Information processing device, information processing method, and program

Abstract

The present technology relates to an information processing device, an information processing method, and a program that make it possible to perform matching of a point cloud representing the shape of an object lacking a distinctive shape. An information processing device according to the present technology comprises: an extraction unit that uses a frequency-modulated continuous wave to extract polarization information from data of a point cloud acquired by a light detection distance measurement unit that measures distances to a plurality of observation points on a surrounding object; and a matching unit that performs matching between point clouds from two viewpoints using the feature amount based on the polarization information. The present technology is applicable to, for example, a sensing system for estimating a self-position and posture using FMCW LiDAR.

Description

Information processing device, information processing method, and program

[0001] The present technology relates to an information processing device, an information processing method, and a program, and in particular to an information processing device, an information processing method, and a program that are capable of matching point clouds that indicate the shape of an object that has little distinctive shape.

[0002] Sensing technologies using cameras and LiDAR (Light Detection and Ranging) are being developed and widely used (see, for example, Patent Document 1).

[0003] Recently, methods have been studied for estimating the self-position and orientation of vehicles using LiDAR. For example, the self-position and orientation can be estimated by matching a point cloud acquired by LiDAR, which indicates the shape of surrounding three-dimensional objects, with previously acquired map information, point clouds acquired in the past, point clouds acquired by other LiDARs, etc.

[0004] International Publication No. 2022 / 195954

[0005] When matching point clouds acquired by LiDAR, it is not possible to distinguish between observation points, so matching is performed based on geometric structure. For example, in environments where there are not enough three-dimensional objects large enough to be captured by LiDAR, such as flat environments, symmetrical environments, environments with repeated identical shapes, or environments with no three-dimensional objects in the surrounding area, the point cloud will have few distinctive shapes, which reduces matching accuracy and makes it difficult to correctly estimate the vehicle's position and orientation.

[0006] For example, in cases where an inspection vehicle moves parallel to the wall of a corridor or tunnel for inspection purposes, it was difficult to recognize the movement of the inspection vehicle using LiDAR.

[0007] The present technology has been developed in light of such circumstances, and makes it possible to perform accurate matching of point clouds that indicate the shapes of objects that lack distinctive shapes.

[0008] An information processing device according to one aspect of the present technology includes an extraction unit that extracts polarization information from point cloud data acquired by a light detection and ranging unit that measures distances to multiple observation points on surrounding objects using frequency-modulated continuous waves, and a matching unit that performs matching between the point clouds from two viewpoints using features based on the polarization information.

[0009] In an information processing method according to one aspect of the present technology, an information processing device extracts polarization information from point cloud data acquired by a light detection and ranging unit that uses frequency-modulated continuous waves to measure distances to multiple observation points on surrounding objects, and matches the point clouds from two viewpoints using features based on the polarization information.

[0010] A program according to one aspect of the present technology causes a computer to extract polarization information from point cloud data acquired by a light detection and ranging unit that uses frequency-modulated continuous waves to measure distances to multiple observation points on surrounding objects, and to execute a process of matching the point clouds from two viewpoints using features based on the polarization information.

[0011] In one aspect of the present technology, polarization information is extracted from point cloud data acquired by an optical detection and ranging unit that uses frequency-modulated continuous waves to measure distances to multiple observation points on surrounding objects, and matching is performed between the point clouds from two viewpoints using features based on the polarization information.

[0012] 12. It is a figure explaining matching between photographed images. It is a figure explaining feature amounts used for matching of feature points. It is a figure explaining matching between point clouds. It is a figure explaining an example of a point cloud when LiDAR is rotated. It is a figure explaining speed information measured by FMCW LiDAR. It is a figure explaining example of polarization information measured by FMCW LiDAR. It is a figure showing another example of polarization information measured by FMCW LiDAR. It is a figure explaining an example of a point cloud when FMCW LiDAR is rotated. It is a block diagram showing an example configuration of a sensing system according to an embodiment of the present technology. It is a flowchart explaining processing performed by a signal processing unit having the configuration of FIG. 9. It is a first figure explaining an example of feature points. It is a block diagram showing an example configuration of a signal processing unit when a numerical value of the degree of polarization is used as a feature amount. It is a flowchart explaining processing performed by a signal processing unit having the configuration of FIG. 12. It is a second figure explaining an example of feature points. It is a block diagram showing an example configuration of a signal processing unit when a shape of a boundary of a same material region is used as a feature amount. It is a flowchart explaining processing performed by a signal processing unit having the configuration of FIG. 15. It is a third figure explaining an example of feature points. It is a block diagram showing an example configuration of a signal processing unit when cluster information is used as a feature amount. It is a flowchart explaining processing performed by a signal processing unit having the configuration of FIG. Fig. 22 is a diagram illustrating a method for matching point-to-face pairs. Fig. 23 is a block diagram illustrating an example configuration of a signal processing unit when normal information is used as a feature. Fig. 24 is a flowchart illustrating processing performed by a signal processing unit having the configuration of Fig. 21. Fig. 25 is a diagram illustrating an application example of a sensing system of the present technology. Fig. 26 is a block diagram illustrating an example configuration of computer hardware.

[0013] Hereinafter, embodiments of the present technology will be described in the following order: 1. Description of matching between images and matching between point clouds 2. Configuration and operation of sensing system 3. Application example

[0014] 1. Description of Matching Between Images and Matching Between Point Clouds Matching Between Captured Images FIG. 1 is a diagram for explaining matching between captured images.

[0015] As shown in FIG. 1, suppose a camera photographs three-dimensional objects Obj1 and Obj2 to obtain a photographed image Pic1, and then changes the viewpoint (position and orientation) to photograph the three-dimensional objects Obj1 and Obj2 to obtain a photographed image Pic2.

[0016] When estimating the amount of movement of the camera using the captured images Pic1 and Pic2, first, feature points are extracted from each of the captured images Pic1 and Pic2. In the example of Fig. 1, the feature points are indicated by black dots.

[0017] Next, feature points are matched between the photographed images Pic1 and Pic2, and corresponding feature points are searched for between the photographed images Pic1 and Pic2.

[0018] Finally, the feature points in the captured images Pic1 and Pic2 are projected onto three-dimensional space, and the amount of movement of the camera is optimized so that the distance in three-dimensional space between a feature point pair consisting of a feature point in the captured image Pic1 and a feature point in the captured image Pic2 that corresponds to that feature point is minimized.

[0019] FIG. 2 is a diagram illustrating feature amounts used in matching feature points.

