JP7743779B2 - Bridge data extraction device, bridge data extraction method, and program - Google Patents

Description

The technical field relates to a bridge data extraction device, a bridge data extraction method, and a program that extracts point cloud data representing bridges.

Non-Patent Document 1 discloses a technology for generating a 3D model of a bridge using point cloud data generated using a ground-based laser scanner and point cloud data generated from images captured by an autonomous unmanned aerial vehicle. The technology in Non-Patent Document 1 also discloses a function for removing noise from the point cloud data.

Yoshinori Tsukada et al., "Study on generating 3D bridge models using point cloud data," Intelligence and Information, vol. 27, No. 5, pp. 796-812, 2015.

However, Non-Patent Document 1 only discloses a function for removing noise, such as outliers caused by moving objects such as people and measurement errors. In other words, it does not disclose a function for accurately extracting bridge point cloud data from point cloud data obtained by measuring a target area (space) that contains bridges and other natural objects and structures around the bridges.

One aspect of the present invention is to provide a bridge data extraction device, a bridge data extraction method, and a program that accurately extract point cloud data representing bridges from point cloud data representing a target area.

In order to achieve the above object, a bridge data extraction device in one aspect includes:
a determination unit that, when the direction of a point determined using a point included in target area point cloud data representing a target area including a bridge and a plurality of points included in a preset range around the point can be considered to be horizontal, determines that the point is lower layer point cloud data representing a pier or abutment of the bridge;
a selection unit that executes a division process based on distances between points included in the target area point cloud data, and selects bridge point cloud data representing the bridge based on the division result;
The present invention is characterized by having the following.

In order to achieve the above object, a bridge data extraction method in one aspect includes:
The computer
If the direction of a point determined using a point included in target area point cloud data representing a target area including a bridge and a plurality of points included in a predetermined range around the point can be considered to be horizontal, the point is determined to be lower layer point cloud data representing a pier or abutment of the bridge,
A division process is performed based on the distances between points included in the target area point cloud data, and bridge point cloud data representing the bridge is selected based on the division result.

Furthermore, in order to achieve the above object, the program in one aspect comprises:
On the computer,
If the direction of a point determined using a point included in target area point cloud data representing a target area including a bridge and a plurality of points included in a predetermined range around the point can be considered to be horizontal, the point is determined to be lower layer point cloud data representing a pier or abutment of the bridge,
A division process is performed based on the distances between points included in the target area point cloud data, and bridge point cloud data representing the bridge is selected based on the division result.

One aspect is that it allows for accurate extraction of point cloud data representing bridges from point cloud data representing the target area.

FIG. 1 is a diagram for explaining an example of a bridge data extraction device according to the first embodiment. FIG. 2 is a diagram illustrating an example of a system including the bridge data extraction device according to the first embodiment. FIG. 3 is a diagram illustrating an example of a target area from which point cloud data is acquired. FIG. 4 is a diagram for explaining horizontal determination. FIG. 5 is a diagram for explaining an example of the operation of the bridge data extraction device in the first embodiment. FIG. 6 is a diagram illustrating an example of the operation of the determination process. FIG. 7 is a diagram illustrating an example of the operation of the selection process. FIG. 8 is a diagram illustrating an example of a bridge data extraction device according to the second embodiment. FIG. 9 is a diagram for explaining an example of the operation of the bridge data extraction device in the second embodiment. FIG. 10 is a diagram illustrating an example of a bridge data extraction device according to the third embodiment. FIG. 11 is a diagram for explaining an example of the operation of the bridge data extraction device according to the third embodiment. FIG. 12 is a diagram illustrating an example of a bridge data extraction device according to the fourth embodiment. FIG. 13 is a diagram for explaining an example of the operation of the bridge data extraction device according to the fourth embodiment. FIG. 14 is a diagram illustrating an example of a computer that realizes the bridge data extraction device according to the first to fourth embodiments.

Embodiments will be described below with reference to the drawings. Note that in the drawings described below, elements having the same or corresponding functions will be given the same reference numerals, and repeated explanations may be omitted.

(Embodiment 1)
The configuration of a bridge data extraction device 10 according to the first embodiment will be described with reference to Fig. 1. Fig. 1 is a diagram for explaining an example of a bridge data extraction device according to the first embodiment.

[Device configuration]
The bridge data extraction device 10 shown in Fig. 1 is a device that accurately extracts point cloud data representing a bridge from point cloud data representing a target area. As shown in Fig. 1, the bridge data extraction device 10 also has a determination unit 11 and a selection unit 12.

If the direction of a point determined using a point included in the target area point cloud data representing a target area including a bridge and multiple points included in a predetermined range around the point can be considered to be horizontal (if the horizontal direction is determined to be dominant), the determination unit 11 determines that the point is lower layer point cloud data representing a bridge pier or abutment.

Both piers and abutments have sides that can be considered perpendicular to the horizontal plane. Therefore, the direction determined by a point included in the lower point cloud data representing the lower part of the bridge (pier or abutment) and multiple points near that point, such as the direction of the normal, can be considered to be horizontal. Therefore, by determining whether the direction of the normal is a direction that can be considered horizontal, it can be determined that a point included in the target area point cloud data is included in the lower point cloud data.

The selection unit 12 performs a division process (a process of dividing the data into groups with similar characteristics) based on the distances between points included in the target area point cloud data, and selects bridge point cloud data representing bridges based on the division results. This division process can also be described as a process of grouping multiple points in the target area point cloud data that can be determined to be located close to each other into a single group. In this case, "close" refers to the result of a relative comparison of the distances between points included in the target area point cloud data.

Specifically, the selection unit 12 performs a clustering process based on the distances between points included in the target area point cloud data to generate clusters, selects a cluster that has points included in the lower layer point cloud data, and determines that the points included in the selected cluster are included in bridge point cloud data representing a bridge.

Bridge point cloud data corresponding to the lower part of the bridge is obtained using clusters that have been clustered based on the distances between points included in the target area point cloud data and the lower part point cloud data, so point cloud data of natural objects, structures, etc. other than the bridge can be removed. As a result, bridge point cloud data corresponding to the lower part of the bridge can be extracted with high accuracy.

