JP7612908B2 - Information processing device, control method, program, and storage medium - Google Patents

Description

The present invention relates to a technology for removing noise from measured data.

Conventionally, there is known a distance measuring device that irradiates a measurement object with light, detects the reflected light from the measurement object, and calculates the distance to the measurement object based on the time difference between the timing at which the light is irradiated to the measurement object and the timing at which the reflected light from the measurement object is detected. In addition, Patent Document 1 discloses a driving assistance system that identifies and removes noise points from multiple target distance measuring points based on at least one of the degree of proximity in the angular direction between multiple target distance measuring points and the degree of proximity in the distance direction between multiple target distance measuring points.

JP 2016-161340 A

Distance data output from distance measuring devices that use laser light, such as LIDAR, generally contains noise that erroneously indicates distance even when no target is present. In such cases, a common method is to remove the noise if the reflected light intensity of the laser light is lower than a predetermined value. However, when the laser light is irradiated onto a distant object or a black object, the reflected light intensity is also low, so the above method has the problem of removing this distance data as well.

The above is one example of a problem that the present invention aims to solve. The main objective of the present invention is to provide an information processing device, a control method, a program, and a storage medium storing the program that can appropriately determine the presence of noise at a measurement point where a distance is measured.

The invention described in the claims is an information processing device having an assignment unit that assigns a flag to a measurement point for which there is another measurement point within a first distance difference, for a plurality of measurement points acquired for each frame, each of which indicates a distance from a reference point, excluding at least measurement points indicating distances within a predetermined range; a grouping unit that performs grouping of the measurement points to which the flag has been assigned into measurement point groups in which the distance difference between the measurement points is within a second distance difference; and a determination unit that performs, for each group, a determination as to whether the measurement points to which the flag has been assigned are noise or not, based on the number of measurement points in each group.

The invention described in the claims is a control method executed by an information processing device, and includes an assignment step of assigning a flag to a plurality of measurement points, each of which indicates a distance from a reference point and which are acquired for each frame, excluding at least measurement points indicating distances within a predetermined range, for which there are other measurement points with a distance difference within a first distance; a grouping step of grouping the measurement points with the flag assigned into groups of measurement points with a distance difference within a second distance difference; and a determination step of determining for each group whether the measurement points with the flag assigned are noise or not, based on the number of measurement points in each group.

The invention described in the claims is a program that causes a computer to function as an assignment unit that assigns a flag to a measurement point where there is another measurement point within a first distance difference for a plurality of measurement points acquired for each frame, each of which indicates a distance from a reference point, excluding at least measurement points that indicate distances within a predetermined range; a grouping unit that performs grouping of the measurement points to which the flag has been assigned into groups of measurement points whose distance difference between the measurement points is within a second distance difference; and a determination unit that performs, for each group, a determination as to whether the measurement points to which the flag has been assigned are noise or not, based on the number of measurement points in each group.

1 shows a schematic configuration of a lidar according to an embodiment. 4 shows the data structures of a first parameter group and a second parameter group. 13 is an example of a flowchart illustrating a procedure for a temporally stable point detection process. FIG. 1 is a diagram showing measurement points measured during each measurement period for three frames as viewed from a lidar. 11 is a graph showing the change over time in reflected light intensity detected by a receiving unit at a certain measurement point at a predetermined time after emission of irradiation light. 11 is an example of a flowchart illustrating a procedure of a first process related to detection of a spatially stable point. This is a distribution map of measurement points that clearly shows the surrounding points for the target measurement point when each measurement point is viewed from above by a lidar. 13 is an example of a flowchart illustrating a procedure of a second process related to detection of a spatially stable point. 13 shows a planar distribution map of measurement points clearly indicating time-stable points and second candidate points. 13 is a flowchart illustrating an example of a procedure of a third process related to detection of a spatially stable point. 10 shows an example of setting of a second valid point flag based on a third process in the distribution of measurement points shown in FIG. 9 . This is a diagram showing measurement points measured in each measurement period of three frames when a moving object is moving, as viewed from a lidar. FIG. 13 shows a configuration diagram of a lidar and a noise removal device according to a second modified example. The reflected light intensity (maximum value) measured at different measurement points at each time is shown. A schematic diagram of a lidar system in variant example 4 is shown.