[0020] Corresponding feature points between the captured images Pic1 and Pic2 are searched for, for example, based on the feature amounts of the feature points, which are calculated, for example, based on the texture around the feature points.

[0021] Assume that four feature points P1 to P4 are extracted from the captured image Pic1 as shown on the left side of Fig. 2. In this case, for example, based on the pixel values ​​of each pixel in an area A4 surrounding the feature point P4, a multi-dimensional sequence is calculated as the feature amounts of the feature points, as shown by the tips of the white arrows in Fig. 2.

[0022] For example, feature points having substantially the same sequence of features are associated as a feature point pair.

[0023] Matching Between Point Clouds FIG. 3 is a diagram for explaining matching between point clouds.

[0024] As shown in Figure 3, the LiDAR 1A measures the distance to the three-dimensional objects Obj1 and Obj2 based on a light reception signal obtained by receiving reflected light of laser light irradiated on the three-dimensional objects Obj1 and Obj2. The LiDAR 1A performs distance measurement at regular angles within the scanning range of the laser light, and acquires a point cloud representing the shapes of the three-dimensional objects Obj1 and Obj2 based on the angle and distance information. The point cloud represents the three-dimensional position of the three-dimensional object as a collection of points on the three-dimensional object (observation points irradiated with laser light). In the example of Figure 3, the observation points are indicated by black dots.

[0025] Assume that the LiDAR 1A acquires a point cloud by measuring the distances to a plurality of observation points on the three-dimensional objects Obj1 and Obj2 from a certain viewpoint, and then acquires a point cloud again by changing the viewpoint as indicated by the arrow in FIG.

[0026] When matching between point clouds acquired by the LiDAR 1A is performed, since it is not possible to distinguish between observation points, matching is performed based on the geometric structure. For example, a point cloud representing the shape of a three-dimensional object Obj1 is searched for from point clouds from two viewpoints, and each observation point included in the point cloud representing the shape of the three-dimensional object Obj1 is associated with each other between the point clouds from the two viewpoints.

[0027] The observation points of the point clouds from each of the two viewpoints are projected onto three-dimensional space, and the amount of movement of LiDAR 1A is optimized so that the distance in three-dimensional space between an observation point pair consisting of an observation point in the point cloud from one viewpoint and a corresponding observation point in the point cloud from the other viewpoint is minimized.

[0028] For example, in environments where there are not enough three-dimensional objects large enough to be captured by LiDAR 1A, such as flat environments, symmetrical environments, environments where the same shape is repeated, or environments where there are no three-dimensional objects in the surrounding area, there will be a lack of distinctive shapes in the point cloud, which will reduce matching accuracy and make it difficult to correctly estimate the self-position and orientation.

[0029] For example, in cases where an inspection vehicle moves parallel to the wall of a corridor, tunnel, etc. for inspection purposes, it was difficult to recognize the movement of the inspection vehicle using the point cloud acquired by LiDAR 1A.

[0030] Furthermore, if the LiDAR 1A rotates without changing the direction it is facing, it may be difficult to recognize the rotation using the point cloud acquired by the LiDAR 1A. For example, as shown in the upper left of Figure 4, assume that the LiDAR 1A acquires a point cloud by measuring the distances of multiple observation points on a planar three-dimensional object Obj11 from a certain viewpoint. In this case, as shown at the tip of the white arrow #1, it appears to the LiDAR 1A that observation points P11 to P13 are arranged in the upper row from left to right in the figure, observation points P14 to P16 are arranged in the middle row from left to right, and observation points P17 to P19 are arranged in the lower row from left to right.

[0031] Now, suppose that LiDAR 1A is rotated 90° to the right without changing its facing direction, as shown in the upper right of Fig. 4. In this case, as shown by the tip of outline arrow #2, to LiDAR 1A, it appears that observation points P13, P16, and P19 are arranged in the upper row from left to right in the figure, observation points P12, P15, and P18 are arranged in the middle row from left to right, and observation points P11, P14, and P17 are arranged in the lower row from left to right.

[0032] However, since it is not possible to distinguish between each observation point in the point cloud acquired before rotation and the point cloud acquired after rotation, it is difficult to detect that LiDAR1A has rotated by matching the observation points between the point cloud acquired before rotation and the point cloud acquired after rotation.

[0033] - About FMCW LiDAR Here, we will explain FMCW (Frequency Modulated Continuous Wave) LiDAR as one of the light detection and ranging methods using LiDAR. FMCW LiDAR uses chirp light as the laser light to be emitted, where the pulse frequency changes linearly over time, for example. In FMCW LiDAR, ranging is performed by coherent detection of the received signal that combines the laser light emitted as chirp light and the reflected light of the laser light.

[0034] Coherent detection is less susceptible to interference from other light sources, making it possible to avoid crosstalk, and is also less susceptible to noise from high-intensity light sources such as sunlight. As a result, measurements using FMCW LiDAR are robust against environmental changes and can stably measure the shape of three-dimensional objects even in low-light environments.

[0035] FMCW LiDAR can measure speed and polarization information simultaneously with distance by utilizing the Doppler effect.

[0036] FIG. 5 is a diagram illustrating velocity information measured by the FMCW LiDAR.

[0037] As shown in A of Figure 5, when the three-dimensional object Obj21 is moving in the direction of vector v1, the FMCW LiDAR 1 measures vector v1' as velocity information. Vector v1' is a movement component obtained by projecting vector v1 onto a line (ray) connecting the FMCW LiDAR 1 and the observation point on the three-dimensional object Obj21. In other words, the FMCW LiDAR 1 measures the velocity of the three-dimensional object (observation point) in the direction of the ray.

[0038] As shown in B of Figure 5, the FMCW LiDAR1 measures velocity information (vectors v11 to v15) of each of multiple observation points P51 to P55 on the three-dimensional object Obj21. The vectors v11 to v15 are movement components in the direction of the light irradiated from the LiDAR1 to each observation point. It is possible to identify the movement of the entire three-dimensional object Obj21 based on the velocity information of the observation points P51 to P55. When the three-dimensional object Obj21 is stationary, it is possible to estimate the movement of the FMCW LiDAR1 based on the velocity information of the observation points P51 to P55.

[0039] Fig. 6 is a diagram showing an example of polarization information measured by an FMCW LiDAR, in which the horizontal axis represents the incident angle of the laser light and the vertical axis represents the degree of polarization (DOP).

[0040] As shown in Figure 6, FMCW LiDAR can measure different degrees of polarization depending on the material and incident angle.