[System configuration]
The configuration of the bridge data extraction device 10 according to the first embodiment will be described in detail with reference to Fig. 2. Fig. 2 is a diagram illustrating an example of a system 200 including the bridge data extraction device according to the first embodiment.

The system 200 includes a bridge data extraction device 10, a sensor 20, and a storage device 30. The bridge data extraction device 10 includes an acquisition unit 13, a direction calculation unit 14, a determination unit 11, and a selection unit 12.

The bridge data extraction device 10 is, for example, a CPU (Central Processing Unit), a programmable device such as an FPGA (Field-Programmable Gate Array), a GPU (Graphics Processing Unit), or an information processing device such as a circuit equipped with one or more of these, a server computer, a personal computer, or a mobile terminal.

Sensor 20 measures the target area and generates target area point cloud data using the measurement results. Sensor 20 measures an area with a preset angle of view.

Specifically, the sensor 20 measures the measurable range and generates point cloud data using the measurement results. The sensor 20 is, for example, a Light Detection and Ranging (LiDAR) sensor.

The LiDAR sensor may also be equipped with an inclinometer, which measures the vertical direction and the tilt of an object relative to the vertical direction.

The LiDAR sensor is also assumed to be installed horizontally using an inclinometer in advance. For example, if the LiDAR sensor is supported by a tripod, the length of the tripod legs is adjusted to install the LiDAR sensor horizontally. However, the sensor 20 is not limited to a LiDAR sensor.

In addition, point cloud data may be created using an imaging device and image processing device instead of a LiDAR sensor. In this case, first, multiple images of the target area are captured by the imaging device. Next, the multiple captured images are processed using Structure from Motion (SfM) software installed on the image processing device to generate point cloud data of the target area. However, the generation of point cloud data is not limited to the process described above.

In the example of Figure 2, the sensor 20 is provided outside the bridge data extraction device 10, but it may also be provided inside the bridge data extraction device 10.

We will now explain the target area point cloud data. Figure 3 is a diagram illustrating an example of a target area from which point cloud data is acquired. In the example of Figure 3, in the target area 1 (target space), a bridge 3, natural objects 4, structures 5, and sensors 20 exist on the ground surface 2. However, the target area is not limited to the target area 1 shown in Figure 3.

The ground surface 2 is not actually flat, but for ease of understanding, it will be considered flat for the sake of convenience. Furthermore, the ground surface 2 may actually be inclined.

The bridge 3 has an upper layer 6 and a lower layer 7 (7a, 7b). The upper layer 6 is composed of components such as girders and deck plates. The lower layer 7 (7a, 7b) is a pier or abutment. Note that in Figure 3, the upper layer 6 and lower layer 7 are used for convenience to make the structure of the bridge 3 easier to understand.

In the example of Figure 3, the natural object 4 is a plant. In the example of Figure 3, the structure 5 is a medium- to high-rise building. However, the type, size, number, and location of the natural object 4 and structure 5 are not limited to the example of Figure 3.

Point cloud data is data that represents a collection of multiple points in a three-dimensional coordinate system. If the sensor 20 shown in Figure 3 is a LiDAR sensor, the point cloud data is represented, for example, in a three-dimensional coordinate system with the position of the sensor 20 as the origin.

In the example of Figure 3, the three-dimensional coordinate system represents the horizontal direction with the x-axis and y-axis, and the vertical direction with the z-axis. Specifically, the x-axis represents the horizontal direction in Figure 3, and the y-axis represents the depth direction in Figure 3. In addition, in the example of Figure 3, the z-axis is perpendicular (vertical) to the horizontal plane. Furthermore, when the three-dimensional coordinate system is represented by the x-axis, y-axis, and z-axis, the point cloud data represents the position of a point using x-coordinate values, y-coordinate values, and z-coordinate values.

However, point cloud data may also be expressed using, for example, voxel data, depth images, etc. Furthermore, point cloud data may also be expressed using something other than the above-mentioned three-dimensional coordinate system, voxel data, and depth images.

In the following explanation, the three-dimensional coordinate system will have its origin at the location where the sensor 20 is installed, with the x-axis and y-axis representing the horizontal direction and the z-axis representing the vertical direction.

Furthermore, the LiDAR sensor does not necessarily have to be installed horizontally. This is because even if the LiDAR sensor is not installed horizontally, it can be converted into a coordinate system in which the x-axis and y-axis represent the horizontal direction and the z-axis represents the vertical direction by performing coordinate conversion processing. A commonly known process is used for the coordinate conversion processing.

The storage device 30 stores bridge point cloud data, which is the result of extracting point cloud data representing bridges from the target area point cloud data. In addition to bridge point cloud data, the storage device 30 may also store, for example, data generated during the process of acquiring bridge point cloud data, target area point cloud data, etc.

The storage device 30 may be, for example, a database, a server computer, a personal computer, or the memory of a mobile device.

In the example of Figure 2, the storage device 30 is provided externally to the bridge data extraction device 10, but it may also be provided internally to the bridge data extraction device 10.

The bridge data extraction device will now be described in detail.
The acquisition unit 13 acquires point cloud data (target area point cloud data P) obtained by measuring part or all of the lower layer 7 , and outputs the acquired target area point cloud data P to the direction calculation unit 14 .

The acquisition unit 13 may acquire the target area point cloud data P in real time, or may acquire the target area point cloud data P stored in a storage device such as the storage device 30.

Furthermore, the target area point cloud data P may be an integration of multiple point cloud data measured by changing the measurement position of the sensor 20.

The target area point cloud data P can be expressed as in Equation 1. In Equation 1, each point included in the target area point cloud data P is represented as pi. The position (3D coordinates) of point pi is represented as (xi, yi, zi).

(Equation 1)
P={pi,i=1,2,3,…,N}
pi=(xi,yi,zi)

The direction calculation unit 14 uses the target area point cloud data P to calculate the direction to set for each point pi included in the target area point cloud data P. Specifically, the direction calculation unit 14 first acquires the target area point cloud data P.

