JP7744064B2 - Map generation device, map generation method, and program - Google Patents

Description

This disclosure relates to a map generation device and a map generation method for generating an environmental map constructed using point cloud data, as well as a program for implementing these.

In recent years, advances in sensing technology have attracted attention to SLAM (Simultaneous Localization and Mapping). SLAM is a technology in which a mobile object equipped with a camera simultaneously estimates its own position and creates an environmental map (see, for example, Patent Document 1). With SLAM, autonomously traveling mobile objects such as robots do not need to move randomly, but can instead move along an autonomous traveling map obtained from an environmental map, thereby improving their movement efficiency.

Specifically, the mobile object acquires image data output from the camera for each frame, searches for feature points in the latest frame that correspond to feature points extracted from previous frames, and extracts pairs of feature points consisting of corresponding feature points. The mobile object then uses the pairs of feature points to calculate the camera matrix for the latest frame, and for each pair of feature points, calculates the three-dimensional coordinates of the feature points using the camera matrix and the two-dimensional coordinates of the feature points in the frame.

After calculating the camera matrix, the mobile body calculates the camera position from the camera matrix and performs self-location estimation. Furthermore, the mobile body generates or updates an environmental map (hereinafter referred to as a "3D environmental map") composed of a 3D point cloud using the feature points whose 3D coordinates have been calculated. The mobile body then generates a map for autonomous driving by converting the 3D environmental map into 2D, and performs autonomous driving using the map for autonomous driving.

In addition to cameras, depth sensors are also used as sensors, capable of measuring the distance (depth) to an object for each pixel of image data. Using a depth sensor allows for a higher density of the point cloud that makes up the environmental map compared to using a camera, enabling highly accurate autonomous driving. A specific example of a depth sensor is a LiDAR (light detection and ranging) sensor (see, for example, Patent Document 2).

Japanese Patent Application Laid-Open No. 2020-76714 Japanese Patent Application Laid-Open No. 2021-124496

With the SLAM technology described above, if an object is present in front of the camera, its feature points are extracted and a point cloud of this object is constructed on the environmental map. However, if there is an obstruction smaller than the object between the object and the camera, incorrect feature points may be extracted at the boundary of the area on the object that is shaded by the obstruction. In such cases, an environmental map with low accuracy is generated, resulting in a significant reduction in the accuracy of autonomous driving.

One example of the objective of this disclosure is to suppress a decrease in the accuracy of the environmental map even when an obstruction exists between the sensor and the object.

In order to achieve the above object, a map generating device according to one aspect of the present disclosure includes:
a data acquisition unit that acquires image data output from the depth sensor on a frame-by-frame basis;
a feature point extraction unit that extracts feature points from the image data for each frame;
a filtering unit that selects a specific feature point from the extracted feature points for each frame, and excludes the selected specific feature point when the depth of the selected specific feature point is greater than the minimum value of depths of a plurality of other feature points that are aligned with the selected specific feature point in a specific direction of the image by a threshold value or more;
an environment map generation unit that identifies pairs of corresponding feature points between frames, calculates three-dimensional coordinates of the feature points identified as pairs, and generates an environment map made up of a collection of the feature points identified as pairs using the calculation results;
The present invention is characterized in that it is provided with:

In order to achieve the above object, a map generation method according to one aspect of the present disclosure includes:
a data acquisition step of acquiring image data output from the depth sensor on a frame-by-frame basis;
a feature point extraction step of extracting feature points from the image data for each frame;
a filtering step of selecting a specific feature point from the extracted feature points for each frame, and excluding the selected specific feature point when the depth of the selected specific feature point is greater than the minimum value of the depths of a plurality of other feature points that are aligned with the selected specific feature point in a specific direction of the image by a threshold value or more;
an environmental map generation step of identifying pairs of corresponding feature points between frames, calculating three-dimensional coordinates of the identified pairs of feature points, and using the calculation results to generate an environmental map configured from a collection of the identified pairs of feature points;
The present invention is characterized by having the following:

Furthermore, in order to achieve the above object, a program according to one aspect of the present disclosure includes:
On the computer,
a data acquisition step of acquiring image data output from the depth sensor on a frame-by-frame basis;
a feature point extraction step of extracting feature points from the image data for each frame;
a filtering step of selecting a specific feature point from the extracted feature points for each frame, and excluding the selected specific feature point when the depth of the selected specific feature point is greater than the minimum value of the depths of a plurality of other feature points that are aligned with the selected specific feature point in a specific direction of the image by a threshold value or more;
an environmental map generation step of identifying pairs of corresponding feature points between frames, calculating three-dimensional coordinates of the identified pairs of feature points, and using the calculation results to generate an environmental map configured from a collection of the identified pairs of feature points;
The method is characterized in that:

As described above, according to the present disclosure, it is possible to suppress a decrease in the accuracy of the environmental map even when an obstruction exists between the sensor and the object.

FIG. 1 is a diagram showing a schematic configuration of a map generating device. FIG. 2 is a diagram specifically showing the configuration of the map generating device. FIG. 3 is an explanatory diagram for explaining the function of the filtering unit. FIG. 4 is a diagram showing an example of a three-dimensional environmental map. FIG. 5 is a flow diagram showing the operation of the map generating device. FIG. 6 is a block diagram showing an example of a computer that realizes the map generating device.

(Embodiment 1)
A map generating device, a map generating method, and a program according to an embodiment will be described below with reference to FIGS.

[Device configuration]
First, the schematic configuration of a map generating device according to an embodiment will be described with reference to Fig. 1. Fig. 1 is a diagram showing the schematic configuration of a map generating device.

In the embodiment shown in Figure 1, the map generation device 10 is a device for generating an environmental map constructed from point cloud data. As shown in Figure 1, the map generation device 10 includes a data acquisition unit 11, a feature point extraction unit 12, a filtering unit 13, and an environmental map generation unit 14.

The data acquisition unit 11 acquires image data output from the depth sensor on a frame-by-frame basis. The feature point extraction unit 12 extracts feature points from the image data acquired by the data acquisition unit 11 for each frame.

The filtering unit 13 selects a specific feature point from the extracted feature points for each frame. Furthermore, if the depth of the selected specific feature point is greater than the minimum depth of multiple other feature points that are aligned with the selected specific feature point in a specific direction of the image by a threshold value or more, the filtering unit 13 excludes the selected feature point.

The environmental map generation unit 14 identifies pairs of corresponding feature points between frames, calculates the three-dimensional coordinates of the feature points identified as pairs, and uses the calculation results to generate an environmental map composed of a collection of feature points identified as pairs.

In this manner, in the embodiment, the map generation device 10 selects a specific feature point, for example, a feature point that is the edge of an object, and excludes the selected feature point if the depth of the selected feature point is greater than the minimum depth value of other feature points aligned horizontally by a certain amount. In this case, the selected feature point is a feature point created by the shadow of an obstruction between the sensor and the object. As a result, according to the embodiment, degradation of the accuracy of the environmental map can be suppressed even when an obstruction exists between the sensor and the object.

Next, the configuration and functions of the map generation device in the embodiment will be specifically explained using Figures 2 and 3. Figure 2 is a block diagram specifically showing the configuration of the map generation device. Figure 3 is an explanatory diagram for explaining the functions of the filtering unit.

As shown in FIG. 2, in this embodiment, the map generating device 10 is mounted on an autonomously moving vehicle 100 such as a robot. In addition to the map generating device 10, the vehicle 100 is equipped with a depth sensor 20, a control device 30, a steering device 40, and a powertrain 50.

In this embodiment, the depth sensor 20 is attached to the moving body 100 so as to capture images in the direction of travel. The depth sensor 20 is a sensor capable of capturing depth-based images and outputs depth-based image data at a set frame rate. In the example of Figure 2, a 3D LiDAR is used as the depth sensor 20.

The control device 30 is constructed by a computer installed on the mobile body 100. The control device 30 controls the direction of travel and speed of travel of the mobile body 100 using the environmental map generated by the map generation device 10. For example, the control device 30 uses the environmental map to set a route from the present to the destination, determines the direction of travel and speed of travel of the mobile body 100 so that the mobile body 100 moves along the set route, and controls the powertrain 50 and steering device 40.