[0041] 7 is a diagram showing another example of polarization information measured by the FMCW LiDAR. In Fig. 6, the horizontal axis represents the incident angle of the laser light, and the vertical axis represents the reflection intensity.

[0042] As shown in Figure 7, FMCW LiDAR can measure different reflection intensities depending on the material and incident angle.

[0043] By setting observation points fairly densely, it is possible to estimate the surface angle of a three-dimensional object, and therefore it is possible to identify polarization information with the effects of the illumination angle canceled out, the material of the three-dimensional object, the normal direction of the observation point, and so on.

[0044] This technology treats polarization information measured by FMCW LiDAR as a feature of each observation point included in the point cloud acquired by FMCW LiDAR, making it possible to match observation points (align point clouds) between point clouds from two viewpoints. For example, the material of each observation point is used as a feature to match observation points between point clouds from two viewpoints.

[0045] For example, suppose that FMCW LiDAR 1 acquires a point cloud by measuring the distances to multiple observation points on a planar three-dimensional object Obj21 from a certain viewpoint, as shown in the upper left of Fig. 8. In this case, as shown at the tip of hollow arrow #11, it appears to FMCW LiDAR 1 that observation points P31 to P33 are arranged in the upper row from left to right in the figure, observation points P34 to P36 are arranged in the middle row from left to right, and observation points P37 to P39 are arranged in the lower row from left to right.

[0046] As shown in the upper right of Fig. 8, assume that the FMCW LiDAR 1 is rotated 90° to the right without changing the direction it is facing. In this case, as indicated by the tip of the white arrow #12, to the FMCW LiDAR 1, it appears that observation points P33, P36, and P39 are arranged in the upper row from left to right in the figure, observation points P32, P35, and P38 are arranged in the middle row from left to right, and observation points P31, P34, and P37 are arranged in the lower row from left to right.

[0047] As shown by the different hatching in the upper left, upper right, lower left, and lower right regions, each region of the three-dimensional object Obj21 is made of a different material. In this case, by treating the material as a feature, it is possible to distinguish each observation point in the point cloud acquired before rotation and the point cloud acquired after rotation. Therefore, by matching the observation points between the point cloud acquired before rotation and the point cloud acquired after rotation, it is possible to detect that the FMCW LiDAR 1 has rotated.

[0048] It should be noted that instead of matching all observation points included in the point cloud acquired by the FMCW LiDAR, matching may be performed on representative points that are representative observation points among the multiple observation points included in the point cloud.

[0049] As described above, in the sensing system of the present technology, polarization information is extracted from point cloud data acquired by a light detection and ranging unit that measures distances to multiple observation points on surrounding objects using frequency-modulated continuous waves, and feature amounts based on the polarization information are used to match point clouds from two viewpoints. Because feature amounts based on the polarization information are used to match point clouds from two viewpoints, it is possible to match point clouds that represent the shape of an environment with few distinctive shapes.

[0050] 2. Configuration and Operation of Sensing System FIG. 9 is a block diagram showing an example configuration of a sensing system according to an embodiment of the present technology.

[0051] The sensing system of FIG. 9 is composed of a sensor unit 11 and a signal processing unit 12 .

[0052] The sensor unit 11 includes a light detection and ranging unit 21. An FMCW LiDAR that measures distances using frequency-modulated continuous waves is used as the light detection and ranging unit 21. Measurement results by the light detection and ranging unit 21 are supplied to the signal processing unit 12 as point cloud data.

[0053] The point cloud data is data of a group of points in which position information, velocity information, and polarization information in three-dimensional space are associated with each observation point on a three-dimensional object.

[0054] The sensor unit 11 may be provided with various sensors such as an RGB sensor and an IMU (Inertial Measurement Unit) in addition to the FMCW LiDAR as the light detection and ranging unit 21. The sensor unit 11 may also be provided with a plurality of FMCW LiDARs.

[0055] The signal processing unit 12 includes a position information extraction unit 31 , a polarization information extraction unit 32 , a feature amount acquisition unit 33 , a matching unit 34 , and a control unit 35 .

[0056] The position information extraction unit 31 extracts position information of a representative point from among the observation points included in the point cloud indicated by the point cloud data supplied from the sensor unit 11 and supplies the extracted information to the matching unit 34 .

[0057] The polarization information extraction unit 32 extracts representative polarization information from among the observation points included in the point cloud indicated by the point cloud data supplied from the sensor unit 11 , and supplies the information to the feature acquisition unit 33 .

[0058] The feature amount acquiring unit 33 acquires the feature amount of each representative point used for matching by the matching unit 34 based on the polarization information supplied from the polarization information extracting unit 32. The feature amount acquiring unit 33 supplies the feature amount acquired based on the polarization information to the matching unit 34 as a polarization feature amount.

[0059] The matching unit 34 matches representative points between point clouds from two viewpoints using the polarization feature of each representative point supplied from the feature acquisition unit 33. The point clouds from two viewpoints to be matched may be a current point cloud and a past point cloud, point clouds acquired by two FMCW LiDARs, or a point cloud acquired by an FMCW LiDAR and a point cloud indicated by map information. Note that the two viewpoints may be different or the same.

[0060] The matching unit 34 estimates the self-position and orientation of, for example, the sensor unit 11 based on the matching result of the representative points, and supplies self-position information indicating the self-position and orientation to the control unit 35. Because the distance between each observation point included in the point cloud and the viewpoint of the FMCW LiDAR (the position and orientation of the sensor unit 11) is known, if the correspondence relationship between the observation points between the point clouds of the two viewpoints is known, it is possible to obtain the relationship between at least one of the positions and orientations of the two viewpoints (hereinafter referred to as the position and orientation relationship). The matching unit 34 can estimate the self-position and orientation of the FMCW LiDAR based on the position and orientation relationship between the two viewpoints.

[0061] When matching is performed between a current point cloud and a past point cloud, for example, the amount of movement from the past FMCW LiDAR viewpoint to the current viewpoint is estimated as self-location information. When matching is performed between point clouds acquired by two FMCW LiDARs, for example, the position and orientation relationship between the viewpoints of the two FMCW LiDARs is estimated as self-location information. When matching is performed between a point cloud acquired by an FMCW LiDAR and a point cloud indicated by map information, for example, the position on the map of the current FMCW LiDAR viewpoint is estimated as self-location information.

[0062] The matching unit 34 uses the position information of each representative point supplied from the position information extraction unit 31 to match the representative points and estimate the self-position and orientation.