Next, the direction calculation unit 14 calculates the direction for each point pi included in the target area point cloud data P. The direction calculation unit 14 calculates the normal using, for example, point pi included in the target area point cloud data P and multiple points in the vicinity of point pi (multiple points surrounding the target point). The vicinity is a pre-set range around the target point. The vicinity may also be a range in which the distance from a certain point satisfies a criterion for determining that the point is in the vicinity. The criterion is, for example, that the distance is shorter than a predetermined threshold.

Note that if the points in the target area point cloud data P are interpolated using a spline curve or the like, the normal may be the normal direction of the spline curve.

The direction set for a point may not only be the normal direction, but also a direction tilted at a specified angle from the normal direction toward the y-axis. The method of expressing the direction at a point is not limited to using the normal line.

Next, the direction calculation unit 14 adds information (directional information) representing the direction calculated for each point pi to each point pi in the target area point cloud data P. Note that hereafter, the information obtained by adding directional information to the target area point cloud data P shown in Equation 1 will be referred to as target area point cloud data P', as shown in Equation 2.

(Equation 2)
P′={p′i,i=1,2,3,…,N}
p´i=(xi, yi, zi, vector(Ni))

The information p'i in Equation 2 is included in the target area point cloud data P'. Furthermore, vector(Ni) is information to which directional information calculated for each point pi has been added. Vector(Ni) represents, for example, a normal unit vector.

The following processing will be explained using normal unit vectors, but the normal direction does not necessarily have to be a unit vector. If the vector is not a unit vector, the processing described below can be achieved by comparing the magnitude of the vector's horizontal component with the magnitude of its vertical component.

The determination unit will now be described.
The determination unit 11 includes a horizontal determination unit 15 and a size determination unit 16 .

The horizontal determination unit will now be described.
The horizontal determination unit 15 selects point cloud data corresponding to the lower layer (lower layer point cloud data C) based on the direction information of each piece of information p'i of the target area point cloud data P' to which direction information has been added.

Specifically, first, the horizontal determination unit 15 acquires the target area point cloud data P'. Next, the horizontal determination unit 15 selects the lower layer point cloud data C representing the lower layer 7 based on the directional information of the target area point cloud data P'.

For example, if the direction information is a normal unit vector, the horizontal determination unit 15 calculates the magnitude of the horizontal component of the normal unit vector (the horizontal component represented by the x-axis and y-axis in Figure 3), and selects a point representing the lower layer 7 from the points included in the target area point cloud data P' based on the magnitude of the calculated horizontal component.

In detail, if the magnitude of the horizontal component of the normal unit vector of a target point included in the target area point cloud data P' is greater than a preset horizontal threshold, the horizontal determination unit 15 determines the target point as a candidate for the lower layer 7.

The side of a pier or abutment is likely to have a vertical side, like the lower layers 7a and 7b shown in Figure 3. In such cases, in the lower layers 7a and 7b, the horizontal component of the normal unit vector of the target point will be more dominant than the vertical component (the horizontal component will be larger), so it is possible to determine whether the target point is a candidate point for the side of the pier or abutment.

The horizontal threshold is a value used to determine whether a point corresponds to the side of a pier or abutment. The horizontal threshold can be determined through experiments and simulations. The horizontal threshold can be set to 0.5, for example.

Furthermore, the lower layer point cloud data C can be expressed as Equation 3. The Horizontal function shown in Equation 3 is a function that calculates the horizontal component from the normal unit vector.

(Equation 3)
C = {p'i | Horizontal(vector(Ni)) > i that satisfies Horizontal Threshold}

Incidentally, since the lower layer point cloud data C is a subset of the target area point cloud data P, new flag information may be added as a field in the target area point cloud data P to distinguish points that correspond to the lower layer point cloud data C.

The size determination unit will now be described.
The size determination unit 16 performs a clustering process based on the distances between points included in the lower layer point cloud data C to generate clusters, and reselects the lower layer point cloud data C based on the points belonging to the generated clusters to generate lower layer point cloud data C'.

Specifically, first, the size determination unit 16 performs, for example, Euclidean clustering processing on the point clouds included in the lower layer point cloud data C to generate clusters.

Next, for each generated cluster, the size determination unit 16 compares the number of points belonging to the cluster with a preset size threshold. Next, the size determination unit 16 selects clusters in which the number of points belonging to the cluster is greater than the size threshold.

The size threshold is determined based on the size of the side of the pier or abutment. The size threshold is determined, for example, through experiments, simulations, etc.

The process of selecting clusters by comparing the number of points belonging to a cluster with a size threshold, as described above, is a process of removing points located far from the lower layer 7 from the lower layer point cloud data C. In other words, it is a process of removing points (noise) such as natural objects 4 and structures 5 other than the lower layer 7 from the lower layer point cloud data C. In other words, it is a process of extracting points corresponding to the lower layer 7.

Therefore, by comparing the number of points belonging to a cluster with a size threshold and selecting the cluster, it is possible to accurately extract points corresponding to the lower layer 7 from the lower layer point cloud data C.

Figure 4 is a diagram for explaining horizontal determination. In the example of Figure 4, the lower layer portions 7a and 7b are each represented as a rectangular prism. Also, in the example of Figure 4, it is assumed that the side surfaces of the lower layer portions 7a and 7b are parallel to the z-axis.

Arrow nv1 in Figure 4 represents the normal unit vector of a point on the side of lower layer 7a. Arrow nv2 in Figure 4 represents the normal unit vector of a point on the side of lower layer 7b. Arrow nv3 in Figure 4 represents the normal unit vector of a point on the floorboard of upper layer 6. Arrow nv4 in Figure 4 represents the normal unit vector of a point on natural object 4. Arrow nv5 in Figure 4 represents the normal unit vector of a point on structure 5. Arrow nv6 in Figure 4 represents the normal unit vector of a point on the ground surface 2.

Furthermore, when the horizontal component of the normal unit vector for arrows nv1 to nv6 is calculated, in the example of Figure 4, arrows nv1, 2, and 5 are pointing horizontally, so the magnitude of the horizontal component becomes the size threshold.