The powertrain 50 is composed of an electric motor for driving, a power transmission mechanism, etc. Tires, caterpillar tracks, etc. are connected to the powertrain 50. The powertrain 50 rotates the tires, caterpillar tracks, etc. in accordance with instructions from the control device 30.

The steering device 40 is equipped with a mechanism for controlling the direction of the steering wheels of the mobile body 100. The steering device 40 determines the direction of the steering wheels in response to instructions from the control device 30. The steering device 40 may also be equipped with a mechanism for controlling the direction of movement by controlling the torque of the left and right drive wheels.

In the embodiment, the map generating device 10 is constructed on a computer mounted on the mobile body 100 by a program described in the embodiment below. The map generating device 10 may also be constructed by a device (e.g., an electronic circuit) separate from the computer mounted on the mobile body 100.

In this embodiment, the data acquisition unit 11 acquires depth-based image data on a frame-by-frame basis from the depth sensor 20. The data acquisition unit 11 sequentially inputs the acquired image data (frames) to the feature point extraction unit 12.

In this embodiment, when a frame is input, the feature point extraction unit 12 extracts feature points of an object for each frame using, for example, the commonly used FAST algorithm. The feature point extraction unit 12 then inputs information identifying the feature points extracted for each frame to the filtering unit 13.

In an embodiment, when feature points for each frame are input, the filtering unit 13 selects a specific feature point from the extracted feature points for each frame. In an embodiment, the filtering unit 13 calculates the curvature of a trajectory connecting a feature point of interest and multiple other feature points that are aligned with the feature point of interest in a specific direction in the image, and selects the feature point of interest as a specific feature point if the calculated curvature is equal to or greater than a threshold. In other words, as shown in Figure 3, the filtering unit 13 selects feature points on the edges of an object as specific feature points (hereinafter referred to as "edge feature points").

In the example of Figure 3, feature points indicated by "●" are selected. Feature points indicated by "●" are feature points on the edge of the object (edge feature points), where the curvature is above a threshold. Feature points indicated by "○" are feature points on the flat surface of the object (hereinafter referred to as "plane feature points"). In Figure 3, 21 indicates the imaging area of the depth sensor 20. 22 indicates a wall that exists in real space, and 23 indicates an obstruction that exists between the depth sensor 20 and the wall (object) 22.

Specifically, the specific direction is the scanning direction of the depth sensor 20 near the feature point of interest. If the depth sensor 20 is a LiDAR, the scanning direction is determined by the LiDAR specifications. The filtering unit 13 first focuses on one of the extracted feature points and identifies multiple other feature points aligned with this feature point (hereinafter referred to as the "feature point of interest") in the scanning direction. For example, the filtering unit 13 identifies five other feature points before and after the feature point of interest (a total of 10) in the scanning direction.

Next, the filtering unit 13 calculates the curvature c of a trajectory connecting the feature point of interest and a plurality of other feature points, for example, using the following equation 1. In the following equation 1, i indicates the feature point of interest, j indicates the other feature points adjacent to the feature point of interest, S indicates the number of the other feature points adjacent to the feature point of interest, ^{X L} _{(k, i)} indicates the coordinates of the feature point of interest i in frame k, and X ^L _{(k, j)} indicates the coordinates of the feature point j in frame k.

The filtering unit 13 selects the feature point of interest as an edge feature point if the calculated value of curvature c is equal to or greater than a threshold (curvature c _i ≧ threshold). The filtering unit 13 can also select an edge feature point only if the number of feature points with curvature c equal to or greater than the threshold within the same frame does not exceed an upper limit. This is because, if the number of feature points exceeds the upper limit, the feature point of interest is likely not a feature point created by a shadow.

Next, for each selected edge feature point, the filtering unit 13 determines whether the depth d _i of the edge feature point is greater than the minimum value among the depths of the other feature points lined up in the scanning direction by a threshold or more. Specifically, the filtering unit 13 determines whether the following equation 2 is satisfied. If the determination result shows that the depth d i is greater than the threshold or more (depth d _i -MIN(d _j )≧threshold), the filtering unit 13 excludes the selected edge feature point.