[0063] The control unit 35 performs various controls by applying the self-position information supplied from the matching unit 34. For example, the control unit 35 outputs a control signal for controlling a vehicle or the like based on the self-position information. Furthermore, for example, the control unit 35 renders an image to be displayed on a display based on the self-position information and outputs the rendering result. Furthermore, for example, the control unit 35 projects point clouds of multiple viewpoints onto a three-dimensional space based on the self-position information, creates a 3D model of a three-dimensional object (modeling), and outputs the modeling result.

[0064] Next, the processing performed by the signal processing unit 12 having the above configuration will be described with reference to the flowchart of FIG.

[0065] In step S1, the signal processing unit 12 acquires point cloud data obtained by the light detection and ranging unit 21 scanning within a predetermined scanning range and measuring the position information, velocity information, and polarization information of each of multiple observation points on surrounding three-dimensional objects.

[0066] In step S2 , the position information extraction unit 31 extracts the position information of each representative point (observation point) from the point cloud data acquired by the light detection and ranging unit 21 .

[0067] In step S3 , the polarization information extraction unit 32 extracts polarization information of each representative point (observation point) from the point cloud data acquired by the light detection and ranging unit 21 .

[0068] In step S4, the signal processing unit 12 determines whether or not the position information and polarization information have been acquired. If it is determined in step S4 that the position information and polarization information have not been acquired, the process ends.

[0069] On the other hand, if it is determined in step S4 that the position information and polarization information have been acquired, the feature acquisition unit 33 performs a feature acquisition process in step S5. The feature acquisition process acquires polarization features of each representative point.

[0070] In step S6, the matching unit 34 performs matching between the point groups from the two viewpoints using the polarization feature amount, and estimates the self-position information of the sensor unit 11 based on the matching result.

[0071] In step S7, the signal processing unit 12 determines whether or not the self-location information has been estimated. If it is determined in step S7 that the self-location information has not been estimated, the process ends.

[0072] On the other hand, if it is determined in step S7 that the self-position has been estimated, in step S8, the control unit 35 performs various controls using the self-position information.

[0073] The above process allows matching between point clouds acquired in an environment with few distinctive shapes. Therefore, even in environments where it was previously difficult to estimate self-position and orientation, it is now possible to estimate self-position and orientation using FMCW LiDAR, enabling users to use applications that utilize self-position and orientation in a variety of environments.

[0074] Next, the information used as the feature amount will be described in detail with reference to FIGS.

[0075] Example of Using the Numerical Value of the Degree of Polarization as a Feature Amount FIG. 11 is a first diagram illustrating an example of a feature point.

[0076] 11 , in this technology, a sequence including numerical values ​​of the degrees of polarization of the nine representative points P31 to P39 themselves and the surrounding observation points is used as the feature amount of the representative points P31 to P39. This allows the matching unit 34 to match the representative points by treating them in the same way as feature points in an image captured by a camera. Therefore, it is possible to match between point groups by utilizing an existing feature point matching algorithm.

[0077] When matching is performed for all observation points included in the point cloud, the numerical value of the degree of polarization of each observation point is used as a feature amount as is.

[0078] Fig. 12 is a block diagram showing an example of the configuration of the signal processing unit 12 when the numerical value of the degree of polarization is used as a feature amount. In Fig. 12, the same components as those in Fig. 9 are denoted by the same reference numerals. Duplicate explanations will be omitted as appropriate.

[0079] The signal processing unit 12 in FIG. 12 differs from the signal processing unit 12 in FIG. 9 in that a feature amount adding unit 51 is provided within the feature amount acquiring unit 33 .

[0080] The polarization information extraction unit 32 extracts polarization information of each observation point included in the point cloud indicated by the point cloud data supplied from the light detection and ranging unit 21 and supplies it to the feature amount assignment unit 51 .

[0081] The feature assigning unit 51 acquires a sequence containing numerical values ​​of the degree of polarization between the representative point itself and the observation points around the representative point based on the polarization information supplied from the polarization information extracting unit 32, and assigns this sequence to each representative point as a polarization feature.

[0082] The processing performed by the signal processing unit 12 having the configuration shown in FIG. 12 will be described with reference to the flowchart shown in FIG.

[0083] The processing in steps S21 to S24 is the same as the processing in steps S1 to S4 in FIG. 10, and therefore a description thereof will be omitted.

[0084] In step S25, the feature amount assigning unit 51 assigns a sequence including the numerical values ​​of the degree of polarization indicated by the polarization information to each representative point as a polarization feature amount.

[0085] In step S26, the matching unit 34 performs matching between the point clouds from the two viewpoints using the polarization feature amount. For example, the matching unit 34 searches for a representative point pair consisting of representative points having approximately the same polarization feature amount between the point clouds from the two viewpoints. The matching unit 34 estimates the self-position information of the sensor unit 11 based on the matching result.

[0086] The processing in steps S27 and S28 is similar to the processing in steps S7 and S8 in FIG. 10, and therefore a description thereof will be omitted.

[0087] As described above, the numerical values ​​(sequence) of the degree of polarization indicated by the polarization information can be used as a feature amount.

[0088] Example of using boundary information of regions with the same material as a feature amount FIG. 14 is a second diagram illustrating an example of feature points.

[0089] As shown in the upper left of Figure 14, by measuring the polarization information of dense observation points on the three-dimensional object Obj31, FMCW LiDAR1 can identify observation points that are adjacent to other observation points made of different materials, as shown at the tip of the white arrow #21.

[0090] Based on observation points adjacent to other observation points of different materials, the boundaries of same-material regions, which are regions made up of a single material, can be identified from the entire three-dimensional object Obj31, as shown by the tip of the white arrow #22. For example, the boundaries of same-material regions are identified by connecting observation points adjacent to other observation points of different materials with lines. Also, for example, a line passing between observation points of different materials is identified as the boundary of a same-material region. In the example of Figure 14, the boundaries of five same-material regions A21 to A51 are identified.

[0091] As shown in the lower left of Figure 14, even when the FMCW LiDAR 1 is tilted to the right, by measuring polarization information from dense observation points on the three-dimensional object Obj 31, it is possible to identify observation points that are adjacent to other observation points made of different materials, as indicated by the tip of the white arrow #31. Because the FMCW LiDAR 1 is tilted to the right, the three-dimensional object Obj 31 appears to the FMCW LiDAR 1 to be tilted to the left.

[0092] As in the case described above, based on an observation point adjacent to another observation point of a different material, the boundaries of the same material areas A21 to A51 within the entire three-dimensional object Obj31 can be identified, as shown at the tip of the white arrow #32.