In addition to the process of selecting clusters by comparing the number of points belonging to a cluster with a size threshold, clusters may also be selected using the following process.

For example, the size determination unit 16 selects only clusters consisting of points within a predetermined area range.

The process of selecting only clusters consisting of points within a preset area range, as described above, is a process of determining that points within a certain range of sizes in the lower layer point cloud data C are the lower layer. For example, if the size of the bridge pier is known in advance, this process is a process of removing points (noise) other than the lower layer 7, such as natural objects 4 and structures 5, based on the number of points.

Instead of the number of points in a cluster, a mesh can be constructed using the points belonging to the cluster, and the area of the constructed mesh can be used.

Next, the size determination unit 16 generates lower layer point cloud data C' using the points included in the selected cluster. However, the processing of the size determination unit 16 described above does not have to be performed.

The selection unit will now be described.
The selection unit 12 includes a height restriction unit 17 , a clustering unit 18 , and a cluster selection unit 19 .

The height limiting unit 17 performs preprocessing on the target area point cloud data P or the target area point cloud data P' based on the height information described below, and extracts point cloud data Q to be used by the clustering unit 18.

Specifically, first, the height limiting unit 17 acquires the target area point cloud data P and the lower layer point cloud data C (if the size determining unit 16 is not present), or the target area point cloud data P and the lower layer point cloud data C' (if the size determining unit 16 is present). Next, the height limiting unit 17 generates a height threshold (height information) based on the lower layer point cloud data C or the lower layer point cloud data C'.

The height threshold is the largest z coordinate value among the points included in the lower level point cloud data C or the lower level point cloud data C'. The height threshold may also be set in advance based on the height of the pier or abutment.

Next, the height limiting unit 17 extracts points whose z coordinate values are equal to or greater than the height threshold from among the points included in the target area point cloud data P or the target area point cloud data P'. The height limiting unit 17 then uses the extracted points as point cloud data Q to be used by the clustering unit 18.

Note that the processing of the height limiting unit 17 does not have to be performed. In that case, the clustering unit 18 uses the lower layer point cloud data C or the lower layer point cloud data C' as the point cloud data Q.

The clustering unit 18 performs clustering processing using the point cloud data Q. Specifically, the clustering unit 18 first acquires the point cloud data Q.

Next, the clustering unit 18 uses the points included in the point cloud data Q to perform Euclidean clustering processing based on the distances between points to generate clusters.

The cluster Bj obtained by the clustering process performed on the point cloud data Q is expressed as in Equation 4.

(Equation 4)
Bj (j=1, 2, 3,...,M)

The cluster selection unit 19 selects a cluster in which at least one point belonging to cluster Bj is included in the lower layer point cloud data C or the lower layer point cloud data C'. The cluster selection unit 19 then generates point cloud data B (bridge point cloud data) using the points belonging to the selected cluster.

[Device operation]
Next, the operation of the bridge data extraction device in embodiment 1 will be described with reference to FIG. 5. FIG. 5 is a diagram for explaining the operation of the bridge data extraction device in embodiment 1. In the following description, reference will be made to the diagram as appropriate. In embodiment 1, the bridge data extraction method is implemented by operating the bridge data extraction device. Therefore, the description of the bridge data extraction method in embodiment 1 will be replaced by the following description of the operation of the bridge data extraction device.

As shown in FIG. 5, first, the acquisition unit 13 acquires point cloud data (target area point cloud data P) obtained by measuring part or all of the lower layer 7, and outputs the acquired target area point cloud data P to the direction calculation unit 14 (step A1).

Next, the direction calculation unit 14 uses the target area point cloud data P to calculate the direction to be set for each point pi included in the target area point cloud data P (step A2).

Specifically, in step A2, first, the direction calculation unit 14 acquires the target area point cloud data P. Next, in step A2, the direction calculation unit 14 calculates the direction for each point pi included in the target area point cloud data P.

Next, if the direction set for each point pi can be considered to be the horizontal direction (if the horizontal direction is determined to be dominant), the determination unit 11 determines that the point pi is included in the lower-layer point cloud data C (Step A3: Determination process).

Next, the selection unit 12 performs a division process (a process of dividing the data into groups with similar characteristics) based on the distances between points included in the target area point cloud data P, and selects bridge point cloud data representing bridges based on the division results (Step A4: Selection process).

The determination process (step A3) will now be described in detail.
6 is a diagram illustrating an example of the operation of the determination process. As shown in Fig. 6, first, the horizontal determination unit 15 selects lower layer point cloud data C corresponding to the lower layer 7 based on the directional information of each piece of information p'i of the target area point cloud data P' to which directional information has been added (step B1).

Specifically, in step B1, the horizontal determination unit 15 first acquires the target area point cloud data P'. Next, in step B1, the horizontal determination unit 15 selects the lower layer point cloud data C representing the lower layer 7 based on the directional information of the target area point cloud data P'.

For example, if the direction information is a normal unit vector, the horizontal determination unit 15 calculates the magnitude of the horizontal component of the normal unit vector (the horizontal component represented by the x-axis and y-axis in Figure 3), and selects a point representing the lower layer 7 from the points included in the target area point cloud data P' based on the magnitude of the calculated horizontal component.

In detail, if the magnitude of the horizontal component of the normal unit vector of a target point included in the target area point cloud data P' is greater than a preset horizontal threshold, the horizontal determination unit 15 determines the target point as a candidate for the lower layer 7.

The side of a pier or abutment is likely to have a vertical side, as shown in the lower layers 7a and 7b in Figure 3. In such cases, in the lower layers 7a and 7b, the normal unit vector of the target point has a dominant horizontal component over the vertical component (the horizontal component is larger), so the target point can be selected as a candidate for the side of the pier or abutment.

Next, the size determination unit 16 performs a clustering process based on the distances between points included in the lower layer point cloud data C to generate clusters, and reselects the lower layer point cloud data C based on the points belonging to the clusters to generate lower layer point cloud data C' (step B2).

Specifically, in step B2, the size determination unit 16 first generates clusters by performing, for example, Euclidean clustering processing on the point clouds included in the lower layer point cloud data C.