The environmental map generation unit 14 first uses the feature points of each frame that were not excluded by the filtering unit 13 to match the feature points extracted from each frame with the feature points extracted from previous frames, and identifies pairs of feature points.

Next, for each frame, the environment map generation unit 14 calculates the camera matrix for that frame using the set of feature points identified in that frame.The environment map generation unit 14 then calculates the three-dimensional coordinates of the feature points using the calculated camera matrix and the two-dimensional coordinates of the feature points in the frame.

Next, after calculating the camera matrix, the environmental map generation unit 14 calculates the camera position from the camera matrix and performs self-location estimation. Furthermore, the mobile object generates or updates an environmental map composed of a three-dimensional point cloud (hereinafter referred to as a "three-dimensional environmental map") using the feature points whose three-dimensional coordinates have been calculated. Figure 4 shows an example of a three-dimensional environmental map.

The environmental map generation unit 14 also inputs the generated three-dimensional environmental map to the control device 30. Furthermore, the environmental map generation unit 14 can also generate a two-dimensional environmental map by converting the three-dimensional environmental map into two dimensions. In this case, the environmental map generation unit 14 passes the generated two-dimensional environmental map to the control device 30. The control device 30 then uses the obtained environmental map to control the direction of travel and movement speed of the moving object.

In the above example, a 3D LiDAR capable of measuring depth in a 3D space is used as the depth sensor 20, but in the embodiment, the depth sensor 20 is not limited to this. In the embodiment, other examples of the depth sensor 20 include a 2D LiDAR with a fixed angle of view that measures depth in a specific plane, and a TOF (Time Of Flight) camera. Furthermore, the LiDAR used as the depth sensor 20 may be a fixed type with a fixed angle of view, or a rotating type with an open angle of view.
[Device operation]
Next, the operation of the map generation device 10 in the embodiment will be described with reference to FIG. 5. FIG. 5 is a flow diagram showing the operation of the map generation device. In the following description, reference will be made to FIGS. 1 to 4 as appropriate. In addition, in the embodiment, a map generation method is implemented by operating the map generation device 10. Therefore, the description of the map generation method in the embodiment will be replaced by the following description of the operation of the map generation device 10.

As shown in Figure 5, first, when depth-added image data is output from the depth sensor 20 on a frame-by-frame basis, the data acquisition unit 11 acquires the output image data (step A1). Furthermore, when the data acquisition unit 11 acquires image data for the set number of frames, it inputs the acquired image data to the feature point extraction unit 12.

Next, when the image data acquired in step A1 is input, the feature point extraction unit 12 extracts feature points for each frame (step A2). The feature point extraction unit 12 also outputs the extracted feature points for each frame to the filtering unit 13.

Next, when the feature points for each frame are input in step A2, the filtering unit 13 selects edge feature points from the extracted feature points for each frame (step A3).

In step A3, the filtering 13 calculates, for each frame, the curvature of the trajectory connecting the feature point of interest and multiple other feature points that are aligned with the feature point of interest in the scanning direction of the depth sensor 20, and if the calculated curvature is greater than or equal to a threshold value, selects the feature point of interest as an edge feature point.

Next, for each edge feature point selected in step A3, the filtering unit 13 determines whether the depth of the edge feature point is greater than the minimum depth of multiple other feature points aligned with the edge feature point in the scanning direction by a threshold value or more (step A4).

Next, the filtering unit 13 eliminates, for each frame, edge feature points that are determined to be greater than or equal to the threshold value in step A4 (step A5).

Next, the environmental map generation unit 14 uses the feature points that were not excluded in step A5 to match, for each frame, the feature points extracted from that frame with the feature points extracted from previous frames, and identifies pairs of feature points (step A6).

Next, the environmental map generation unit 14 calculates the three-dimensional coordinates of the feature points identified as pairs in step A6, and uses the calculation results to generate an environmental map consisting of a collection of feature points identified as pairs (step A7).

Then, the environmental map generation unit 14 inputs the environmental map generated in step A7 to the control device 30. As a result, the control device 30 uses the obtained environmental map to control the direction of travel and the speed of movement of the moving object.

