two-dimensional Dornay triangles T1 are generated with each of the three laser reflection points P included in the two-dimensional laser reflection point cloud G0a as vertices, and a second generation step (step S103) in which multiple three-dimensional Dornay triangles T2 are generated by assigning Z coordinates to each vertex of the multiple two-dimensional Dornay triangles T1. In this way, by involving the two-dimensional laser reflection point cloud G0a and the two-dimensional Dornay triangles T1 when generating the three-dimensional Dornay triangles T2, the process becomes easier to understand.

[0041] Furthermore, the point cloud acquisition method according to this embodiment includes an offset calculation step (step S106) that calculates the offset of each of a plurality of planes based on the equation of the plane. Then, in the point cloud acquisition step (step S107), a reference point cloud G1 is detected that includes a plurality of laser reflection points P that constitute a three-dimensional Dornay triangle T2 included in a reference plane Rp, which is a plane defined by a frequently occurring normal vector Vf and an offset (constant d). This makes it possible to separate different planes defined by similar frequently occurring normal vectors Vf (for example, frequently occurring normal vectors Va, Vb having the same angle).

[0042] Furthermore, in the point cloud acquisition method according to this embodiment, the point cloud acquisition step includes a noise reduction step (step S108) in which isolated laser reflection points P and / or laser reflection points P close to the ground surface, which are included in the reference point cloud G1, are removed from the reference point cloud G1 as noise to generate a point cloud G2 after noise reduction. This makes it possible to acquire the shape of the planar portion of a structure more accurately.

[0043] Furthermore, in the point cloud acquisition method according to this embodiment, the point cloud acquisition step includes a proximity reflection point detection step (step S109) in which a laser reflection point P located on the same plane as the reference plane Rp and adjacent to a three-dimensional Dornay triangle T2 included in the reference plane Rp is detected as a proximity reflection point Pn, and the proximity reflection point Pn is included in the reference point cloud G1 (in this case, point cloud G2). As a result, it becomes possible to acquire the shape of the edge portion of the planar part of the structure more accurately.

[0044] The above embodiments illustrate one aspect of the present invention. Therefore, the present invention is not limited to the above embodiments and can be modified as needed.

[0045] For example, in the above embodiment, an example was described in which, as a preliminary step before the second generation unit 32 generates a three-dimensional Dornay triangle T2 in step S103, the projection unit 2 projects the laser reflection point cloud G0 onto the XY plane in step S101, or the first generation unit 31 generates a two-dimensional Dornay triangle T1 in step S102. However, steps S101 and S102 may be omitted, and the triangle generation unit 3 may directly generate a three-dimensional Dornay triangle T2 from the three-dimensional laser reflection point cloud G0.

[0046] Furthermore, in the above embodiment, an example was described in which, after the normal vector calculation unit 5 calculates the frequently occurring normal vector Vf in step S105, the offset calculation unit 6 calculates the plane offset in step S106. However, step S106 may be performed before step S105, or steps S105 and S106 may be performed in parallel.

[0047] Furthermore, in the above embodiment, step S108 described an example in which the noise reduction unit 71 of the point cloud acquisition device 1 removes isolated laser reflection points P and / or laser reflection points P close to the ground surface from the reference point cloud G1 as noise. However, the operator may manually remove isolated laser reflection points P and / or laser reflection points P close to the ground surface as noise from the reference point cloud G1 by visually inspecting the point cloud displayed on the display of the point cloud acquisition device 1, etc., and generate point cloud G2. Similarly, in step S109, the operator may manually detect nearby reflection points Pn by visually inspecting the point cloud displayed on the display of the point cloud acquisition device 1, etc., and generate point cloud G3.

[0048] Furthermore, noise reduction in step S108 (generation of point cloud G2) and detection of nearby reflection point Pn in step S109 (generation of point cloud G3) are not mandatory; the reference point cloud G1 may be acquired in step S107 and obtained as a planar reflection point cloud.

[0049] Furthermore, examples of planar parts of a structure are not limited to roofs, but include any planar structural parts such as the side walls of a structure. [Explanation of Symbols]

[0050] 1...Point cloud acquisition device, 2...Projection unit, 3...Triangle generation unit, 4...Plane calculation unit, 5...Normal vector calculation unit, 6...Offset calculation unit, 7...Point cloud acquisition unit, 31...First generation unit, 32...Second generation unit, 71...Noise reduction unit, 72...Nearby reflection point detection unit, G0...Laser reflection point cloud, G1...Reference point cloud, G3...Point cloud (Plane reflection point cloud), T1...2D Dornay triangle, T2...3D Dornay triangle, P...Laser reflection point, Pn...Nearby reflection point, Rp...Reference plane, V...Normal vector, Va, Vb, Vf...Frequently occurring normal vectors.