In a preferred embodiment of the present invention, the information processing device has an assignment unit that assigns a flag to a measurement point that has other measurement points within a first distance difference for a plurality of measurement points acquired for each frame, each of which indicates a distance from a reference point; a grouping unit that performs grouping of the flagged measurement points into measurement point groups in which the distance difference between the measurement points is within a second distance difference; and a determination unit that performs, for each group, a determination as to whether the flagged measurement point is noise or not, based on the number of measurement points in each group. The "reference point" may be the position of the information processing device, or, if the device that measures the distance is separate from the information processing device, may be the position of the device that measures the distance, or may be the position of a moving body that carries the information processing device.

Generally, since an object has a spatial extent, the measurement points on the object form a cluster (a measurement point cloud) that indicates similar distances. In the case of measurement points generated by noise, even if the above-mentioned cluster is formed, the number of clusters formed is small. Taking the above into consideration, the information processing device groups measurement points that have other measurement points within the first distance difference into measurement point clusters where the distance difference between the measurement points is within the second distance difference, and performs noise judgment for each group of measurement points based on the number of measurement points in each group. This allows the information processing device to accurately judge which measurement points are noise.

In one aspect of the information processing device, the assigning unit assigns the flag to a measurement point among the plurality of measurement points, the measurement point having a distance difference within the first distance from a distance indicated by any measurement point having a measurement direction from the reference point that is close to the first distance. With this aspect, the information processing device can assign a flag to a measurement point that is not an isolated point, and appropriately determine the measurement point as a target for grouping.

In another aspect of the information processing device, the information processing device further includes a time-stable point detection unit that detects, as a time-stable point, a measurement point where at least one of the distance from the reference point or the reflected light intensity is measured, where the variance of the distance between multiple frames or the reflected intensity is less than a predetermined value, and the grouping unit performs grouping for the measurement points to which the flag is assigned and the time-stable points, for each measurement point group where the distance difference between the measurement points is within the second distance difference. The time-stable points, which are measurement points measured stably in time, are presumed to be measurement points that indicate the actual distance of an object. Therefore, the information processing device performs grouping including such time-stable points, and performs noise judgment for each group of the measurement points based on the number of measurement points in each group, thereby more accurately determining which measurement points are noise.

In another aspect of the information processing device, the assigning unit assigns the flag to a measurement point where there is another measurement point within the first distance, for a plurality of measurement points, excluding the temporally stable point and the measurement point showing a distance within a predetermined range, among the measurement points at which the distance from the reference point is measured. The above-mentioned "predetermined range" refers to, for example, a range of distances where accurate measurement is not possible and a range of distances where noise does not occur. By assigning a flag to measurement points excluding the temporally stable point and the measurement point showing a distance within a predetermined range, the information processing device can exclude measurement points for which it has already been determined whether they are noise or not from the targets for determining whether they are noise or not.

In another aspect of the information processing device, the time-stable point detection unit identifies the correspondence between the measurement points for each frame based on the speed of the moving object on which the information processing device is mounted, and calculates the variance of at least one of the distance or reflected light intensity for each corresponding measurement point. With this aspect, the information processing device can accurately detect the time-stable point even when the measurement is performed while moving.

In another aspect of the information processing device, the grouping unit executes a process for each measurement point to which the flag is assigned, in which a combination of measurement points whose measurement directions from the reference point are similar and whose distance difference is within the second distance difference is classified into the same group. With this aspect, the information processing device can accurately group each group of measurement points whose distance difference between the measurement points is within the second distance difference.

In another aspect of the information processing device, the information processing device further includes a point cloud information generating unit that generates point cloud information in which the distance from the reference point, calculated based on the timing at which the reflected light intensity detected after each emission of the pulsed light while changing the angle, is associated with the emission direction of the pulsed light, and an output unit that outputs noise-removed point cloud information in which distance information in the emission direction corresponding to the measurement points determined by the determination unit to be noise is removed from the point cloud information. With this aspect, the information processing device can suitably remove information on the distance measured due to noise from the point cloud information indicating the distance for each measurement point.

In another aspect of the information processing device, the grouping unit performs the grouping for a plurality of measurement points measured by a plurality of measuring devices, by grouping measurement points that are determined to be adjacent to each other on the boundary of the measurement ranges of the measuring devices based on the position and orientation information of each of the measuring devices into the same group. With this aspect, the information processing device can preferably prevent the measurement points of an object from being determined to be noise, even if an object is present on the boundary of the measurement ranges of the plurality of measuring devices.