Next, in step B2, the size determination unit 16 compares the number of points belonging to each generated cluster with a preset size threshold. Next, in step B2, the size determination unit 16 selects clusters in which the number of points belonging to the cluster is greater than the size threshold.

Next, in step B2, the size determination unit 16 sets the points included in the selected cluster as lower-layer point cloud data C'. However, the processing of step B2 described above does not necessarily have to be performed.

The process of selecting clusters by comparing the number of points belonging to a cluster with a size threshold, as described above, is a process of removing points located far from the lower layer 7 from the lower layer point cloud data C. In other words, it is a process of removing points (noise) such as natural objects 4 and structures 5 other than the lower layer 7 from the lower layer point cloud data C. In other words, it is a process of extracting points corresponding to the lower layer 7.

Therefore, by comparing the number of points belonging to a cluster with a size threshold and selecting the cluster, it is possible to accurately extract points corresponding to the lower layer 7 from the lower layer point cloud data C.

In addition to the process of selecting clusters by comparing the number of points belonging to a cluster with a size threshold, clusters may also be selected using the following process.

For example, the size determination unit 16 selects only clusters consisting of points within a predetermined area range.

The process of selecting only clusters consisting of points within a preset area range, as described above, is a process of determining that points within a certain range of sizes in the lower layer point cloud data C are the lower layer. For example, if the size of the bridge pier is known in advance, this process is a process of removing points (noise) other than the lower layer 7, such as natural objects 4 and structures 5, based on the number of points.

Instead of the number of points in a cluster, a mesh can be constructed using the points belonging to the cluster, and the area of the constructed mesh can be used.

The selection process (step A4) will now be described in detail.
7 is a diagram illustrating an example of the operation of the selection process. As shown in Fig. 7, first, the height limiting unit 17 performs preprocessing on the target area point cloud data P or the target area point cloud data P' based on height information described later, and extracts point cloud data Q to be used by the clustering unit 18 (step C1).

Specifically, in step C1, the height limiting unit 17 first acquires the target area point cloud data P and the lower layer point cloud data C (if the size determining unit 16 is not present), or the target area point cloud data P and the lower layer point cloud data C' (if the size determining unit 16 is present).

Next, in step C1, the height limiting unit 17 generates a height threshold (height information) based on the lower level point cloud data C or the lower level point cloud data C'. The height threshold is the largest z coordinate value among the points included in the lower level point cloud data C or the lower level point cloud data C'. The height threshold may also be set in advance based on the height of the pier or abutment.

Next, in step C1, the height limiting unit 17 extracts points whose z coordinate values are equal to or greater than the height threshold from among the points included in the target area point cloud data P or the target area point cloud data P'.

Then, in step C1, the height limiting unit 17 uses the extracted points to extract point cloud data Q to be used by the clustering unit 18.

Note that the processing of the height limiting unit 17 does not have to be performed. In that case, the clustering unit 18 uses the lower layer point cloud data C or the lower layer point cloud data C' as the point cloud data Q.

Next, the clustering unit 18 performs clustering processing using the point cloud data Q (step C2).

Specifically, in step C2, the clustering unit 18 first acquires point cloud data Q.

Next, in step C2, the clustering unit 18 performs Euclidean clustering processing based on the distances between points using the points included in the point cloud data Q to generate cluster Bj.

Next, the cluster selection unit 19 selects a cluster in which at least one point belonging to cluster Bj is included in the lower level point cloud data C or the lower level point cloud data C', and sets the points belonging to the selected cluster as point cloud data B (bridge point cloud data) (step C3).

[Effects of the First Embodiment]
According to embodiment 1, if the direction set for a point included in the target area point cloud data representing a target area including a bridge can be regarded as the horizontal direction (if the horizontal direction is determined to be dominant), it can be determined that the point is included in the lower level point cloud data representing the bridge pier or abutment.

In other words, both piers and abutments have sides that can be considered perpendicular to the horizontal plane. Therefore, the direction determined by a point included in the lower level point cloud data representing the pier or abutment and multiple points in its vicinity, such as the direction of the normal, is a direction that can be considered horizontal. Therefore, by using the direction of the normal, it is possible to determine whether a point is included in the lower level point cloud data.

In addition, a division process (a process of dividing the data into groups with similar characteristics) is performed based on the distance between points contained in the target area point cloud data, and bridge point cloud data representing bridges can be selected based on the division results.

Specifically, clusters are generated by performing clustering processing based on the distances between points included in the target area point cloud data, a cluster containing points included in the lower layer point cloud data is selected, and the points included in the selected cluster are used as bridge point cloud data representing the bridge.

Therefore, bridge point cloud data corresponding to the lower part of the bridge is obtained by using clusters that have been clustered based on the distances between points included in the target area point cloud data and the lower part point cloud data, so point cloud data that becomes noise, such as natural objects and structures other than the bridge, can be removed. As a result, bridge point cloud data corresponding to the lower part of the bridge can be extracted with high accuracy.

(Embodiment 2)
The configuration of a bridge data extraction device 80 according to the second embodiment will be described with reference to Fig. 8. Fig. 8 is a diagram for explaining an example of a bridge data extraction device according to the second embodiment.

[System configuration]
The bridge data extraction device 80 shown in Fig. 8 is a device that accurately extracts point cloud data representing a bridge from point cloud data representing a target area. As shown in Fig. 8, the bridge data extraction device 80 includes an acquisition unit 13, a direction calculation unit 14, a determination unit 11, a removal unit 81, and a selection unit 82.

The acquisition unit 13, direction calculation unit 14, and determination unit 11 have already been described in embodiment 1, so descriptions of the acquisition unit 13, direction calculation unit 14, and determination unit 11 will be omitted.

The removal unit 81 removes the lower layer point cloud data C or the lower layer point cloud data C' from the target area point cloud data P or the target area point cloud data P'.

Specifically, the removal unit 81 first acquires the target area point cloud data P or the target area point cloud data P' and the lower layer point cloud data C or the lower layer point cloud data C'.

