JP5339953B2 - 3D map correction apparatus and 3D map correction program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To acquire a three-dimensional map having error accuracy equivalent to a large scale (e.g. scale 1/500) without carrying out individual surveys separately from the ground. <P>SOLUTION: The three-dimensional map correcting device is provided with a wall vector-computing section 4 for computing wall vector data D indicating the position and shape of the wall surface of a building which is a feature existing around a vehicle, based on: feature point group data A indicating a point group of three-dimensional positions of features existing around the vehicle; and travel vector data C indicating the travel locus of the vehicle. A map polygon-correcting section 9 corrects wall polygon data E and roof polygon data F of the three-dimensional map stored in a three-dimensional map storage section 8 according to the wall vector data D computed by the wall vector-computing section 4. <P>COPYRIGHT: (C)2010,JPO&amp;INPIT

Description

  The present invention relates to a three-dimensional map correction apparatus and a three-dimensional map correction program for correcting a three-dimensional map (particularly, the position and shape of a building wall and a roof) with low accuracy.

Conventionally, a two-dimensional map has been created by aerial surveying or surveying on the ground by a surveying instrument.
Moreover, the height of the roof of the building is recorded by performing a stereoscopic view by aerial survey or by adding altitude measurement including a laser measurement device and a GPS (Global Positioning System) device in aerial survey. Three-dimensional maps have been created.
These 3D maps are adjusted to a standard error of 1.75 m (Ministry of Land, Infrastructure, Transport and Tourism) when the scale is 1/2500 due to restrictions such as the distance between the surveying device by aerial survey and the ground to be surveyed. There is a need.
Thus, in order to reduce the error of the 3D map, it is necessary to correct the 3D map by surveying individual things that are troublesome.

In Non-Patent Document 1 below, a laser measurement device, a GPS device, an inertial navigation device, and an integrated odometer are mounted on a vehicle, so that the position of the feature surface is held in three-dimensional position coordinates. Ground point cloud acquisition means for acquiring as point cloud data is disclosed.
However, this point cloud data does not hold information such as the meaning of the feature (such as a building wall) or the three-dimensional structure of the feature (the position of the three-dimensional plane representing the wall).

In Patent Document 1 below, a three-dimensional building model group is created from three-dimensional point cloud data obtained by irradiating a laser from an aircraft to the ground, and a side surface of an urban area is continuously measured by a vehicle or the like. A technique for automatically creating side surface irregularities and integrating the three-dimensional building model group with the side irregularities is disclosed.
However, Patent Document 1 does not disclose a technique for correcting an error that is inevitably included in position information of a building obtained from an aircraft.
In addition, the data obtained by measuring the side of an urban area has effects that cannot be avoided (effects of pedestrians, vehicles, utility poles, planted trees, etc.), but no technology is disclosed for removing those effects. .

Patent Document 2 below discloses a technique for creating a building model from three-dimensional point cloud data obtained by irradiating a laser from a vehicle and an image obtained from a camera mounted on the vehicle. .
However, in the case of Patent Document 2, it is necessary to always obtain an image of a building from a vehicle, and Patent Document 2 does not disclose a technique for correcting an error of an existing three-dimensional city model.

In Patent Document 3 below, three-dimensional point cloud data obtained by irradiating a ground with laser from an aircraft is divided into ground and features (buildings), and contour data such as buildings is obtained by processing aerial photographs. And a technique for creating a three-dimensional city model from these three-dimensional point cloud data and contour data is disclosed.
However, Patent Document 3 does not disclose a technique for correcting an error that is inevitably included in building position information obtained from an aircraft.

Patent Document 4 below discloses a technique for creating a building model by separating the ground and the building based on the height information of a three-dimensional point group obtained by irradiating a laser beam from an aircraft to the ground. .
However, Patent Document 4 does not disclose a technique for correcting an error of the building model.

Patent Document 5 below discloses a technique for creating a three-dimensional model by associating three-dimensional point cloud data obtained by laser irradiation with image data. This image data is configured to perform position correction assuming an error due to vehicle travel.
However, Patent Document 5 does not disclose a technique for correcting a model having an existing error because it is assumed that the three-dimensional point cloud data and the image data have the same error due to vehicle travel.

Yoshida, et al. “Mobile Mapping System”, Mitsubishi Electric Technical Report Vol. 81 no. 7 2007

JP 2003-256871 A (paragraph numbers [0068] to [0069]) JP 2002-031528 A (paragraph number [0008]) JP 2002-74323 A (paragraph number [0010]) JP 2003-296706 A (paragraph number [0006]) JP 2004-348575 A (paragraph number [0012])

Since the conventional three-dimensional map correction device is configured as described above, a three-dimensional map having an error accuracy equivalent to 1/500 scale is generated from a point cloud acquired by a radar mounted on an aircraft. In order to have this error accuracy, there has been a problem that separate surveying is required from the ground.
In addition, it is possible to create road surfaces and wall surfaces of roads as three-dimensional maps having error accuracy equivalent to 1/500 scale from points obtained by radar mounted on vehicles. In addition to the difficulty of creating the roof portion, there is a problem that it is difficult to correct the three-dimensional map by the point cloud acquired with a small scale (for example, a scale of 1/2500).
If all the ground features are configured as 3D objects from the point cloud acquired by the radar mounted on the vehicle, pedestrians, vehicles, roadside trees, It is necessary to eliminate the effects of traffic signs, but it was difficult to eliminate the effects.

  The present invention has been made to solve the above-described problems, and has an error accuracy corresponding to a large scale (for example, a scale of 1/500) without performing separate surveying that requires labor from the ground. It is an object of the present invention to obtain a 3D map correction apparatus and a 3D map correction program capable of acquiring a 3D map.