[0093] A rough change in the viewpoint of FMCW LiDAR1 is estimated based on the change in the boundaries of the same material areas A21 to A51 between the point cloud of the viewpoint before FMCW LiDAR1 is tilted and the point cloud of the viewpoint after FMCW LiDAR1 is tilted.

[0094] In this technology, the shape of the boundary of the same-material region is used as a feature of the entire point cloud. The matching unit 34 can estimate the position and orientation relationship (change) between two viewpoints based on the shape of the boundary of each same-material region.

[0095] Fig. 15 is a block diagram showing an example of the configuration of the signal processing unit 12 when the shape of the boundary of the same material region is used as a feature. In Fig. 15, the same components as those in Fig. 9 are denoted by the same reference numerals. Duplicate explanations will be omitted where appropriate.

[0096] The signal processing unit 12 in Figure 15 differs from the signal processing unit 12 in Figure 9 in that a material information assignment unit 71, a boundary extraction unit 72, and a boundary information storage unit 73 are provided within the feature acquisition unit 33, and in that a boundary information comparison unit 81 and a point cloud matching unit 82 are provided within the matching unit 34.

[0097] The polarization information extraction unit 32 extracts polarization information of each observation point included in the point cloud indicated by the point cloud data supplied from the light detection and ranging unit 21 and supplies it to the material information addition unit 71 .

[0098] The material information assigning unit 71 identifies the material of each observation point based on the polarization information supplied from the polarization information extracting unit 32, and assigns material information to each observation point. The material information assigning unit 71 supplies the point cloud data to which the material information has been assigned to the boundary extracting unit 72.

[0099] Boundary extraction unit 72 extracts boundaries of same-material regions based on the material information assigned to each observation point by material information assignment unit 71. For example, boundary extraction unit 72 extracts observation points adjacent to other observation points of different materials, and extracts boundaries of same-material regions based on these observation points.

[0100] The boundary extraction unit 72 supplies boundary information indicating the boundaries of the extracted same-material regions to a boundary information storage unit 73 and a boundary information comparison unit 81 .

[0101] The boundary information storage unit 73 stores the boundary information supplied from the boundary extraction unit 72 and the self-position information supplied from the matching unit 34 .

[0102] The boundary information comparison unit 81 acquires boundary information corresponding to point cloud data acquired at a past time and self-position information from the boundary information storage unit 73. The boundary information comparison unit 81 compares the past boundary information with the current boundary information supplied from the boundary extraction unit 72, and estimates a rough amount of change in the self-position and orientation of the light detection and ranging unit 21.

[0103] The boundary information comparison unit 81 supplies the estimated rough amount of change of the light detection and ranging unit 21 to the point cloud matching unit 82 as auxiliary information for matching between point clouds. In addition, the boundary information comparison unit 81 supplies the point cloud matching unit 82 with past self-position information acquired from the boundary information storage unit 73.

[0104] The point cloud matching unit 82 matches the point cloud of the current viewpoint with the point cloud of the past viewpoint based on the auxiliary information supplied from the boundary information comparison unit 81. The point cloud matching unit 82 estimates the precise self-position and orientation of the light detection and ranging unit 21 based on the matching result, and supplies the self-position information to the control unit 35. The point cloud matching unit 82 uses the position information supplied from the position information extraction unit 31 and the past self-position information supplied from the boundary information comparison unit 81 to match between point clouds and estimate the self-position and orientation.

[0105] In other words, the boundary information comparison unit 81 estimates the position and orientation relationship between the two viewpoints based on the boundary information, and the point cloud matching unit 82 matches the point clouds of the two viewpoints based on the estimation result of the position and orientation relationship between the two viewpoints by the boundary information comparison unit 81.

[0106] The processing performed by the signal processing unit 12 having the configuration shown in FIG. 15 will be described with reference to the flowchart shown in FIG.

[0107] The processing in steps S41 to S44 is the same as the processing in steps S1 to S4 in FIG. 10, and therefore a description thereof will be omitted.

[0108] In step S45, the material information assigning unit 71 identifies the material of each observation point based on the polarization information, and assigns material information to each observation point.

[0109] In step S46, the boundary extraction unit 72 extracts the boundary of the same material region based on the material information assigned to each observation point, and acquires boundary information.

[0110] In step S47, the signal processing unit 12 determines whether or not the boundary information has been acquired. If it is determined in step S47 that the boundary information has not been acquired, the process ends.

[0111] On the other hand, if it is determined in step S47 that the boundary information has been acquired, the boundary information comparison unit 81 compares the past boundary information with the current boundary information in step S48 to acquire auxiliary information.

[0112] In step S49, the point cloud matching unit 82 performs matching between the point cloud of the current viewpoint and the point cloud of the past viewpoint based on the auxiliary information, and estimates the precise self-position and orientation of the light detection and ranging unit 21 based on the matching results.

[0113] The processing in steps S50 and S51 is similar to the processing in steps S7 and S8 in FIG. 10, and therefore a description thereof will be omitted.

[0114] As described above, boundary information can be used as a feature.

[0115] Example of Using Cluster Information as Feature Amount FIG. 17 is a third diagram illustrating an example of feature points.

[0116] As shown by the arrow A in Fig. 17, suppose that FMCW LiDAR1 measures the distances to multiple observation points on three-dimensional objects Obj41 and Obj42 from a certain viewpoint to acquire point cloud PC1 (B in Fig. 17), and then changes the viewpoint to acquire point cloud PC2 (B in Fig. 17). Here, three-dimensional objects Obj41 and Obj42 are made of different materials.

[0117] As described above, the material of each observation point is identified based on the polarization information measured by the FMCW LiDAR 1. As shown in B of Fig. 17, the signal processing unit 12 clusters, in each of the point clouds PC1 and PC2, a set of observation points of the material that makes up the three-dimensional object Obj41 as cluster C1, and a set of observation points of the material that makes up the three-dimensional object Obj42 as cluster C2. In this way, observation points of the same material are clustered into one cluster.

[0118] In this technology, cluster information indicating which cluster each observation point is clustered into is used as a feature amount. By matching observation points for each cluster between point clouds from two viewpoints, the matching unit 34 can reduce erroneous matching, such as matching different observation points into an observation point pair.

[0119] Fig. 18 is a block diagram showing an example of the configuration of the signal processing unit 12 when cluster information is used as a feature. In Fig. 18, the same components as those in Fig. 9 are denoted by the same reference numerals. Duplicate explanations will be omitted where appropriate.