In another preferred embodiment of the present invention, a control method executed by an information processing device includes an assignment step of assigning a flag to a plurality of measurement points acquired for each frame, each of which indicates a distance from a reference point, for which there are other measurement points within a first distance difference; a grouping step of grouping the flagged measurement points into groups of measurement points whose distance difference between the measurement points is within a second distance difference; and a determination step of determining for each group whether the flagged measurement points are noise or not, based on the number of measurement points in each group. By executing this control method, the information processing device can accurately determine which measurement points are noise.

In another preferred embodiment of the present invention, the program causes the computer to function as an assignment unit that assigns a flag to a measurement point where there is another measurement point within a first distance difference for a plurality of measurement points acquired for each frame, each of which indicates a distance from a reference point; a grouping unit that performs grouping of the flagged measurement points into groups of measurement points where the distance difference between the measurement points is within a second distance difference; and a determination unit that determines for each group whether the flagged measurement points are noise or not based on the number of measurement points in each group. By executing this program, the computer can accurately determine which measurement points are noise. Preferably, the program is stored in a storage medium.

Below, a preferred embodiment of the present invention will be described with reference to the drawings.

[Device configuration]
FIG. 1 shows a schematic configuration of a lidar 100 according to this embodiment. The lidar 100 is mounted on a vehicle that performs driving assistance such as automatic driving, for example. The lidar 100 irradiates a laser light (also called "irradiation light") in a predetermined angular range in the horizontal and vertical directions, and receives light (also called "reflected light") that is reflected by an object and returns, thereby discretely measuring the distance from the lidar 100 to the object and generating point cloud information indicating the three-dimensional position of the object. Furthermore, the lidar 100 generates and outputs point cloud information (also called "noise-removed point cloud information") in which data generated by erroneously detecting an object in the point cloud information (so-called false alarm data) is suitably removed.

As shown in FIG. 1, the lidar 100 mainly includes a transmitter 1, a receiver 2, a beam splitter 3, a scanner 5, a piezoelectric sensor 6, a control unit 7, and a memory 8.

The transmitter 1 is a light source that emits pulsed irradiation light toward the beam splitter 3. The transmitter 1 includes, for example, an infrared laser light-emitting element. The transmitter 1 is driven based on a drive signal "S1" supplied from the control unit 7.

The receiver 2 is, for example, an avalanche photodiode, generates a detection signal "S2" corresponding to the amount of light received, and supplies the generated detection signal S2 to the controller 7.

The beam splitter 3 transmits the pulsed irradiation light emitted from the transmitter 1. The beam splitter 3 also reflects the light reflected by the scanner 5 toward the receiver 2.

The scanner 5 is, for example, an electrostatically driven mirror (MEMS mirror), and the inclination (i.e., the angle of the optical scanning) changes within a predetermined range based on the drive signal "S3" supplied from the control unit 7. The scanner 5 reflects the irradiated light that has passed through the beam splitter 3 toward the outside of the lidar 100, and also reflects the reflected light that is incident from the outside of the lidar 100 toward the beam splitter 3. In addition, the points on which the irradiated light is irradiated within the measurement range of the lidar 100 are also called "measurement points". Furthermore, the measurement points that correspond to data generated by erroneously detecting an object are called "noise generation points", and the measurement points that correspond to data generated by detecting an actual object are also called "effective points".

The scanner 5 is also provided with a piezoelectric sensor 6. The piezoelectric sensor 6 detects distortion caused by stress in a torsion bar that supports the mirror portion of the scanner 5. The piezoelectric sensor 6 supplies the generated detection signal "S4" to the control unit 7. The detection signal S4 is used to detect the orientation of the scanner 5.

The memory 8 is composed of various types of memory such as a RAM (Random Access Memory), a ROM (Read Only Memory), and a flash memory. The memory 8 stores programs necessary for the control unit 7 to execute a predetermined process. The memory 8 also stores a first parameter group 81 and a second parameter group 82 referenced by the control unit 7. Details of the first parameter group 81 and the second parameter group 82 will be described later. The memory 8 also stores point cloud information for a predetermined number of the most recent frames that have been generated. These pieces of point cloud information are referenced when noise removal processing is performed on the point cloud information being processed.