As described above, in the embodiment, feature points caused by the shadow of an obstruction between the depth sensor and the object are eliminated, thereby avoiding the extraction of an incorrect set of feature points. Therefore, according to the embodiment, even if an obstruction exists between the depth sensor and the object, a decrease in the accuracy of the environmental map can be suppressed.

[program]
The program in the embodiment may be any program that causes a computer to execute steps A1 to A7 shown in Figure 5. By installing and executing this program on a computer, the map generation device 10 and map generation method in the embodiment can be realized. In this case, the processor of the computer functions as a data acquisition unit 11, a feature point extraction unit 12, a filtering unit 13, and an environmental map generation unit 14 and performs processing.

Another example of a computer is a computer installed on the mobile object 100. Other examples of computers include general-purpose PCs, smartphones, and tablet terminal devices.

The program in the embodiment may also be executed by a computer system constructed by multiple computers. In this case, for example, each computer may function as one of the data acquisition unit 11, feature point extraction unit 12, filtering unit 13, and environmental map generation unit 14.

[Physical configuration]
A computer that realizes the map generating device 10 by executing a program according to the embodiment will now be described with reference to Fig. 6. Fig. 6 is a block diagram showing an example of a computer that realizes the map generating device.

As shown in FIG. 7, the computer 110 comprises a CPU (Central Processing Unit) 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. Each of these components is connected to each other via a bus 121 so that they can communicate data with each other.

Furthermore, the computer 110 may be equipped with a GPU (Graphics Processing Unit) or an FPGA (Field-Programmable Gate Array) in addition to or instead of the CPU 111. In this embodiment, the GPU or FPGA can execute the programs in the embodiments.

The CPU 111 deploys the program in this embodiment, which is composed of a group of codes stored in the storage device 113, into the main memory 112 and executes each code in a predetermined order to perform various calculations. The main memory 112 is typically a volatile storage device such as a DRAM (Dynamic Random Access Memory).

In addition, the program in this embodiment is provided in a state stored on a computer-readable recording medium 120. The program in this embodiment may also be distributed over the Internet connected via the communication interface 117.

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 120, reads programs from the recording medium 120, and writes the processing results of the computer 110 to the recording medium 120. The communication interface 117 mediates data transmission between the CPU 111 and other computers.

Specific examples of recording medium 120 include general-purpose semiconductor storage devices such as CF (Compact Flash (registered trademark)) and SD (Secure Digital), magnetic recording media such as flexible disks, or optical recording media such as CD-ROMs (Compact Disk Read Only Memory).

Note that the map generation device 10 in the embodiment can be realized not by a computer with a program installed, but by hardware corresponding to each part, such as electronic circuits. Furthermore, the map generation device 10 may be realized in part by a program and the remaining part by hardware. In the embodiment, the computer is not limited to the computer shown in FIG. 7.

Some or all of the above-described embodiments can be expressed by (Appendix 1) to (Appendix 12) described below, but are not limited to the following descriptions.

(Appendix 1)
a data acquisition unit that acquires image data output from the depth sensor on a frame-by-frame basis;
a feature point extraction unit that extracts feature points from the image data for each frame;
a filtering unit that selects a specific feature point from the extracted feature points for each frame, and excludes the selected specific feature point when the depth of the selected specific feature point is greater than the minimum value of depths of a plurality of other feature points that are aligned with the selected specific feature point in a specific direction of the image by a threshold value or more;
an environment map generation unit that identifies pairs of corresponding feature points between frames, calculates three-dimensional coordinates of the feature points identified as pairs, and generates an environment map made up of a collection of the feature points identified as pairs using the calculation results;
Equipped with
A map generating device characterized by:

(Appendix 2)
the filtering unit calculates a curvature of a locus connecting a feature point of interest and a plurality of other feature points aligned with the feature point of interest in the specific direction, and selects the feature point of interest as the specific feature point when the calculated curvature is equal to or greater than a threshold;
excluding the selected specific feature point when the depth of the selected specific feature point is greater than the minimum value of the depths of the other plurality of feature points by a threshold or more;
2. The map generating device of claim 1.