[0120] The signal processing unit 12 in Figure 18 differs from the signal processing unit 12 in Figure 9 in that a material information assignment unit 101 and a clustering unit 102 are provided within the feature acquisition unit 33, and in that intra-cluster matching units 111A to 111Z and an integration unit 112 are provided within the matching unit 34.

[0121] The polarization information extraction unit 32 extracts polarization information of each observation point included in the point cloud indicated by the point cloud data supplied from the light detection and ranging unit 21 and supplies it to the material information addition unit 101 .

[0122] The material information assigning unit 101 identifies the material of each observation point based on the polarization information supplied from the polarization information extracting unit 32, and assigns material information to each observation point. The material information assigning unit 101 supplies the point cloud data to which the material information has been assigned to the clustering unit 102.

[0123] Clustering unit 102 clusters observation points of the same material into one cluster based on the material information assigned to each observation point by material information assignment unit 101. Clustering unit 102 supplies the results of clustering the observation points according to the material as cluster information to intra-cluster matching units 111A to 111Z.

[0124] The intra-cluster matching units 111A to 111Z perform matching of representative points within clusters in which representative points of corresponding materials are clustered, between point clouds from two viewpoints. The intra-cluster matching units 111A to 111Z estimate the self-position and orientation of the light detection and ranging unit 21 based on each matching result, and supply the self-position and orientation estimation results to the integrating unit 112. The position information supplied from the position information extraction unit 31 is used for matching between point clouds and estimating the self-position and orientation by the intra-cluster matching units 111A to 111Z.

[0125] The integrating unit 112 integrates the estimation results of the self-position and orientation by the intra-cluster matching units 111A to 111Z to estimate the self-position and orientation of the sensor unit 11. Specifically, the integrating unit 112 calculates the average of the self-position and orientation estimated by the intra-cluster matching units 111A to 111Z, and adds up the self-position and orientation estimated by the intra-cluster matching units 111A to 111Z using weights according to the number of observation points included in each cluster.

[0126] In other words, the intra-cluster matching units 111A to 111Z perform matching between point clouds from two viewpoints for each cluster, and the integration unit 112 integrates the matching results between point clouds from two viewpoints for each cluster performed by the intra-cluster matching units 111A to 111Z.

[0127] The integration unit 112 supplies the control unit 35 with self-position information indicating the self-position and orientation obtained by integrating the results of estimation of the self-position and orientation by the intra-cluster matching units 111A to 111Z.

[0128] The processing performed by the signal processing unit 12 having the configuration shown in FIG. 18 will be described with reference to the flowchart shown in FIG.

[0129] The processing in steps S71 to S74 is the same as the processing in steps S1 to S4 in FIG. 10, and therefore a description thereof will be omitted.

[0130] In step S75, the material information assigning unit 101 identifies the material of each observation point based on the polarization information, and assigns material information to each observation point.

[0131] In step S76, the clustering unit 102 performs clustering of the observation points based on the material information assigned to each observation point, and acquires cluster information.

[0132] In step S77, the signal processing unit 12 determines whether or not the cluster information has been acquired. If it is determined in step S77 that the cluster information has not been acquired, the process ends.

[0133] On the other hand, if it is determined in step S77 that cluster information has been acquired, in step S78, the intra-cluster matching units 111A to 111Z perform matching between point clouds from two viewpoints for each cluster, and estimate the self-position and orientation of the light detection and ranging unit 21 based on the matching results.

[0134] In step S79, the integration unit 112 integrates the results of estimation of the self-position and orientation by the intra-cluster matching units 111A to 111Z, estimates the self-position and orientation of the sensor unit 11, and acquires self-position information.

[0135] The processing in steps S80 and S81 is similar to the processing in steps S7 and S8 in FIG. 10, and therefore a description thereof will be omitted.

[0136] As described above, cluster information can be used as a feature. Because observation points of the same material are paired, it is possible to reduce erroneous matching, such as pairing different observation points. By using accurately matched observation point pairs, it is possible to improve the accuracy of self-localization and orientation estimation.

[0137] - Example of using normal information as a feature In addition to the method of optimizing the distance between observation points (point-to-point), there is also a method of matching between point clouds (point-to-face) that uses the normals of the observation points of the point cloud that is the search source and the observation points of the point cloud that is the search destination. Generally, point-to-point matching can find a solution faster than point-to-point matching, but it is necessary to separately calculate the normal direction of the observation points of the point cloud that is the search destination.

[0138] As described above, the normal direction of each observation point can be identified based on the polarization information measured by the FMCW LiDAR 1. In this technology, normal information indicating the normal direction identified based on the polarization information is used as a feature. By performing point-to-point matching using the normal direction identified based on the polarization information, the signal processing unit 12 can skip the normal direction estimation process, which takes a long time in general point-to-point matching. Therefore, it is possible to significantly reduce the processing amount and processing time.

[0139] FIG. 20 is a diagram for explaining a point-to-point matching technique.

[0140] As shown in FIG. 20, when matching is performed between a search source point cloud PCy and a search destination point cloud PCx, for example, a vector v yx and the normal vector v of the observation point Px x The observation point pairs are optimized so as to minimize the inner product of

[0000] and

[0000] In the example of Fig. 20, the observation points included in the point cloud PCy are indicated by black dots, and the observation points included in the point cloud PCx are indicated by white dots.

[0141] In general, a plane including the observation point Px and the observation points in the vicinity of the observation point Px is obtained, and the normal vector of the plane is the normal vector v x The signal processing unit 12 of the present technology converts a vector indicating the normal direction specified based on the polarization information measured by the FMCW LiDAR 1 into a normal vector v x By using it as an observation point Px を Since it is not necessary to calculate the containing plane and the normal vector of that plane, matching can be performed efficiently.

[0142] Fig. 21 is a block diagram showing an example of the configuration of the signal processing unit 12 when normal vector information is used as a feature amount. In Fig. 21, the same components as those in Fig. 9 are denoted by the same reference numerals. Duplicate explanations will be omitted as appropriate.

[0143] The signal processing unit 12 in FIG. 21 differs from the signal processing unit 12 in FIG. 9 in that a normal calculation unit 131 is provided in the feature amount acquisition unit 33 .

[0144] The normal calculation unit 131 calculates the normal direction of each representative point based on the polarization information supplied from the polarization information extraction unit 32, and assigns normal information indicating the identified normal direction to each representative point as a polarization feature.

[0145] The processing performed by the signal processing unit 12 having the configuration shown in FIG. 21 will be described with reference to the flowchart shown in FIG.