The three-dimensional map correction apparatus according to the present invention includes a travel vector calculating means for calculating a travel vector indicating a direction in which the vehicle is traveling from the travel locus data of the vehicle, and a three-dimensional position of a feature existing around the vehicle. Wall vector calculation for calculating wall vector data indicating the position and shape of the wall surface of the building, which is a feature existing around the vehicle, from the feature point cloud data indicating the point cloud and the travel vector calculated by the travel vector calculation means And the map polygon correction means corrects the wall polygon data and the roof polygon data of the 3D map stored in the 3D map storage means according to the wall vector data calculated by the wall vector calculation means. those walls vector calculation means, from the point cloud of the three-dimensional position indicated by the feature point group data stored by the data storage means, outline Extracted by the vertical point cloud extraction unit according to the horizontal distance between the vertical point cloud extraction unit that extracts the point cloud of the three-dimensional position continuous in the vertical direction and the travel vector calculated by the travel vector calculation means. A vertical point cloud alignment unit that divides a point cloud at a three-dimensional position into a plurality of sets, and a travel vector calculated by the travel vector calculation means in the plurality of point cloud groups divided by the vertical point cloud alignment unit; And a wall point cloud vectorization unit that calculates, as wall vector data, vector data indicating a region where a point cloud having the maximum horizontal distance exists.

  According to this invention, the travel vector calculation means for calculating the travel vector indicating the direction in which the vehicle is traveling from the travel locus data of the vehicle, and the ground indicating the point group of the three-dimensional positions of the features existing around the vehicle. Wall vector calculation means for calculating wall vector data indicating the position and shape of the wall surface of the building, which is a feature existing around the vehicle, from the object point cloud data and the travel vector calculated by the travel vector calculation means; The map polygon correction means is configured to correct the wall polygon data and the roof polygon data of the 3D map stored in the 3D map storage means according to the wall vector data calculated by the wall vector calculation means. 3D map with error accuracy equivalent to a large scale (for example, 1/500 scale) without carrying out individual surveys There is an effect that can be obtained.

It is a block diagram which shows the three-dimensional map correction apparatus by Embodiment 1 of this invention. It is explanatory drawing which shows the data format of the feature point cloud data A and the driving | running | working locus data B which are memorize | stored by the on-vehicle measurement data storage part 2. FIG. It is explanatory drawing which shows the data format of the traveling vector data C output from the traveling vector calculation part 3. FIG. It is explanatory drawing which shows the data format of the wall vector data D calculated by the wall vector calculation part 4. FIG. It is explanatory drawing which shows the data format of the wall polygon data E and the roof polygon data F of the three-dimensional map memorize | stored by the three-dimensional map memory | storage part 8. FIG. It is explanatory drawing which shows the processing content of the wall vector calculation part. It is explanatory drawing which shows the processing content of the map polygon correction | amendment part. It is a block diagram which shows the principal part of the three-dimensional map correction apparatus by Embodiment 3 of this invention. It is explanatory drawing which shows the processing content of the road surface point group supplement part.

Embodiment 1 FIG.
1 is a block diagram showing a three-dimensional map correction apparatus according to Embodiment 1 of the present invention.
In FIG. 1, an on-vehicle point cloud acquisition unit 1 is a device in which, for example, a laser measuring device, a GPS device, an inertial navigation device, and an integrated odometer are combined. Implement the process of measuring the three-dimensional position.
Here, as disclosed in Non-Patent Document 1, the GPS device can measure its own position with an accuracy of a horizontal error of about 2 cm. Further, the laser measuring device can measure the distance with an accuracy of 10 cm with respect to the distance of about 10 m. From this, the three-dimensional position can acquire the position of the surface of the feature with the accuracy (error within 25 cm) of the map of 1/500 scale.
The on-vehicle point cloud acquisition unit 1 constitutes a feature position measuring unit.

The on-vehicle measurement data storage unit 2 stores the feature point cloud data A indicating the point cloud of the three-dimensional position measured by the on-vehicle point cloud acquisition unit 1, and the traveling locus data B indicating the traveling locus of the vehicle. The process to memorize is executed. The on-vehicle measurement data storage unit 2 constitutes data storage means.
The traveling vector calculation unit 3 determines the direction in which the vehicle is traveling from the traveling locus data B of the portion of the traveling locus that is relatively close to the straight line in the traveling locus data B stored in the on-vehicle measurement data storage unit 2. A process of calculating the travel vector shown and outputting travel vector data C including information on the travel direction of the vehicle is performed. The travel vector calculation unit 3 constitutes a travel vector calculation unit.

The wall vector calculation unit 4 includes a vertical point group extraction unit 5, a vertical point group alignment unit 6, and a wall point group vectorization unit 7, and the feature point group data stored in the on-vehicle measurement data storage unit 2. A process of calculating wall vector data D indicating the position and shape of the wall surface of the building, which is a feature existing around the vehicle, is performed from A and the travel vector data C calculated by the travel vector calculation unit 3. The wall vector calculation unit 4 constitutes a wall vector calculation means.
The vertical point group extraction unit 5 is a point of the three-dimensional position that is substantially continuous in the vertical direction from the point group of the three-dimensional position indicated by the feature point group data A stored in the on-vehicle measurement data storage unit 2. A process of extracting a group is performed.

The vertical point group aligning unit 6 divides the point group of the three-dimensional position extracted by the vertical point group extracting unit 5 into a plurality of sets according to the horizontal distance from the traveling vector data C calculated by the traveling vector calculating unit 3. Perform the process.
The wall point group vectorization unit 7 is a point group having a maximum horizontal distance from the travel vector data C calculated by the travel vector calculation unit 3 among the plurality of point groups divided by the vertical point group alignment unit 6. A process of calculating vector data indicating a region where the is present as wall vector data is performed.

  The three-dimensional map storage unit 8 is, for example, a recording medium such as a database, and represents wall polygon data E that represents the wall surface of the building on the ground as a polygon (polygon) and the roof of the building on the ground as a polygon (polygon). A three-dimensional map composed of the roof polygon data F is stored. The 3D map storage unit 8 constitutes 3D map storage means.