The control unit 7 includes a processor such as a CPU (Central Processing Unit) or a GPU (Graphics Processing Unit). The control unit 7 executes a program stored in the memory 8 to perform a predetermined process. The control unit 7 is an example of a computer that executes a program.

The control unit 7 functionally has a transmission drive block 70, a scanner drive block 71, a point cloud information generation block 72, a temporally stable point detection block 73, a spatially stable point detection block 74, and an output block 75.

The transmission drive block 70 outputs a drive signal S1 that drives the transmission unit 1. The drive signal S1 includes information for controlling the emission time of the laser light-emitting element included in the transmission unit 1 and the emission intensity of the laser light-emitting element. The transmission drive block 70 controls the emission intensity of the laser light-emitting element included in the transmission unit 1 based on the drive signal S1.

The scanner driving block 71 outputs a driving signal S3 for driving the scanner 5. This driving signal S3 includes a horizontal driving signal corresponding to the resonant frequency of the scanner 5 and a vertical driving signal for vertical scanning. The scanner driving block 71 also detects the scanning angle of the scanner 5 (i.e., the emission direction of the irradiation light) by monitoring the detection signal S4 output from the piezoelectric sensor 6. The emission direction of the irradiation light is an example of the "measurement direction from the reference point" in this invention.

Based on the detection signal S2 supplied from the receiving unit 2, the point cloud information generation block 72 generates point cloud information indicating the distance and direction to the object irradiated with the irradiation light for each measurement point, with the lidar 100 as the reference point. In this case, the point cloud information generation block 72 calculates the time from when the irradiation light is emitted until the receiving unit 2 detects the reflected light as the flight time of light. Then, the point cloud information generation block 72 generates point cloud information indicating the distance according to the calculated flight time and the irradiation direction of the irradiation light corresponding to the reflected light received by the receiving unit 2 for each measurement point, and supplies the generated point cloud information to the time-stable point detection block 73. Hereinafter, the point cloud information obtained by one scan of all measurement points is regarded as point cloud information for one frame. In addition, the point cloud information generation block 72 stores the generated point cloud information for the latest predetermined number of frames in the memory 8. The point cloud information generation block 72 is an example of the "point cloud information generation unit" in the present invention.

The point cloud information can be considered as an image in which the measurement points are pixels and the distance indicated by each measurement point is the pixel value. In this case, the irradiation direction of the light at the elevation and depression angles of each measurement point differs when arranged vertically, and the irradiation direction of the light at the horizontal angle differs when arranged horizontally.

Based on the point cloud information supplied from the point cloud information generation block 72, the time-stable point detection block 73 detects measurement points (also called "time-stable points Pt") at which distances with little variation in time series are measured. The time-stable point detection block 73 then supplies the point cloud information in which a flag (also called "first valid point flag F1") indicating the judgment result regarding the time-stable point Pt is associated with each measurement point to the spatially stable point detection block 74. The time-stable point Pt is different from a noise generation point that occurs with temporal variation, and is a measurement point where the distance can be measured stably in the time series. Therefore, the control unit 7 regards the time-stable point Pt as a valid point where an object is actually measured. The time-stable point detection block 73 is an example of a "time-stable point detection unit" in this invention.

The spatially stable point detection block 74 detects measurement points (also called "spatially stable points Ps") whose distances are measured in a spatially stable manner based on the point cloud information for one frame to which the first effective point flag F1 is added, which is supplied from the temporally stable point detection block 73. The spatially stable points Ps are measurement points that constitute a measurement point cloud in which the distance difference between the measurement points is small. The spatially stable point detection block 74 then supplies to the output block 75 point cloud information in which a flag (also called "second effective point flag F2") indicating the judgment result regarding the spatially stable points Ps is associated with the first effective point flag F1 for each measurement point. The spatially stable points Ps are measurement points where an object can be measured in a spatially stable manner, unlike noise generation points that occur with spatial variation. Therefore, the control unit 7 considers the spatially stable points Ps to be effective points that actually measure the object. The spatially stable point detection block 74 is an example of the "assignment unit", "grouping unit", and "judgment unit" in this invention.