(Appendix 3)
the specific direction is a scanning direction of the depth sensor,
the filtering unit calculates a curvature of a locus connecting the feature point of interest and a plurality of other feature points aligned with the feature point of interest in the specific direction, with the feature point of interest as a center;
3. The map generating device of claim 2.

(Appendix 4)
The depth sensor is mounted on a moving body.
4. A map generating device according to any one of claims 1 to 3.

(Appendix 5)
a data acquisition step of acquiring image data output from the depth sensor on a frame-by-frame basis;
a feature point extraction step of extracting feature points from the image data for each frame;
a filtering step of selecting a specific feature point from the extracted feature points for each frame, and excluding the selected specific feature point when the depth of the selected specific feature point is greater than the minimum value of the depths of a plurality of other feature points that are aligned with the selected specific feature point in a specific direction of the image by a threshold value or more;
an environmental map generation step of identifying pairs of corresponding feature points between frames, calculating three-dimensional coordinates of the identified pairs of feature points, and using the calculation results to generate an environmental map configured from a collection of the identified pairs of feature points;
having
A map generating method comprising:

(Appendix 6)
in the filtering step, a curvature of a locus connecting a feature point of interest and a plurality of other feature points aligned with the feature point of interest in the specific direction is calculated, and if the calculated curvature is equal to or greater than a threshold, the feature point of interest is selected as the specific feature point;
excluding the selected specific feature point when the depth of the selected specific feature point is greater than the minimum value of the depths of the other plurality of feature points by a threshold or more;
6. The map generation method of claim 5.

(Appendix 7)
the specific direction is a scanning direction of the depth sensor,
In the filtering step, a curvature of a locus connecting the feature point of interest and a plurality of other feature points aligned with the feature point of interest in the specific direction with the feature point of interest as a center is calculated.
7. The map generation method of claim 6.

(Appendix 8)
The depth sensor is mounted on a moving body.
8. A map generation method according to any one of appendices 5 to 7.

(Appendix 9)
On the computer,
a data acquisition step of acquiring image data output from the depth sensor on a frame-by-frame basis;
a feature point extraction step of extracting feature points from the image data for each frame;
a filtering step of selecting a specific feature point from the extracted feature points for each frame, and excluding the selected specific feature point when the depth of the selected specific feature point is greater than the minimum value of the depths of a plurality of other feature points that are aligned with the selected specific feature point in a specific direction of the image by a threshold value or more;
an environmental map generation step of identifying pairs of corresponding feature points between frames, calculating three-dimensional coordinates of the identified pairs of feature points, and using the calculation results to generate an environmental map configured from a collection of the identified pairs of feature points;
A program that executes.

(Appendix 10)
in the filtering step, a curvature of a locus connecting a feature point of interest and a plurality of other feature points aligned with the feature point of interest in the specific direction is calculated, and if the calculated curvature is equal to or greater than a threshold, the feature point of interest is selected as the specific feature point;
excluding the selected specific feature point when the depth of the selected specific feature point is greater than the minimum value of the depths of the other plurality of feature points by a threshold or more;
10. The program according to claim 9.

(Appendix 11)
the specific direction is a scanning direction of the depth sensor,
In the filtering step, a curvature of a locus connecting the feature point of interest and a plurality of other feature points aligned with the feature point of interest in the specific direction with the feature point of interest as a center is calculated.
11. The program according to claim 10.

(Appendix 12)
The depth sensor is mounted on a moving body.
12. The program according to any one of appendices 9 to 11.

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

This application claims priority based on Japanese Patent Application No. 2022-91129, filed on June 3, 2022, the disclosure of which is incorporated herein in its entirety.

As described above, according to the present disclosure, it is possible to suppress a decrease in the accuracy of the environmental map even when an obstruction exists between the sensor and the object. This disclosure is useful in fields where SLAM is used.