The map polygon correction unit 9 includes a wall polygon search unit 10, a wall polygon correction unit 11, and a related polygon correction unit 12, and a 3D map storage unit 8 according to the wall vector data D calculated by the wall vector calculation unit 4. The processing for correcting the wall polygon data E and the roof polygon data F of the three-dimensional map stored in the above is executed. The map polygon correction unit 9 constitutes a map polygon correction unit.
The wall polygon search unit 10 is a wall of a building existing in the vicinity of the position indicated by the wall vector data D calculated by the wall vector calculation unit 4 from the wall polygon data stored in the 3D map storage unit 8. A process of searching for the wall polygon data E related to is performed.

The wall polygon correction unit 11 performs processing for correcting the position and shape represented by the wall polygon data searched by the wall polygon search unit 10 according to the wall vector data D.
The related polygon correction unit 12 is connected adjacent to the uncorrected wall polygon data E in accordance with the position and shape represented by the corrected wall polygon data G, which is the wall polygon data E corrected by the wall polygon correction unit 11. Processing for correcting the position and shape represented by the wall polygon data E and the roof polygon data F is performed.

  The corrected three-dimensional map storage unit 13 is a recording medium such as a database, for example, and stores the corrected wall polygon data G whose position and shape are corrected by the map polygon correction unit 9 and the corrected roof polygon whose position and shape are corrected. Data H is stored.

  In the example of FIG. 1, an on-vehicle point cloud acquisition unit 1, an on-vehicle measurement data storage unit 2, a travel vector calculation unit 3, a wall vector calculation unit 4, and a map polygon correction unit 9 that are components of the three-dimensional map correction device. It is assumed that each is composed of dedicated hardware (for example, a semiconductor integrated circuit on which a CPU is mounted), but if the 3D map correction device is composed of a computer, an on-vehicle point cloud is acquired. Storing a three-dimensional map correction program in which the processing contents of the unit 1, on-vehicle measurement data storage unit 2, travel vector calculation unit 3, wall vector calculation unit 4 and map polygon correction unit 9 are described in the memory of the computer, The CPU of the computer may execute a 3D map correction program stored in the memory.

FIG. 2 is an explanatory diagram showing the data format of the feature point cloud data A and the traveling locus data B stored in the on-vehicle measurement data storage unit 2.
FIG. 3 is an explanatory diagram showing a data format of the traveling vector data C output from the traveling vector calculation unit 3.
FIG. 4 is an explanatory diagram showing the data format of the wall vector data D calculated by the wall vector calculation unit 4.
FIG. 5 is an explanatory diagram showing the data format of the wall polygon data E and the roof polygon data F of the 3D map stored in the 3D map storage unit 8.

Next, the operation will be described.
The on-vehicle point cloud acquisition unit 1 is, for example, a device in which a laser measurement device and a GPS device are combined, and is installed on the roof of the vehicle.
Here, the laser measurement device irradiates laser light in the direction around the vehicle and receives the laser light that has been reflected back to the features existing around the vehicle, so that the vehicle to the features can be received. Earn distance and direction.
The GPS device has a GPS receiving antenna installed at a position where the laser measuring device is installed, and the GPS receiving antenna receives a GPS signal transmitted from a GPS satellite, so that the position (latitude, longitude) of the laser measuring device is received. ) And the reception time t of the GPS signal.

When the GPS device acquires the position (latitude, longitude, altitude) of the laser measurement device and the laser measurement device acquires the distance and direction from the vehicle to the feature, the on-vehicle point cloud acquisition unit 1 A three-dimensional position (x (longitude), y (latitude), z (elevation) coordinate value) of points constituting the surface of the feature is obtained.
As shown in FIG. 2A, the on-vehicle point cloud acquisition unit 1 receives x (longitude), y (latitude), z (elevation) coordinate values of points constituting the surface of the feature, and GPS signal reception. The time t (coordinate value acquisition time) is output to the on-vehicle measurement data storage unit 2 as the feature point cloud data A.

The unit of coordinate values is, for example, a planar rectangular coordinate system, in which x (distance in the east-west direction) and y (distance in the north-south direction) from the origin of each system are expressed in m units.
The density of the point group is, for example, 10 points or more at 1 cm intervals at a distance of 1 m from the laser measuring apparatus and 2 points or more at 1 cm intervals at a distance of 5 m.
Further, the on-vehicle point cloud acquisition unit 1 acquires the position measured by the GPS device at a minimum interval of 0.1 seconds as the traveling locus data B, and the traveling locus data B is stored in the on-vehicle measurement data storage unit 2. Output.
As shown in FIG. 2 (b), the traveling locus data B includes a traveling locus of a vehicle (specifically, a GPS antenna in the vehicle), x (longitude), y (latitude), z (elevation) coordinate values, This is expressed by the coordinate value acquisition time t.
Note that the method of expressing the longitude and latitude in the feature point cloud data A and the travel locus data B is the same.

  The on-vehicle measurement data storage unit 2 stores the feature point cloud data A and the traveling locus data B output from the on-vehicle point cloud acquisition unit 1.

The traveling vector calculation unit 3 is a direction in which the vehicle is traveling from the traveling locus data B of the portion of the traveling locus that is relatively close to the straight line in the traveling locus data B stored in the on-vehicle measurement data storage unit 2. The travel vector data C including the travel direction information of the vehicle is output to the wall vector calculation unit 4.
As shown in FIG. 3, the traveling vector data C is a starting point (x (longitude), y (latitude), z (elevation) coordinate values) of a portion of the traveling locus that is relatively close to a straight line in the traveling locus data B. And the end point (x (longitude), y (latitude), z (elevation) coordinate values) and the traveling direction (the direction of the end point as seen from the start point: a numerical value expressed up to 360 degrees clockwise with 0 degrees north). It represents.