Next, the removal unit 81 removes the lower layer point cloud data C or the lower layer point cloud data C' from the target area point cloud data P or the target area point cloud data P' to obtain point cloud data D (point cloud data that does not contain a point cloud corresponding to the lower layer (bridge or abutment)). Next, the removal unit 81 outputs the point cloud data D to the selection unit 82.

The selection unit 82 performs clustering processing based on the distances between points included in the point cloud data D to generate clusters, and generates point cloud data B (bridge point cloud data) using the points belonging to the generated clusters and the points included in the lower level point cloud data C or the lower level point cloud data C'.

Specifically, the selection unit 82 first acquires point cloud data D and lower layer point cloud data C or lower layer point cloud data C'. Next, the selection unit 82 performs Euclidean clustering processing based on the distances between points included in point cloud data D to generate clusters.

Next, the selection unit 82 calculates the distance between each point belonging to the generated cluster and the points included in the lower layer point cloud data C or the lower layer point cloud data C'.

Next, the selection unit 82 compares the calculated distance with a preset distance threshold, and if the calculated distance is less than or equal to the distance threshold (if the distance is short), it selects a cluster that includes points whose distance is less than or equal to the distance threshold. The selection unit 82 then sets the points that belong to the selected cluster as point cloud data B (bridge point cloud data).

[Device operation]
Next, the operation of the bridge data extraction device in the second embodiment will be described with reference to FIG. 9. FIG. 9 is a diagram for explaining an example of the operation of the bridge data extraction device in the second embodiment. In the following description, the diagram will be referenced as appropriate. Furthermore, in the second embodiment, a bridge data extraction method is implemented by operating the bridge data extraction device. Therefore, the description of the bridge data extraction method in the second embodiment will be replaced by the following description of the operation of the bridge data extraction device.

In embodiment 2, the processes of steps A1 to A3 and steps D1 to D2 shown in Figure 9 are executed. Since the processes of steps A1 to A3 have already been explained, a description of the processes of steps A1 to A3 will be omitted.

As shown in FIG. 9, first, the removal unit 81 removes the lower layer point cloud data C or the lower layer point cloud data C' from the target area point cloud data P or the target area point cloud data P' (step D1).

Specifically, in step D1, the removal unit 81 first acquires the target area point cloud data P or the target area point cloud data P' and the lower layer point cloud data C or the lower layer point cloud data C'.

Next, in step D1, the removal unit 81 removes the lower layer point cloud data C or the lower layer point cloud data C' from the target area point cloud data P or the target area point cloud data P' to generate point cloud data D (point cloud data that does not contain a point cloud corresponding to the lower layer (bridge or abutment)).

Next, the selection unit 82 performs a clustering process based on the inter-point distances of the points included in the generated point cloud data D to generate clusters (step D2).

Specifically, in step D2, the selection unit 82 first acquires point cloud data D and lower layer point cloud data C or lower layer point cloud data C'. Next, in step D2, the selection unit 82 performs a clustering process based on the distances between points included in point cloud data D to generate clusters.

Next, in step D2, the selection unit 82 calculates the distance between each point belonging to the generated cluster and the points included in the lower layer point cloud data C or the lower layer point cloud data C'.

Next, in step D2, the selection unit 82 compares the calculated distance with a preset distance threshold, and if the calculated distance is less than or equal to the distance threshold (if the distance is short), selects a cluster that includes points whose distance is less than or equal to the distance threshold. The selection unit 82 then sets the points belonging to the selected cluster as point cloud data B (bridge point cloud data).

[Effects of the Second Embodiment]
According to the second embodiment, by performing a clustering process based on the distances between points included in the point cloud data D obtained by removing the lower layer point cloud data from the target area point cloud data, even if the points of the upper layer 6 (girders and deck panels), the points of the ground surface 2, and the points of the lower layer 7 (7a, 7b) belong to the same cluster, the points of the upper layer 6 and the points of the ground surface 2 can be separated into different clusters. Therefore, even if point cloud data corresponding to the ground surface 2 exists, bridge point cloud data can be extracted with high accuracy.

(Embodiment 3)
The configuration of the bridge data extraction device 100 according to the third embodiment will be described with reference to Fig. 10. Fig. 10 is a diagram for explaining an example of the bridge data extraction device according to the third embodiment.

[System configuration]
The bridge data extraction device 100 shown in Fig. 10 is a device that accurately extracts point cloud data representing bridges from point cloud data representing a target area. As shown in Fig. 10, the bridge data extraction device 100 includes an acquisition unit 13, a direction calculation unit 14, a determination unit 11, a selection unit 12, and a bridge determination unit 101.

The acquisition unit 13, direction calculation unit 14, determination unit 11, and selection unit 12 have already been described in embodiments 1 and 2, so descriptions of the acquisition unit 13, direction calculation unit 14, determination unit 11, and selection unit 12 will be omitted.

When the bridge determination unit 101 performs a clustering process based on the distances between points included in point cloud data B and the direction of the points belonging to the generated cluster can be considered to be vertical, the points included in the cluster to which the point belongs are considered to be bridge point cloud data (point cloud data A) representing a bridge.

Specifically, the bridge determination unit 101 first acquires point cloud data B. Next, the bridge determination unit 101 performs Euclidean clustering processing based on the distances between points included in point cloud data B to generate clusters.

Note that the clustering process in the bridge determination unit 101 is the same as the clustering process in the selection unit 12, so the cluster Bj generated by the selection unit 12 may be used.

Next, the bridge determination unit 101 determines whether the cluster corresponds to a bridge based on the normals of the points belonging to the cluster. The bridge determination unit 101 calculates the vertical component of the normals of the points belonging to the cluster, and determines whether the vertical component of the normals is greater than a preset vertical threshold. The vertical threshold can be set to 0.5, for example.

Next, if the direction of a point belonging to a cluster can be considered to be vertical, the bridge determination unit 101 determines that the points included in the cluster to which the point belongs are bridge point cloud data (point cloud data A) representing a bridge.