REFERENCE SIGNS LIST 10 Map generation device 11 Data acquisition unit 12 Feature point extraction unit 13 Filtering unit 14 Environmental map generation unit 20 Depth sensor 30 Control device 40 Steering device 50 Power train 100 Mobile body 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 120 Recording medium 121 Bus

Claims (12)

a data acquisition unit that acquires image data output from the depth sensor on a frame-by-frame basis;
a feature point extraction unit that extracts feature points from the image data for each frame;
a filtering unit that selects a specific feature point from the extracted feature points for each frame, and excludes the selected specific feature point when the depth of the selected specific feature point is greater than the minimum value of depths of a plurality of other feature points that are aligned with the selected specific feature point in a specific direction of the image by a threshold value or more ;
an environment map generation unit that identifies pairs of corresponding feature points between frames, calculates three-dimensional coordinates of the feature points identified as pairs, and generates an environment map made up of a collection of the feature points identified as pairs using the calculation results;
Equipped with
A map generating device characterized by: the filtering unit calculates a curvature of a locus connecting a feature point of interest and a plurality of other feature points aligned with the feature point of interest in the specific direction, and selects the feature point of interest as the specific feature point when the calculated curvature is equal to or greater than a threshold;
excluding the selected specific feature point when the depth of the selected specific feature point is greater than the minimum value of the depths of the other plurality of feature points by a threshold or more;
The map generating device according to claim 1 . the specific direction is a scanning direction of the depth sensor,
the filtering unit calculates a curvature of a locus connecting the feature point of interest and a plurality of other feature points aligned with the feature point of interest in the specific direction, with the feature point of interest as a center;
The map generating device according to claim 2 . The depth sensor is mounted on a moving body.
The map generating device according to claim 1 . The image data output from the depth sensor is acquired frame by frame,
extracting feature points from the image data for each frame;
selecting a specific feature point from the extracted feature points for each frame, and excluding the selected specific feature point when the depth of the selected specific feature point is greater than the minimum value of the depths of a plurality of other feature points that are aligned with the selected specific feature point in a specific direction of the image by a threshold value or more;
Identifying pairs of corresponding feature points between frames, calculating three-dimensional coordinates of the feature points identified as pairs, and generating an environment map composed of the sets of feature points identified as pairs using the calculation results.
A map generating method comprising: in the selection, a curvature of a locus connecting a feature point of interest and a plurality of other feature points aligned with the feature point of interest in the specific direction is calculated, and if the calculated curvature is equal to or greater than a threshold, the feature point of interest is selected as the specific feature point;
excluding the selected specific feature point when the depth of the selected specific feature point is greater than the minimum value of the depths of the other plurality of feature points by a threshold or more;
The map generating method according to claim 5 . the specific direction is a scanning direction of the depth sensor,
In the calculation, a curvature of a locus that connects the feature point of interest and a plurality of other feature points that are aligned with the feature point of interest in the specific direction with the feature point of interest as a center is calculated.
The map generating method according to claim 6. The depth sensor is mounted on a moving body.
The map generating method according to claim 5 . On the computer,
The image data output from the depth sensor is acquired frame by frame,
extracting feature points from the image data for each frame;
selecting a specific feature point from the extracted feature points for each frame, and excluding the selected specific feature point when the depth of the selected specific feature point is greater than the minimum value of the depths of a plurality of other feature points that are aligned with the selected specific feature point in a specific direction of the image by a threshold value or more;
Identifying pairs of corresponding feature points between frames, calculating three-dimensional coordinates of the feature points identified as pairs, and generating an environmental map composed of the sets of feature points identified as pairs using the calculation results.
program . in the selection, a curvature of a locus connecting a feature point of interest and a plurality of other feature points aligned with the feature point of interest in the specific direction is calculated, and if the calculated curvature is equal to or greater than a threshold, the feature point of interest is selected as the specific feature point;
excluding the selected specific feature point when the depth of the selected specific feature point is greater than the minimum value of the depths of the other plurality of feature points by a threshold or more;
The program according to claim 9. the specific direction is a scanning direction of the depth sensor,
In the calculation, a curvature of a locus that connects the feature point of interest and a plurality of other feature points that are aligned with the feature point of interest in the specific direction with the feature point of interest as a center is calculated.
The program according to claim 10. The depth sensor is mounted on a moving body.
The program according to claim 9.