Hereinafter, the calculation processing of the traveling vector data C in the traveling vector calculation unit 3 will be specifically described.
First, the travel vector calculation unit 3 assumes a vector starting from the coordinate values x and y of the travel start time and having the coordinate value at the time when a certain time (for example, 10 seconds) has elapsed from the travel start time as the end point. .
If the distance between the assumed vector and a point on the travel locus data B is less than a predetermined threshold (for example, 0.3 m) at all times between the start point and the end point, the travel vector calculation unit 3 The assumed vector is registered as travel vector data C.
On the other hand, if the distance between the assumed vector and the point on the travel locus data B is equal to or greater than a predetermined threshold (for example, 0.3 m), the same process is performed by shifting the start point by a certain time (for example, 2 seconds). I do.

After the traveling vector data C is registered, the traveling vector calculation unit 3 assumes a new vector whose starting point is the same as the starting point of the traveling vector data C and whose end point is shifted by a certain time (for example, 2 seconds).
If the distance between the newly assumed vector and a point on the travel locus data B is less than a predetermined threshold (for example, 0.3 m), the travel vector calculation unit 3 uses the travel vector data C registered earlier. Reject and register a newly assumed vector as travel vector data C.
In this way, the portion of the traveling locus that is relatively close to a straight line in the traveling locus data B is calculated as the traveling vector data C.

When the travel vector calculation unit 3 calculates the travel vector data C, the wall vector calculation unit 4 calculates the vehicle surroundings from the travel vector data C and the feature point cloud data A stored in the on-vehicle measurement data storage unit 2. The wall vector data D indicating the position and shape of the wall surface of the building, which is a feature existing in, is calculated.
FIG. 6 is an explanatory diagram showing the processing contents of the wall vector calculation unit 4.
Hereinafter, the calculation process of the wall vector data D in the wall vector calculation part 4 is demonstrated concretely.

First, the vertical point group extraction unit 5 of the wall vector calculation unit 4 is arranged in a substantially vertical direction from point groups at three-dimensional positions indicated by the feature point group data A stored in the on-vehicle measurement data storage unit 2. Extract a point cloud of consecutive 3D positions.
Here, FIG. 6A illustrates a point group in which the vertical point group extraction unit 5 is vertically connected to adjacent points from a processing space (a space formed by a point group at a three-dimensional position indicated by the feature point group data A). A state in which (a point group of three-dimensional positions substantially continuous in the vertical direction) is extracted is schematically shown.

Specifically, a point group in which adjacent points are vertically connected is extracted as follows.
First, the vertical point cloud extraction unit 5 rounds off the x and y coordinate values of the feature point cloud data A in the space based on the point cloud to a unit of 0.1 m, for example.
Therefore, for example, the point (x = 3600.512, y = 200.456) and the point (x = 3600.508, y = 200.461) are the same (x = 3600.5, y = 200). .5) is treated as a point.
Next, the vertical point group extraction unit 5 sorts the points having the same rounding value in descending order of the z coordinate (elevation).

Among the points having the same rounded value, the vertical point cloud extraction unit 5 has a maximum z coordinate value that is equal to or greater than a predetermined threshold value (for example, 3 m), and a difference between the maximum z coordinate value and the minimum value is a predetermined threshold value. If the number of points having the same rounding value (for example, 1 m) is equal to or greater than a predetermined threshold (for example, 10 points), the corresponding points may be points constituting a wall. Therefore, the plurality of points are extracted as a vertical point group because they can be regarded as substantially vertical points (the difference between the x coordinate and the y coordinate is within 0.1 m each).
A point group in which adjacent points are connected almost vertically does not include a point that constitutes a road, a pedestrian, or a hedge because the maximum value of the z coordinate does not match. Further, since the difference between the maximum value and the minimum value of the z coordinate does not match, the protruding eaves is not included.
However, it is considered that the point group in which adjacent points are connected substantially vertically includes points constituting street trees, utility poles, etc. in addition to the walls.

When the vertical point cloud extraction unit 5 extracts the point cloud connected vertically, the vertical point cloud alignment unit 6 of the wall vector calculation unit 4 extracts the horizontal distance from the travel vector data C calculated by the travel vector calculation unit 3. Accordingly, the vertically connected point group is divided into a plurality of sets.
Here, FIG. 6B schematically shows a state in which the vertical point group aligning unit 6 divides the vertically connected point group into a plurality of sets according to the horizontal distance from the traveling vector data C. It represents.

Specifically, a point group in which adjacent points are connected substantially vertically is divided into a plurality of sets as follows.
First, the vertical point group aligning unit 6 uses an arbitrary point in a point group in which adjacent points are vertically connected as a reference, and the horizontal distance from that point to the traveling vector data C (only the x coordinate value and the y coordinate value). The distance in which the z coordinate value is not taken into consideration) is within a predetermined threshold (for example, 0.1 m), and the horizontal distance in the traveling direction of the traveling vector data C is the predetermined threshold (for example, 0.15 m). The points that are within are connected.
Next, the vertical point group aligning unit 6 uses a set of points in which the horizontal distance of a point farthest from the travel vector data C among a set of connected points is a predetermined threshold (for example, 2 m) or more. Get a set of point clouds aligned along C.
However, it is considered that the point cloud set includes points that constitute walls, but does not include points that constitute street trees, utility poles, and the like.

The wall point group vectorization unit 7 of the wall vector calculation unit 4 divides the point group, to which the vertical point group alignment unit 6 is vertically connected, into a plurality of sets. Vector data indicating a region where a point group having the maximum horizontal distance from the travel vector data C calculated by the calculation unit 3 is present is calculated as wall vector data D.
Here, FIG. 6C shows a wall in which the wall point cloud vectorization unit 7 shows a region where a point cloud having the maximum horizontal distance from the traveling vector data C exists among a plurality of pairs of point clouds. A mode that vector data D is computed is expressed typically.