[Device operation]
Next, the operation of the bridge data extraction device in embodiment 3 will be described with reference to FIG. 11. FIG. 11 is a diagram for explaining an example of the operation of the bridge data extraction device in embodiment 3. In the following description, the diagram will be referenced as appropriate. Furthermore, in embodiment 3, a bridge data extraction method is implemented by operating the bridge data extraction device. Therefore, the description of the bridge data extraction method in embodiment 3 will be replaced by the following description of the operation of the bridge data extraction device.

In embodiment 3, the processes of steps A1 to A4 and step E1 shown in Figure 11 are executed. Since the processes of steps A1 to A4 have already been explained, a description of the processes of steps A1 to A4 will be omitted.

As shown in Figure 11, first, the bridge determination unit 101 generates clusters by performing Euclidean clustering processing based on the distances between points included in point cloud data B. Next, if the direction of a point belonging to a generated cluster can be considered to be vertical, the bridge determination unit 101 sets the points included in the cluster to which the point belongs as bridge point cloud data (point cloud data A) representing a bridge (step E1).

Specifically, in step E1, the bridge determination unit 101 first acquires point cloud data B. Next, in step E1, the bridge determination unit 101 performs Euclidean clustering processing based on the inter-point distances of points included in point cloud data B to generate clusters.

Note that the clustering process in the bridge determination unit 101 is the same as the clustering process in the selection unit 12, so the cluster Bj generated by the selection unit 12 may be used.

Next, in step E1, the bridge determination unit 101 determines whether the cluster corresponds to a bridge based on the normals of the points belonging to the cluster. The bridge determination unit 101 calculates the vertical component of the normals of the points belonging to the cluster, and determines whether the vertical component of the normals is greater than a preset vertical threshold.

Next, in step E1, if the direction of a point belonging to a cluster can be considered to be vertical, the bridge determination unit 101 determines that the points included in the cluster to which the point belongs are bridge point cloud data (point cloud data A) representing a bridge.

[Effects of the Third Embodiment]
As shown in Figure 3, the bridge 3 has an upper layer 6 (girders and deck plates) supported by a lower layer 7 (piers or abutments). Also, the arrow nv3 representing the normal unit vector of the upper layer 6 (girders and deck plates) in Figure 4 has a large vertical component.

In this way, clusters with points with large horizontal components, such as arrows nv1 and nv2 representing the normal unit vector of the side of the lower layer 7, and clusters with points with large vertical components, such as arrow nv3 representing the normal unit vector of the upper layer 6, can be determined to represent bridges.

Therefore, bridge point cloud data including points of the upper layer 6 and lower layer 7 can be extracted from the target area point cloud data.

(Embodiment 4)
The configuration of the bridge data extraction device 120 according to the fourth embodiment will be described with reference to Fig. 12. Fig. 12 is a diagram for explaining an example of the bridge data extraction device according to the fourth embodiment.

[System configuration]
The bridge data extraction device 120 shown in Fig. 12 is a device that accurately extracts point cloud data representing bridges from point cloud data representing a target area. As shown in Fig. 12, the bridge data extraction device 120 includes an acquisition unit 13, a direction calculation unit 14, a determination unit 11, a removal unit 81, a selection unit 82, and a bridge determination unit 101.

The acquisition unit 13, direction calculation unit 14, determination unit 11, removal unit 81, selection unit 82, and bridge determination unit 101 have already been described in embodiments 1, 2, and 3, so descriptions of the acquisition unit 13, direction calculation unit 14, determination unit 11, removal unit 81, selection unit 82, and bridge determination unit 101 will be omitted.

[Device operation]
Next, the operation of the bridge data extraction device in embodiment 4 will be described with reference to FIG. 13. FIG. 13 is a diagram for explaining an example of the operation of the bridge data extraction device in embodiment 4. In the following description, the diagram will be referenced as appropriate. Furthermore, in embodiment 4, a bridge data extraction method is implemented by operating the bridge data extraction device. Therefore, the description of the bridge data extraction method in embodiment 4 will be replaced by the following description of the operation of the bridge data extraction device.

In embodiment 4, the processes of steps A1 to A3, steps D1 to D2, and step E1 shown in FIG. 13 are executed. Since the processes of steps A1 to A3, steps D1 to D2, and step E1 have already been explained, explanation of the processes of steps A1 to A3, steps D1 to D2, and step E1 will be omitted.

[Effects of the Fourth Embodiment]
According to the fourth embodiment, even when point cloud data corresponding to the ground surface 2 exists, bridge point cloud data can be extracted with high accuracy.

According to embodiment 4, bridge point cloud data including points of the upper layer 6 and lower layer 7 can be extracted from the target area point cloud data.

[program]
The program in the first embodiment may be a program that causes a computer to execute steps A1 to A4 shown in FIG. 5, steps B1 to B2 shown in FIG. 6, and steps C1 to C3 shown in FIG.

The program in embodiment 2 may be a program that causes a computer to execute steps A1 to A3 and steps D1 to D2 shown in Figure 9.

The program in embodiment 3 may be a program that causes a computer to execute steps A1 to A4 and step E1 shown in Figure 9.

The program in embodiment 4 may be a program that causes a computer to execute steps A1 to A3, steps D1 to D2, and step E1 shown in FIG. 13.

By installing and running the above-mentioned program on a computer, the bridge data extraction device and bridge data extraction method according to embodiments 1 to 4 can be realized.

In the first embodiment, the computer processor functions as an acquisition unit 13, a direction calculation unit 14, a determination unit 11 (horizontal determination unit 15, size determination unit 16), and a selection unit 12 (height limitation unit 17, clustering unit 18, cluster selection unit 19), and performs processing.

In the case of embodiment 2, the computer processor functions as the acquisition unit 13, direction calculation unit 14, determination unit 11, removal unit 81, and selection unit 82, and performs processing.

In the third embodiment, the computer processor functions as the acquisition unit 13, direction calculation unit 14, determination unit 11, selection unit 12, and bridge determination unit 101, and performs processing.

In the fourth embodiment, the computer processor functions as the acquisition unit 13, direction calculation unit 14, determination unit 11, removal unit 81, selection unit 82, and bridge determination unit 101, and performs processing.