The output block 75 determines the noise generation point based on the point cloud information in which the first valid point flag F1 and the second valid point flag F2 are associated with each measurement point, which is supplied from the spatially stable point detection block 74. The output block 75 then generates point cloud information from which data corresponding to the noise generation point has been removed as noise-removed point cloud information "S5". The output block 75 then outputs the generated noise-removed point cloud information S5 to a device that controls driving assistance such as automatic driving of the vehicle (also called a "driving assistance device"). In this case, the driving assistance device may be, for example, the ECU (Electronic Control Unit) of the vehicle, or an in-vehicle device such as a car navigation device electrically connected to the vehicle. The output block 75 is an example of an "output unit" in the present invention.

2(A) shows an example of the data structure of the first parameter group 81 stored in the memory 8. FIG. 2(B) shows an example of the data structure of the second parameter group 82 stored in the memory 8. The first parameter group 81 is a parameter group referenced by the temporal stable point detection block 73, and includes the minimum ranging distance "Dth1", the stable ranging distance "Dth2", the number of calculation frames "Nf", and the effective variance threshold "Vth". The second parameter group 82 is a parameter group referenced by the spatial stable point detection block 74, and includes the candidate determination threshold "dDth1", the clump determination distance threshold "dDth2", and the clump determination score threshold "Ngth".

The minimum measurement distance Dth1 is a threshold value indicating the minimum distance to be measured. For measurement points indicating distances less than the minimum measurement distance Dth1, the temporally stable point detection block 73 determines that an accurate distance cannot be measured, and determines that the measurement points are noise generation points that do not correspond to either the temporally stable points Pt or the spatially stable points Ps.

The stable measurement distance Dth2 is a threshold value indicating the upper limit of the distance at which stable measurement is possible without generating noise. The temporally stable point detection block 73 considers, for example, a measurement point indicating a distance between the minimum measurement distance Dth1 and the stable measurement distance Dth2 to be a valid point. In addition to the stable measurement distance Dth2, the first parameter group 81 may include a parameter indicating the lower limit of the distance at which stable measurement is possible without generating noise. The distance range outside the measurement range specified by the minimum measurement distance Dth1 and the distance range at which stable measurement is possible without generating noise, specified by the stable measurement distance Dth2, are examples of the "predetermined range" in this invention.

The number of calculation frames Nf indicates the number of frames that the time-stable point detection block 73 refers to when calculating the variance value (also called "distance variance V") that indicates the temporal variation in distance for each measurement point. The use of the number of calculation frames Nf will be described in detail later.

The effective variance threshold Vth is a threshold for determining whether the target measurement point corresponds to a time stable point Pt based on the distance variance V. The time stable point detection block 73 determines that the target measurement point is a time stable point Pt when the distance variance V is less than the effective variance threshold Vth. On the other hand, the time stable point detection block 73 considers a measurement point whose distance variance V is equal to or greater than the effective variance threshold Vth to not be a time stable point Pt. A measurement point that is not considered to be a time stable point Pt based on the distance variance V is a candidate measurement point for a spatial stable point Ps (also called a "first candidate point"), and the spatial stable point detection block 74 determines whether it is a spatial stable point Ps. The effective variance threshold Vth is an example of a "predetermined value" in the present invention.

The candidate determination threshold dDth1 is a threshold for determining whether the first candidate point is an isolated point that is isolated in distance from the surrounding measurement points (also called "peripheral points"). In other words, the candidate determination threshold dDth1 is a threshold for the spatially stable point detection block 74 to determine whether the distance between the first candidate point and the peripheral points is similar. Here, the peripheral points refer to measurement points whose measurement direction from the lidar 100 is similar to that of the target measurement point. In other words, the peripheral points refer to measurement points that are adjacent to the target measurement point when the measurement points are projected onto a virtual plane on which the illumination light is irradiated. In further words, the peripheral points refer to measurement points that correspond to pixels that are adjacent (for example, within a predetermined number of pixels) vertically and horizontally to the pixel corresponding to the target measurement point when the point cloud information is regarded as an image and the measurement points are regarded as pixels in the image.

Then, the first candidate point (also called the "second candidate point") that is not considered to be an isolated point is further judged as to whether it is a spatially stable point Ps using the clump determination distance threshold dDth2 and the clump determination score threshold Ngth described below. The candidate determination threshold dDth1 is an example of the "first distance difference" in the present invention.