Specifically, the wall vector data D in FIG. 4 is calculated as follows.
First, the wall point cloud vectorization unit 7 calculates the point having the highest altitude among the points having the closest horizontal distance to the starting point of the traveling vector data C for each of the point clouds constituting the plurality of point cloud sets. The coordinates are set as the coordinates of the start point of the wall vector data, and the coordinates of the point having the highest altitude among the points closest to the end point of the traveling vector data C are set as the coordinates of the end point of the wall vector data.
Further, a wall vector data candidate is calculated in which the direction from the start point coordinate to the end point coordinate is expressed as an angle in the clockwise direction from north to the wall direction.
Next, the wall point cloud vectorization unit 7 checks the horizontal distance from the travel vector data C for wall vector data candidates substantially parallel to the travel vector data C, and among the wall vector data candidates, the travel vector data C The farthest wall vector data candidate is set as wall vector data D.

Among the features existing around the vehicle, the three-dimensional map storage unit 8 includes wall polygon data E representing the wall surface of the building as a polygon (polygon) and the building roof as a polygon (polygon). A three-dimensional map consisting of roof polygon data F represented by is stored.
The wall polygon data E and the roof polygon data F are stored in the data format shown in FIG. 5, and the generality is not lost.
In addition to the wall polygon data E and the roof polygon data F, the 3D map storage unit 8 may store polygons representing the ground.
Note that the maximum error between the wall polygon data E and the roof polygon data F is 1.75 m if the scale is 1/2500.

When the wall vector calculation unit 4 calculates the wall vector data D, the map polygon correction unit 9 calculates the 3D map wall polygon data E and the roof polygon stored in the 3D map storage unit 8 according to the wall vector data D. Data F is corrected.
FIG. 7 is an explanatory diagram showing the processing contents of the map polygon correction unit 9.
Hereinafter, the correction processing of the wall polygon data E and the roof polygon data F in the map polygon correction unit 9 will be specifically described.

First, the wall polygon search unit 10 of the map polygon correction unit 9 indicates the position indicated by the wall vector data D calculated by the wall vector calculation unit 4 from the wall polygon data E stored in the 3D map storage unit 8. The wall polygon data E related to the wall of the building existing in the vicinity of is searched.
Here, FIG. 7A schematically shows that the wall polygon search unit 10 searches for the wall polygon data E related to the wall of the building existing in the vicinity of the position indicated by the wall vector data D. Represents.

Specifically, the wall polygon data E is searched as follows.
First, the wall polygon search unit 10 searches for wall polygon data E substantially parallel to the wall vector data D calculated by the wall vector calculation unit 4 from the wall polygon data E stored in the 3D map storage unit 8. Explore.
That is, the wall polygon search unit 10 has a difference between the wall vector data D and the wall direction in the wall polygon data E stored in the 3D map storage unit 8 within a predetermined threshold (for example, 10 degrees). A certain wall polygon data E is searched.

  Next, the wall polygon search unit 10 selects the wall polygon data E whose horizontal distance from the wall vector data D is less than a predetermined threshold (for example, 1 m) among the wall polygon data E substantially parallel to the wall vector data D. The wall polygon data E is searched for the wall polygon data E having the minimum horizontal distance from the wall vector data D.

When the wall polygon search unit 10 searches for the wall polygon data E corresponding to the wall vector data D, the wall polygon correction unit 11 corrects the position and shape represented by the wall polygon data E according to the wall vector data D.
Here, FIG. 7B shows a state where the wall polygon correction unit 11 corrects the position and shape represented by the wall polygon data E.

Specifically, the position and shape represented by the wall polygon data E are corrected as follows.
The wall polygon correction unit 11 replaces the x and y coordinate values of the wall polygon data E searched by the wall polygon search unit 10 with the x and y coordinate values of the corresponding wall vector data D, so that the wall polygon data E is searched. The polygon data E is corrected.
However, the laser beam emitted from the laser measuring device of the on-vehicle point cloud acquisition means 1 has a limited range, for example, it is assumed that it does not reach the rooftop of a high-rise building. The z coordinate value is not replaced with the z coordinate value of the wall vector data D.

When the wall polygon correction unit 11 corrects the wall polygon data E, the related polygon correction unit 12 corrects the corrected wall polygon data G according to the position and shape represented by the corrected wall polygon data G that is the corrected wall polygon data E. The position and shape represented by the wall polygon data E and the roof polygon data F that are connected adjacent to the wall polygon data E before the correction is corrected.
In other words, the related polygon correction unit 12 represents the wall polygon represented by the wall polygon data connected adjacent to the wall polygon data E before being corrected by the wall polygon correction unit 11 by the corrected wall polygon data E. Correct by moving in the same way as a wall polygon.
The roof polygon data F is corrected by moving the roof polygon so as to cover the moved wall polygon.

Here, FIG. 7C shows a state in which the related polygon correction unit 12 corrects the position and shape represented by the roof polygon data F connected adjacent to the wall polygon data E before correction. Yes.
Whether the wall polygon data E and the roof polygon data F are adjacently connected to the wall polygon data E before correction is determined based on whether the wall polygon data E and the roof polygon data F are connected. This can be determined by determining whether or not a shared outline is present.

The corrected 3D map storage unit 13 stores the corrected wall polygon data G and the corrected roof polygon data H output from the map polygon correcting unit 9.
In this way, the three-dimensional map correction device uses the on-vehicle point cloud acquisition unit 1 with less average error for the wall polygon data E and the roof polygon data F for calibrating the building stored in the three-dimensional map storage unit 8. Is corrected based on the feature point cloud data A and the traveling locus data B acquired in step (3) to obtain a three-dimensional map with a smaller average error.