Claims

1. A point cloud acquisition method for obtaining a planar reflection point cloud corresponding to the planar portion of a structure from a laser reflection point cloud of a target region including a plurality of laser reflection points having three-dimensional coordinates consisting of X, Y, and Z coordinates, A triangle generation step that generates multiple three-dimensional Dornay triangles with each of the three aforementioned laser reflection points as its vertex, Following the triangle generation step, a plane calculation step is performed to calculate a plane that includes each of the multiple three-dimensional Dornay triangles based on the equation of the plane. Following the plane calculation step, a normal vector calculation step is performed to calculate the normal vectors of each of the multiple planes and to find the frequently occurring normal vector where the angle appears most frequently among the multiple normal vectors. A point cloud acquisition step is performed after the vector calculation step, by detecting a group of reference points including a plurality of laser reflection points that constitute the three-dimensional Dornay triangle, which is included in the reference plane that is defined by the frequently occurring normal vector, thereby acquiring the plane reflection point group. A point cloud acquisition method comprising the following features.

2. Prior to the triangle generation step, a projection step is provided to obtain a two-dimensional laser reflection point cloud by projecting the laser reflection point cloud onto an XY plane including the X and Y coordinates. The aforementioned triangle generation step is: A first generation step of generating a plurality of two-dimensional Dornay triangles whose vertices are each of the three laser reflection points included in the two-dimensional laser reflection point group, A second generation step is performed after the first generation step, in which a plurality of three-dimensional Dornay triangles are generated by assigning the Z coordinates to each vertex of the plurality of two-dimensional Dornay triangles, including, The point cloud acquisition method according to claim 1.

3. The offset calculation step, performed after the plane calculation step and before the point cloud acquisition step, calculates the offset of each of the multiple planes based on the equation of the plane. In the point cloud acquisition step, the reference point cloud is detected, which includes a plurality of laser reflection points that constitute the three-dimensional Dornay triangle included in the reference plane, which is the plane defined by the frequently occurring normal vector and the offset. The point cloud acquisition method according to claim 1.

4. The point cloud acquisition step includes a noise reduction step of removing isolated laser reflection points and / or laser reflection points close to the ground surface from the reference point cloud as noise. The point cloud acquisition method according to claim 1.

5. The point cloud acquisition step includes a proximity reflection point detection step which detects the laser reflection points located on the same plane as the reference plane and adjacent to the three-dimensional Dornay triangle included in the reference plane as proximity reflection points, and includes the proximity reflection points in the reference point cloud. A method for acquiring a point cloud according to any one of claims 1 to 4.

6. A point cloud acquisition device for obtaining a planar reflection point cloud corresponding to the planar portion of a structure from a laser reflection point cloud of a target region including a plurality of laser reflection points having three-dimensional coordinates consisting of X, Y, and Z coordinates, A triangle generation unit that generates multiple three-dimensional Dornay triangles with each of the three aforementioned laser reflection points as its vertex, A plane calculation unit calculates a plane that includes each of the multiple three-dimensional Dornay triangles based on the equation of the plane, A normal vector calculation unit calculates the normal vectors of each of the multiple planes and finds the frequently occurring normal vector among the multiple normal vectors where the angle appears most frequently, A point cloud acquisition unit acquires the plane reflection point cloud by detecting a group of reference points that include a plurality of laser reflection points constituting the three-dimensional Dornay triangle, which is included in the reference plane that is defined by the frequently occurring normal vector, A point cloud acquisition device equipped with the following features.

7. A point cloud acquisition program for causing a computer to function as a point cloud acquisition device for acquiring a planar reflection point cloud corresponding to the planar portion of a structure from a laser reflection point cloud of a target region including multiple laser reflection points having three-dimensional coordinates consisting of X, Y, and Z coordinates, To the aforementioned computer, A triangle generation function that generates multiple three-dimensional Dornay triangles with each of the three aforementioned laser reflection points as its vertex, A plane calculation function that calculates a plane containing each of the multiple three-dimensional Dornay triangles based on the equation of the plane, A normal vector calculation function that calculates the normal vectors of each of the multiple planes and finds the most frequent normal vector among the multiple normal vectors where the angle appears most frequently, A point cloud acquisition function that acquires the plane reflection point cloud by detecting a group of reference points including a plurality of laser reflection points that constitute the three-dimensional Dornay triangle, which is included in the reference plane that is defined by the frequently occurring normal vector, To make it happen Point cloud acquisition program.

Images