The programs in embodiments 1 to 4 may also be executed by a computer system constructed using multiple computers. In this case, for example, each computer may function as one of the acquisition unit 13, direction calculation unit 14, determination unit 11, selection unit 12, removal unit 81, selection unit 82, and bridge determination unit 101.

[Physical configuration]
Here, a computer that realizes the bridge data extraction device by executing the programs according to the first to fourth embodiments will be described with reference to Fig. 14. Fig. 14 is a diagram for explaining an example of a computer that realizes the bridge data extraction device according to the first to fourth embodiments.

As shown in FIG. 14, the computer 110 includes a CPU 111, a main memory 112, a storage device 113, an input interface 114, a display controller 115, a data reader/writer 116, and a communication interface 117. These components are connected to each other via a bus 121 so that they can communicate data with each other. Note that the computer 110 may include a GPU or FPGA in addition to or instead of the CPU 111.

The CPU 111 performs various calculations by expanding the programs (codes) in the embodiment stored in the storage device 113 into the main memory 112 and executing them in a predetermined order. The main memory 112 is typically a volatile storage device such as a DRAM (Dynamic Random Access Memory). The programs in the embodiment are provided in a state stored on a computer-readable recording medium 122. The programs in the embodiment may also be distributed over the Internet connected via the communication interface 117. The recording medium 122 is a non-volatile recording medium.

Specific examples of the storage device 113 include a hard disk drive and semiconductor storage devices such as flash memory. The input interface 114 mediates data transmission between the CPU 111 and input devices 118 such as a keyboard and mouse. The display controller 115 is connected to the display device 119 and controls the display on the display device 119.

The data reader/writer 116 mediates data transmission between the CPU 111 and the recording medium 122, reads programs from the recording medium 122, and writes the processing results of the computer 110 to the recording medium 122. The communication interface 117 mediates data transmission between the CPU 111 and other computers.

Specific examples of the recording medium 122 include general-purpose semiconductor storage devices such as CF (Compact Flash (registered trademark)) and SD (Secure Digital), magnetic recording media such as flexible disks, and optical recording media such as CD-ROMs (Compact Disk Read Only Memory).

Note that the bridge data extraction device in the embodiments can also be realized using hardware corresponding to each component, rather than a computer with a program installed. Furthermore, the bridge data extraction device may be partially realized by a program and the remaining components by hardware.

The invention has been described above with reference to embodiments, but the invention is not limited to the above-described embodiments. Various modifications that would be understandable to a person skilled in the art can be made to the configuration and details of the invention within the scope of the invention.

The above description makes it possible to accurately extract point cloud data representing bridges from point cloud data representing a target area. This is also useful in fields where bridge analysis is required.

REFERENCE SIGNS LIST 1 Target area 2 Ground surface 3 Bridge 4 Natural object 5 Structure 6 Upper layer 7, 7a, 7b Lower layer 10, 80, 100, 120 Bridge data extraction device 11 Determination unit 12, 82 Selection unit 13 Acquisition unit 14 Direction calculation unit 15 Horizontal determination unit 16 Size determination unit 17 Height limitation unit 18 Clustering unit 19 Cluster selection unit 20 Sensor 30 Storage device 81 Removal unit 101 Bridge determination unit 110 Computer 111 CPU
112 Main memory 113 Storage device 114 Input interface 115 Display controller 116 Data reader/writer 117 Communication interface 118 Input device 119 Display device 121 Bus 122 Recording medium 200 System

Claims (9)

a determination means for determining that a point is lower layer point cloud data representing a pier or abutment of the bridge when the direction of the point determined using a point included in target area point cloud data representing a target area including a bridge and a plurality of points included in a preset range around the point can be regarded as horizontal;
a selection means for executing a division process based on distances between points included in the target area point cloud data, and selecting bridge point cloud data representing the bridge based on the division result;
A bridge data extraction device having the above. The bridge data extraction device according to claim 1,
When the direction is a normal unit vector, the determination means calculates the magnitude of the horizontal component of the normal unit vector, and determines whether the point is included in the lower layer point cloud data based on the calculated magnitude of the horizontal component. 3. The bridge data extraction device according to claim 1 or 2,
The determination means performs a clustering process based on distances between points included in the lower layer point cloud data to generate clusters, and generates the lower layer point cloud data based on the points belonging to the generated clusters. 4. The bridge data extraction device according to claim 1,
The selection means performs a clustering process based on distances between points included in the target area point cloud data to generate clusters, selects a cluster including points included in the lower layer point cloud data, and sets the points included in the selected cluster as the bridge point cloud data. 5. The bridge data extraction device according to claim 4,
The determination means generates height information for limiting the points included in the target area point cloud data by vertical height based on the points included in the lower layer point cloud data, and extracts point cloud data to be used in the clustering process based on the height information. 6. The bridge data extraction device according to claim 1,
The selection means generates clusters by performing a clustering process based on distances between points included in the point cloud data obtained by removing the lower layer point cloud data from the target area point cloud data, selects the clusters based on points included in the lower layer point cloud data, and sets points belonging to the selected clusters as the bridge point cloud data. 7. The bridge data extraction device according to claim 6 ,
A bridge data extraction device that performs clustering processing based on the distances between points belonging to the selected cluster, and when the direction of the points belonging to the cluster generated can be considered to be vertical, the points included in the cluster to which the point belongs are set as the bridge point cloud data. The computer
If the direction of a point determined using a point included in target area point cloud data representing a target area including a bridge and a plurality of points included in a predetermined range around the point can be considered to be horizontal, the point is determined to be lower layer point cloud data representing a pier or abutment of the bridge,
A bridge data extraction method comprising: performing a division process based on distances between points included in the target area point cloud data; and selecting bridge point cloud data representing the bridge based on the division result. On the computer,
If the direction of a point determined using a point included in target area point cloud data representing a target area including a bridge and a plurality of points included in a predetermined range around the point can be considered to be horizontal, the point is determined to be lower layer point cloud data representing a pier or abutment of the bridge,
A program that executes a division process based on distances between points included in the target area point cloud data, and selects bridge point cloud data representing the bridge based on the division results.