  As is apparent from the above, according to the first embodiment, the traveling vector calculation unit 3 calculates traveling vector data C including traveling vector information indicating the traveling direction of the vehicle from the traveling locus data B of the vehicle. From the feature point cloud data A indicating the point cloud of the three-dimensional position of the feature existing around the vehicle and the travel vector data C calculated by the travel vector calculation unit 3, the feature existing around the vehicle A wall vector calculation unit 4 for calculating wall vector data D indicating the position and shape of a wall surface of a building, and a map polygon correction unit 9 according to the wall vector data D calculated by the wall vector calculation unit 4 Since it is configured to correct the wall polygon data E and the roof polygon data F of the three-dimensional map stored in the storage unit 8, an individual survey that takes time and labor from the ground separately. Without an effect capable of obtaining a three-dimensional map with error accuracy corresponding to the large scale (e.g., scale 500 minutes 1).

Embodiment 2. FIG.
In the first embodiment, the on-vehicle point cloud acquisition unit 1 mounted on the vehicle has acquired the feature point cloud data A and the travel locus data B. However, the laser measurement device and the GPS device are placed on the ground. Measure the feature point cloud data A, then move the laser measuring device and the GPS device to measure the new feature point cloud data A, and the trajectory connecting both points as the running trajectory data B You may make it memorize | store.

In the first embodiment, the absolute position (latitude, longitude, altitude) of the laser measuring device by the GPS device constituting the on-vehicle point cloud acquisition unit 1, the direction of laser irradiation by the laser measuring device, and the features Although it has shown about what specifies the three-dimensional position (x (longitude), y (latitude), z (elevation) coordinate value) of the point which comprises the surface of a feature from distance, the point which comprises the surface of a feature In order to correct the three-dimensional position, an individual precision measurement method that is a known technique such as a distance measuring instrument may be used.
Further, since the accuracy of position measurement by the GPS device is affected by the number of GPS satellites that can be captured by the GPS device, the feature point cloud data A with poor position measurement accuracy is not used for calculating the wall vector data D. It can also be configured.

In Embodiment 1 described above, the wall is described as being vertically configured from the ceiling portion to the portion in contact with the ground. However, the actual wall surface is often different from the vertical surface in the portion close to the ground.
Therefore, when the vertical point cloud extraction unit 5 extracts a vertical part where point clouds are continuous, for example, from the ceiling part to the part in contact with the ground, for example, from the ground to an altitude of 1 m, from an altitude of 1 m to 3 m, or The wall may be expressed by a plurality of planes according to the distance from the travel vector for each layer by calculating the relative altitude from the ground, such as an altitude of 3 m or more.

In Embodiment 1 described above, the vertical point cloud extraction unit 5 has been described as extracting a vertical portion where point clouds are continuous. However, the vertical point cloud extraction unit 5 constitutes a vertical cliff by a threshold value that determines verticality. A group may be extracted.
Thereby, not only the wall surface of the building but also a cliff may be extracted and applied to grasping a landslide.

  In the first embodiment, the vertical point group aligning unit 6 has a set of points whose horizontal distance of the point farthest from the travel vector data C is a threshold (for example, 2 m) or more and is aligned along the travel vector data C. As shown in the figure, the most distant point and a point whose horizontal distance is within a threshold (for example, 1 m) are included, so that not only the outer wall of the building but also the inner wall behind the veranda Windows can be considered walls.

In the first embodiment, even when there is a noise object such as a pedestrian, a road sign, or a tree between the vehicle and the wall, the wall of the building is not reflected by the noise object, and the point cloud that reaches the wall of the building Although described as being detected, there are cases in which the wall surface of the low-rise part of the building cannot be detected as a point cloud at all because it is blocked by a continuous low-rise protective wall.
In this case, since there is no point group of less than a certain altitude in the vertical point group that is far from the travel vector data C, the wall vector data D calculated by the vertical point group aligning unit 6 indicates “low-layer position uncertain”. A flag is added, and the corrected wall polygon data G, which is the wall polygon data E corrected by the wall vector data D, is also flagged as “low-layer position uncertain”, and the created 3D map is displayed to the user. You may be made to call attention.

  In the first embodiment, the wall polygon correction unit 11 selects the wall polygon data E whose position and direction are similar to the wall vector data D from the wall polygon data E stored in the 3D map storage unit 8. Although the correction target is shown, for example, even on a three-dimensional map with a scale of 1/2500, wall polygon data E created by precise surveying to have an error accuracy of 1/500 on a scale. May be excluded from the object of correction by the wall polygon correction unit 11 by adding a flag “with high accuracy”.

In the first embodiment, the case where the wall polygon data E and the roof polygon data F have a contour line shared to determine that both are adjacently connected is shown. There is also a method of originally expressing the wall polygon data E and the roof polygon data F as an integrated three-dimensional object.
In the case of a three-dimensional map expressed according to such a method, a related polygon such as a roof polygon can be obtained by correcting the wall polygon by the wall polygon correcting unit 11 without using the related polygon correcting unit 12. It can be configured to automatically correct.

In the first embodiment, the description has been made so that only one surface in the wall polygon data E constituting the building is corrected. However, the on-vehicle point cloud may be used by an intersecting road or a vehicle traveling on an adjacent road. As a result, the plurality of wall polygon data E constituting the building may be corrected.
In this case, only the uncorrected wall polygon data E and roof polygon data F can be corrected by the related polygon correction unit 12.

The first embodiment has been described on the assumption that a building exists in the vicinity of the position indicated by the wall vector data D and the wall polygon data E relating to the wall of the building exists. When polygon data E does not exist, it is determined that there is a possibility that a new building has been formed due to secular change, and new corrected wall polygon data G is generated based on the corresponding wall vector data E You can also
However, in this case, since it is impossible to generate the roof polygon data H, the correction wall polygon data G is flagged as “created from on-vehicle measurement data” and the 3D map creator's attention is given. It can also be configured to prompt.