[0146] The processing in steps S101 to S104 is the same as the processing in steps S1 to S4 in FIG. 10, and therefore a description thereof will be omitted.

[0147] In step S105, the normal calculation unit 131 calculates the normal direction of each representative point based on the polarization information, and acquires normal information.

[0148] In step S106, the signal processing unit 12 determines whether or not the normal information has been acquired. If it is determined in step S106 that the normal information has not been acquired, the process ends.

[0149] On the other hand, if it is determined in step S106 that the normal information has been acquired, in step S107 the matching unit 34 performs matching between the point groups of the two viewpoints using the normal information assigned to each representative point.

[0150] The processing in steps S108 and S109 is similar to the processing in steps S7 and S8 in FIG. 10, and therefore a description thereof will be omitted.

[0151] As described above, by using normal information as a feature, it becomes possible to efficiently perform point-to-point matching.

[0152] 3. Application Example FIG. 23 is a diagram showing an application example of the sensing system of the present technology.

[0153] The sensing system of the present technology can be applied to a control system that controls a moving body such as a vehicle, an aircraft, a ship, an autonomous mobile robot, etc. Fig. 23A shows an example in which the FMCW LiDAR 1 is mounted on a vehicle 201 as a moving body. The FMCW LiDAR 1 is disposed on the top or side of the vehicle 201, for example.

[0154] The sensing system of this technology can be used for odometry estimation while driving on public roads, and absolute position estimation on private property such as parking lots. By using FMCW LiDAR1, it is possible to identify differences in materials such as buildings (concrete), roadways (asphalt), crosswalks (glass beads), signs (steel), trees, and leaves, making it possible to accurately estimate the self-position of a moving object.

[0155] The polarization information measured by the FMCW LiDAR 1 may itself be used to control the vehicle. For example, the shape of a three-dimensional object identified based on the polarization information is used to control the vehicle.

[0156] Specifically, by determining the flatness of the ground surface (road) in the direction of travel of a vehicle based on polarization information and controlling the vehicle's output according to that flatness, it becomes possible to realize a vehicle with higher fuel efficiency. For example, the vehicle's control system may increase the vehicle's driving force on roads with high resistance and decrease the driving force on roads with low resistance.

[0157] Furthermore, for example, a control system for a moving body can determine the level of friction on the ground, such as whether the ground surface in the direction of travel is icy, based on polarization information, and can assist the driver in steering operations or improve the ride comfort of passengers based on the level of friction on the ground.

[0158] The speed information measured by the FMCW LiDAR 1 may be used to control the vehicle. Since the speed of objects around the vehicle can be determined based on the speed information, it is possible to improve the performance of safety functions such as obstacle detection. Furthermore, the vehicle's control system can identify the type of object around the vehicle based on the polarization information and perform hazard prediction based on the type of object.

[0159] The sensing system of the present technology can be applied to, for example, a filming system used in virtual production. B of Fig. 23 shows an example in which the FMCW LiDAR 1 is mounted on a cinema camera 202 in a studio where filming for virtual production is performed.

[0160] The sensing system of this technology is used to estimate the position and orientation of a cinema camera 202, which is required for real-time rendering of background images displayed on an LED (Light Emitting Diode) wall. By using FMCW LiDAR1, it is possible to identify differences in materials such as the LED wall, performers, floors, and walls, making it possible to accurately estimate the self-position and orientation of the cinema camera 202.

[0161] The polarization information measured by the FMCW LiDAR 1 itself may be used for modeling, for example. Modeling using the point cloud acquired by the FMCW LiDAR 1 generates a 3D model with geometric precision that cannot be achieved using only images captured by a camera. At the same time, using material information identified based on the polarization information for modeling can improve the realism of the texture of the 3D model. For example, it is possible to generate a 3D model that reflects material information such as glass and acrylic, which are difficult to distinguish in captured images.

[0162] <Regarding the Computer> The above-described series of processes can be executed by hardware or software. When the series of processes are executed by software, the program constituting the software is installed from a program recording medium into a computer incorporated in dedicated hardware, or into a general-purpose personal computer, etc.

[0163] 24 is a block diagram showing an example of the hardware configuration of a computer that executes the above-mentioned series of processes by a program. The signal processing unit 12 is, for example, configured by a PC having a configuration similar to that shown in FIG.

[0164] A CPU (Central Processing Unit) 501 , a ROM (Read Only Memory) 502 , and a RAM (Random Access Memory) 503 are interconnected by a bus 504 .

[0165] An input / output interface 505 is also connected to the bus 504. An input unit 506 including a keyboard, a mouse, etc., and an output unit 507 including a display, a speaker, etc. are connected to the input / output interface 505. Also connected to the input / output interface 505 are a storage unit 508 including a hard disk, a nonvolatile memory, etc., a communication unit 509 including a network interface, etc., and a drive 510 that drives removable media 511.

[0166] In a computer configured as described above, the CPU 501 performs the above-described series of processes by, for example, loading a program stored in the storage unit 508 into the RAM 503 via the input / output interface 505 and the bus 504 and executing it.

[0167] The program executed by the CPU 501 is installed in the storage unit 508 by being recorded on, for example, a removable medium 511 or provided via a wired or wireless transmission medium such as a local area network, the Internet, or digital broadcasting.

[0168] The program executed by the computer may be a program that processes in chronological order according to the order described in this specification, or may be a program that processes in parallel or at the required timing, such as when called.

[0169] In this specification, a system refers to a collection of multiple components (devices, modules (components), etc.), regardless of whether all of the components are housed in the same housing. Therefore, multiple devices housed in separate housings and connected via a network, and a single device housed in a single housing with multiple modules, are both systems.

[0170] The effects described in this specification are merely examples and are not limiting, and other effects may also be present.

[0171] The embodiments of the present technology are not limited to the above-described embodiments, and various modifications are possible without departing from the spirit of the present technology.

[0172] For example, the present technology can be configured as a cloud computing system in which a single function is shared and processed collaboratively by a plurality of devices via a network.

[0173] Furthermore, each step described in the above flowchart can be executed by one device, or can be shared and executed by a plurality of devices.

[0174] Furthermore, when one step includes multiple processes, the multiple processes included in that one step can be executed by one device or can be shared and executed by multiple devices.

[0175] <Examples of Combinations of Configurations> The present technology can also have the following configurations.