  The first embodiment has been described on the assumption that a building exists in the vicinity of the position indicated by the wall vector data D and the wall polygon data E relating to the wall of the building exists. When the polygon data E does not exist, not only the correction of the corresponding wall polygon data is performed, but it is determined that there is a possibility that the building has disappeared due to secular change. It is also possible to generate correction wall polygon data G with a flag of “possibility” and call the attention of the 3D map creator.

  In Embodiment 1 described above, the processing is all performed automatically. However, the 3D map storage unit 8 stores a photograph or video image in which the wall direction is taken from the road, and the wall polygon correction unit 11 The image of the frame of the picture or video image in which the relevant wall direction was taken from the road is displayed on the screen according to the position, and the operator selects the corresponding wall interactively while viewing the image, or The corresponding correction wall polygon data G indicated by the wall polygon correction unit 11 may be further corrected by the operator.

In the first embodiment, the three-dimensional map storage unit 8 has been described in which the wall polygon data E and the roof polygon data F of the building, which is a feature, are stored. In the case where the wall surface including the ceiling is integrated data, the same effect as this description can be obtained if the processing is divided into a portion near vertical and a portion near horizontal for convenience.
In the case of a tunnel, the travel locus data B can be acquired by an inertial navigation device and an integrated odometer, not by a GPS device.

In the first embodiment, the correction of the three-dimensional map on the ground based on the data acquired by the on-vehicle point cloud acquisition unit 1 has been described. However, an underwater tertiary representing the distribution of underwater ruins and fishing reefs. The original map may be corrected.
When correcting an underwater three-dimensional map, a point cloud of an underwater feature is acquired by an underwater laser measuring device towed by a ship, and the point cloud is handled as the feature point cloud data A.
In addition, the position of the underwater laser measurement device itself can be obtained by a GPS device or an inertial navigation device installed on the ship and a laser point cloud emitted from the ship toward the underwater laser measurement device. The travel locus data B is handled.

In the first embodiment, the correction of a three-dimensional map representing a building on the ground is described.
On the other hand, in a model of a museum, a movie set, various experimental facilities, etc., a three-dimensional design drawing is obtained and an object is arranged according to the design drawing. However, it may be difficult to verify whether or not the object is correctly arranged.
In such a case, a three-dimensional design drawing is stored in the three-dimensional storage unit 8, a laser measuring device is installed on the carriage to obtain the feature point cloud data A, and the carriage is moved to Based on the position, travel locus data B is obtained by an inertial navigation device and an integrated odometer.
Thus, the corrected wall polygon data G may be obtained and output to prompt the object locator to reposition the object.

Embodiment 3 FIG.
FIG. 8 is a block diagram showing a main part of a three-dimensional map correction apparatus according to Embodiment 3 of the present invention. In the figure, the same reference numerals as those in FIG.
The road surface point group replenishment unit 20 is blocked by a point group (for example, a point group on a space constituting a ground object such as a roadside tree, a telephone pole, a pedestrian, and a parked vehicle) that is substantially continuous in a vertical direction. Then, a process of replenishing the point cloud is performed on the road surface that does not reach the laser beam of the laser measuring device and does not constitute the point cloud.
The wall vector calculation unit 4 and the road surface point group supplementation unit 20 constitute a wall vector calculation unit.

FIG. 9 is an explanatory diagram showing the processing contents of the road surface point group supplementing unit 20.
In FIG. 9, among the point groups of the three-dimensional position indicated by the feature point group data A stored in the on-vehicle measurement data storage unit 2, the vertical point group (the three-dimensional position substantially continuous in the vertical direction). A state in which points on the road surface obstructed by the points constituting the ground object are supplemented as supplementary road surface point group data I to the space of the remaining point group excluding the point group) is shown.

Next, the operation will be described.
The process of calculating the wall vector data D from the feature point cloud data A and the traveling vector data C, and correcting the wall polygoning data E and the roof polygon data F of the three-dimensional map based on the wall vector data D is performed as described above. This is the same as the first embodiment.

In the feature point cloud data A, a point cloud on the road surface such as a road or a site is also acquired by the on-board point cloud acquisition unit 1.
Therefore, only points with low elevation are regarded as a point group on the road surface, and further, the same processing as that of the wall point group vectorization unit 7 is performed on the point points of the road surface that are connected almost horizontally, thereby making it possible to vectorize the road surface. Is possible. Further, it is possible to correct the road surface in the three-dimensional map based on the vectorized road surface.
However, when there are ground objects such as roadside trees, power poles, pedestrians, and parked vehicles on the ground, the laser light emitted from the laser measuring device of the on-vehicle point cloud acquisition unit 1 does not reach the road surface. As shown in the space by the point cloud in a), the point cloud where the point cloud is not formed is generated on the road surface behind the ground object. For this reason, lack of vectorization of the road surface occurs.

Therefore, in the third embodiment, in order to cope with chipping in the point cloud, the point cloud on the road surface lacking by the ground object is supplemented.
The space formed by the residual point cloud is obtained by removing the point cloud connected vertically by the vertical point cloud extraction unit 5 from the point cloud at the three-dimensional position indicated by the feature point cloud data A (FIG. 9). See).
On the other hand, as shown in the first embodiment, by removing the point group corresponding to the wall from the vertically connected point group by the vertical point group aligning unit 6, the roadside tree, the telephone pole, the pedestrian, There are ground objects such as parked vehicles, which are distinguished from walls.

The road surface point group supplementation unit 20 acquires the minimum value of the altitude z in the surrounding range (for example, a range of 0.1 m) in the x and y coordinates of the ground object, and the minimum value is the road surface on which the corresponding ground object exists. To be regarded as an altitude.
Next, as shown in FIG. 9, the road surface point group supplementation unit 20 draws the normal line N in the horizontal direction from the travel locus data B at regular intervals (e.g., 0.05 m). Points are temporarily arranged on N at regular intervals (for example, 0.05 m).
The normal N passes through a certain area of the ground object and is temporarily arranged until reaching the maximum range where the point cloud exists (for example, 30 m from the travel locus data) or the area separated as a wall. These points are arranged as supplementary road surface point cloud data I.
The x and y coordinates of the supplementary road surface point group data I are the coordinates of the points temporarily placed on the normal line N, and the z coordinate is the altitude of the road surface on which the corresponding ground object exists.

Thereby, the point cloud is replenished on the road surface on the ground where the point cloud is blocked by the vertical dot cloud.
After the point cloud is replenished on the road surface, the road surface point cloud is vectorized by applying the same processing as the wall point cloud vectorization unit 7 to the road surface cloud.
The road surface vector can be used to correct road surface polygon data (not shown) on the three-dimensional map.

  1 on-vehicle point cloud acquisition unit (feature position measurement unit), 2 on-vehicle measurement data storage unit (data storage unit), 3 travel vector calculation unit (travel vector calculation unit), 4 wall vector calculation unit (wall vector calculation unit) ) 5 vertical point group extraction unit, 6 vertical point group alignment unit, 7 wall point group vectorization unit, 8 3D map storage unit (3D map storage unit), 9 map polygon correction unit (map polygon correction unit), DESCRIPTION OF SYMBOLS 10 Wall polygon search part, 11 Wall polygon correction | amendment part, 12 Related polygon correction | amendment part, 13 Correction | amendment 3D map memory | storage part, 20 Road surface point group supplement part (wall vector calculation means)

Claims (4)

A feature position measuring means for measuring a three-dimensional position of a feature existing around a traveling vehicle and feature point cloud data indicating a point cloud of the three-dimensional position measured by the feature position measuring means are stored. In addition, data storage means for storing travel locus data indicating the travel locus of the vehicle, and travel vector calculation for calculating a travel vector indicating the direction in which the vehicle is traveling from the travel locus data stored in the data storage means. The position and shape of the wall surface of the building, which is a feature existing around the vehicle, from the means, the feature point cloud data stored by the data storage means, and the travel vector calculated by the travel vector calculation means. Wall vector calculation means for calculating wall vector data to be shown, wall polygon data representing the wall surface of the building on the ground, and roof polygon data representing the roof of the building 3D map storage means storing a 3D map, and wall polygon data of the 3D map stored by the 3D map storage means according to the wall vector data calculated by the wall vector calculation means, Map polygon correction means for correcting the roof polygon data, and the wall vector calculation means is substantially vertical from the point group at the three-dimensional position indicated by the feature point cloud data stored in the data storage means. Is extracted by the vertical point cloud extraction unit according to the horizontal distance between the vertical point cloud extraction unit that extracts point clouds at three-dimensional positions that are continuous with the travel vector calculated by the travel vector calculation means. A vertical point cloud aligning unit that divides a point cloud at a three-dimensional position into a plurality of sets, and the traveling vector among the plurality of point cloud sets divided by the vertical point cloud aligning unit. Characterized in that it comprises a wall point cloud vectorization unit that calculates, as wall vector data, vector data indicating a region in which a point cloud having the maximum horizontal distance from the travel vector calculated by the output means exists. three-dimensional map correction device that. The wall vector calculating means is provided with a road surface point cloud supplementing unit that supplements the point cloud on a ground road surface that is blocked by a point cloud of a three-dimensional position that is substantially continuous in the vertical direction and does not exist. The three-dimensional map correction apparatus according to claim 1, wherein The map polygon correcting means is a wall related to the wall of the building existing in the vicinity of the position indicated by the wall vector data calculated by the wall vector calculating means from the wall polygon data stored in the three-dimensional map storage means. A wall polygon search unit that searches for polygon data, a wall polygon correction unit that corrects the position and shape represented by the wall polygon data searched by the wall polygon search unit according to the wall vector data, and correction by the wall polygon correction unit The position and shape represented by the wall polygon data and the roof polygon data connected adjacent to the wall polygon data before the wall polygon data is corrected are corrected according to the position and shape represented by the wall polygon data. according to claim 1 or claim 2 wherein, characterized in that it is composed of a relevant polygon correction unit Dimension map correction device. A feature position measurement processing procedure for measuring a three-dimensional position of a feature existing around a traveling vehicle, and feature point cloud data indicating a point cloud of the three-dimensional position measured by the feature position measurement processing procedure. A data storage processing procedure for storing the travel trajectory data indicating the travel trajectory of the vehicle and a travel vector indicating the direction in which the vehicle is traveling is calculated from the travel trajectory data stored by the data storage processing procedure. A building which is a feature existing around the vehicle from the travel vector calculation processing procedure to be performed, the feature point cloud data stored by the data storage processing procedure and the travel vector calculated by the travel vector calculation processing procedure A wall vector calculation processing procedure for calculating wall vector data indicating the position and shape of the wall surface, and a wall vector calculated by the wall vector calculation processing procedure. To execute a map polygon correction procedure for correcting the three-dimensional map of the roof polygon data representing a walls polygon data and the building roof of a three-dimensional map representing a wall surface of the ground building computer according Rudeta, said wall vector calculation The processing procedure is a vertical point group that extracts a point group at a three-dimensional position that is substantially continuous in the vertical direction from the point groups at the three-dimensional position indicated by the feature point cloud data stored by the data storage processing procedure. Vertical that divides the point group of the three-dimensional position extracted by the vertical point group extraction processing procedure into a plurality of sets according to the horizontal distance between the extraction processing procedure and the travel vector calculated by the travel vector calculation processing procedure The travel calculated by the travel vector calculation processing procedure in the set of a plurality of point groups divided by the point cloud alignment processing procedure and the vertical point cloud alignment processing procedure Three-dimensional map correction program for the horizontal distance between the vector which is characterized in that it run on the wall point cloud vectorization processing procedure for calculating the vector data as a wall vector data indicating an area in which there is a maximum point group .

Images