[0176] (1) An information processing device comprising: an extraction unit that extracts polarization information from point cloud data acquired by a light detection and ranging unit that measures distances to multiple observation points on surrounding objects using frequency-modulated continuous waves; and a matching unit that performs matching between the point clouds from two viewpoints using features based on the polarization information. (2) The information processing device described in (1), further comprising a feature acquisition unit that acquires the features based on the polarization information. (3) The information processing device described in (2), in which the feature acquisition unit acquires, as the feature, a numerical value of the degree of polarization of the observation point included in the polarization information. (4) The information processing device described in (3), in which the matching unit matches a representative point among the multiple observation points between the point clouds from the two viewpoints. (5) The information processing device described in (4), in which the feature acquisition unit acquires, as the feature, a numerical sequence including the numerical value of the degree of polarization of the representative point and the numerical values ​​of the degree of polarization of the observation points surrounding the representative point. (6) The information processing device described in (2), wherein the feature acquisition unit acquires boundary information indicating boundaries of regions in the object that are made up of the same material as the feature. (7) The information processing device described in (6), wherein the matching unit estimates a position and orientation relationship between the two viewpoints based on the boundary information, and performs matching between the point clouds of the two viewpoints based on the estimation result of the position and orientation relationship between the two viewpoints. (8) The information processing device described in (6) or (7), wherein the feature acquisition unit identifies the boundaries of the regions that are made up of the same material based on an observation point that is adjacent to another observation point that is made up of a different material. (9) The information processing device described in (2), wherein the feature acquisition unit acquires, as the feature, results of clustering the observation points according to material. (10) The information processing device described in (9), wherein the matching unit performs matching between the point clouds of the two viewpoints for each cluster obtained by clustering the observation points according to material, and integrates the matching results between the point clouds of the two viewpoints for each cluster. (11) The information processing device according to (2), wherein the feature amount acquisition unit acquires normal information indicating a normal direction of the observation point as the feature amount.(12) The information processing device according to (11), wherein the matching unit matches the point clouds from the two viewpoints based on the normal direction of a first observation point included in the point cloud of a search source and a second observation point included in the point cloud of a search destination. (13) The information processing device according to any of (1) to (12), wherein the light detection and ranging unit is an FMCW LiDAR. (14) The information processing device according to any of (1) to (13), further comprising: a control unit that controls a moving body equipped with the light detection and ranging unit based on the polarization information and a matching result between the point clouds from the two viewpoints by the matching unit. (15) The information processing device according to any of (1) to (13), further comprising: a control unit that creates a 3D model of the object based on the polarization information and a matching result between the point clouds from the two viewpoints by the matching unit. (16) An information processing method in which an information processing device extracts polarization information from point cloud data acquired by a light detection and ranging unit that uses frequency-modulated continuous waves to measure distances to multiple observation points on surrounding objects, and performs matching between the point clouds from two viewpoints using feature quantities based on the polarization information. (17) A program for causing a computer to execute processing in which:

[0177] REFERENCE SIGNS LIST 1 FMCW LiDAR, 11 sensor unit 112 signal processing unit, 21 light detection and ranging unit, 31 position information extraction unit, 32 polarization information extraction unit, 33 feature acquisition unit, 34 matching unit, 35 control unit, 51 feature assignment unit, 71 material information assignment unit, 72 boundary extraction unit, 73 boundary information storage unit, 81 boundary information comparison unit, 82 point cloud matching unit, 101 material information assignment unit, 102 clustering unit, 111A to 111Z intra-cluster matching unit, 112 integration unit, 131 normal calculation unit

Claims

1. An information processing device comprising: an extraction unit that extracts polarization information from point cloud data acquired by a light detection and ranging unit that measures distances to multiple observation points on surrounding objects using frequency-modulated continuous waves; and a matching unit that performs matching between the point clouds from two viewpoints using features based on the polarization information.

2. The information processing device according to claim 1, further comprising a feature acquisition unit that acquires the feature based on the polarization information.

3. The information processing device according to claim 2, wherein the feature acquisition unit acquires a numerical value of the degree of polarization at the observation point included in the polarization information as the feature.

4. The information processing device according to claim 3, wherein the matching unit performs matching of representative points among the plurality of observation points between the point clouds of the two viewpoints.

5. The information processing device according to claim 4, wherein the feature acquisition unit acquires, as the feature, a sequence including the numerical value of the degree of polarization of the representative point and the numerical values ​​of the degree of polarization of the observation points surrounding the representative point.

6. The information processing device according to claim 2, wherein the feature acquisition unit acquires boundary information indicating boundaries between areas of the object that are made of the same material as the feature.

7. The information processing device according to claim 6, wherein the matching unit estimates a position and orientation relationship between the two viewpoints based on the boundary information, and performs matching between the point groups of the two viewpoints based on a result of the estimation of the position and orientation relationship between the two viewpoints.

8. The information processing device according to claim 6, wherein the feature acquisition unit identifies the boundaries of the regions made of the same material based on the observation points adjacent to other observation points made of different materials.

9. The information processing device according to claim 2, wherein the feature acquisition unit acquires, as the feature, a result of clustering the observation points according to a material.

10. The information processing device according to claim 9, wherein the matching unit performs matching between the point clouds from the two viewpoints for each cluster into which the observation points are clustered according to material, and integrates the matching results between the point clouds from the two viewpoints for each cluster.

11. The information processing device according to claim 2, wherein the feature acquisition unit acquires normal information indicating a normal direction of the observation point as the feature.

12. The information processing device according to claim 11, wherein the matching unit performs matching between the point clouds of the two viewpoints based on the normal direction of a first observation point included in the point cloud of the search source and a second observation point included in the point cloud of the search destination.

13. The information processing device according to claim 1, wherein the light detection and ranging unit is an FMCW LiDAR.

14. The information processing device according to claim 1, further comprising a control unit that controls a moving body equipped with the light detection and ranging unit based on the polarization information and a matching result between the point clouds from the two viewpoints by the matching unit.

15. The information processing device according to claim 1, further comprising a control unit that creates a 3D model of the object based on the polarization information and a matching result between the point clouds from the two viewpoints obtained by the matching unit.

16. An information processing method in which an information processing device extracts polarization information from point cloud data acquired by a light detection and ranging unit that measures distances to multiple observation points on surrounding objects using frequency-modulated continuous waves, and matches the point clouds from two viewpoints using features based on the polarization information.

17. A program for causing a computer to execute a process of extracting polarization information from point cloud data acquired by a light detection and ranging unit that uses a frequency-modulated continuous wave to measure distances to multiple observation points on surrounding objects, and matching the point clouds from two viewpoints using features based on the polarization information.

Images