The clump determination distance threshold dDth2 is a threshold for determining whether or not the same clump (cluster) is formed based on the distance difference between measurement points whose measurement directions from the lidar 100 are similar. The clump determination distance threshold dDth2 is an example of the "second distance difference" in the present invention. The clump determination point number threshold Ngth is a threshold for determining whether or not the second candidate point is a noise generation point for each group based on the number of measurement points grouped into clumps (also called "group measurement point number Ng").

[Temporal Stable Point Detection Processing]
First, the process of detecting the time-stable point Pt by the time-stable point detection block 73 will be described.

Figure 3 is an example of a flowchart showing the procedure for the process of detecting a temporally stable point Pt. The temporally stable point detection block 73 repeatedly executes the process shown in Figure 3. Here, as an example, the temporally stable point detection block 73 performs a process of assigning a first valid point flag F1 of "1" to a measurement point that is a temporally stable point Pt, assigning a first valid point flag F1 of "2" to a measurement point that is a first candidate point for a spatially stable point Ps, and assigning a first valid point flag F1 of "0" to the other measurement points.

First, the time-stable point detection block 73 sets the initial value of the first valid point flag F1 of the measurement point to be processed to "0" indicating that it is invalid (step S11). Then, the time-stable point detection block 73 determines whether the distance corresponding to the target measurement point is equal to or greater than the minimum distance Dth1 (step S12). Then, if the distance corresponding to the target measurement point is less than the minimum distance Dth1 (step S12; No), the time-stable point detection block 73 leaves the first valid point flag F1 of the target measurement point at "0" (step S13). That is, in this case, the time-stable point detection block 73 considers the target measurement point to be not a valid point, and leaves the first valid point flag F1 at "0" indicating that it is invalid.

On the other hand, if the distance corresponding to the target measurement point is equal to or greater than the minimum distance Dth1 (step S12; Yes), the time-stable point detection block 73 determines whether the distance corresponding to the measurement point to be processed exceeds the stable measurement distance Dth2 (step S14). If the distance corresponding to the measurement point to be processed is equal to or less than the stable measurement distance Dth2 (step S14; No), the time-stable point detection block 73 sets the first valid point flag F1 of the target measurement point to "1" indicating validity (step S15). That is, in this case, since a distance at which stable distance measurement is possible without noise has been measured, the time-stable point detection block 73 considers the target measurement point to be a valid point and sets the first valid point flag F1 to the effect that the target measurement point is valid.

On the other hand, if the distance corresponding to the target measurement point exceeds the stable distance measurement distance Dth2 (step S14; Yes), the temporally stable point detection block 73 performs detailed determination processing in steps S16 and onward as to whether the target measurement point is a temporally stable point Pt, since the distance indicated by the target measurement point is outside the range of distances that can be stably measured without generating noise.

In this case, the temporally stable point detection block 73 first refers to the number of calculation frames Nf included in the first parameter group 81 for the target measurement point, and calculates the distance variance V based on the distance for the number of calculation frames Nf (step S16). In this case, the temporally stable point detection block 73 calculates the distance variance V by extracting the distance for the latest number of calculation frames Nf of the target measurement point based on the past point cloud information (i.e., "Nf-1" pieces of point cloud information) stored in the memory 8 and the current point cloud information. The process of step S16 will be described in detail later.

Then, the time-stable point detection block 73 determines whether the calculated distance variance V is equal to or greater than the effective variance threshold Vth (step S17). If the distance variance V is less than the effective variance threshold Vth (step S17; No), the time-stable point detection block 73 sets the first effective point flag F1 of the target measurement point to "1" indicating that it is effective (step S18). That is, in this case, the time-stable point detection block 73 considers the target measurement point to be a time-stable point Pt (i.e., a valid point) because a distance with little temporal variation was stably measured at the target measurement point.

On the other hand, if the distance variance V is equal to or greater than the effective variance threshold Vth (step S17; Yes), the temporal stable point detection block 73 sets the first effective point flag F1 of the target measurement point to "2" (step S19). That is, in this case, the temporal stable point detection block 73 determines that the target measurement point is at least not a temporal stable point Pt because a distance with temporal variation was measured at the target measurement point. Then, the temporal stable point detection block 73 sets the first effective point flag F1 of the target measurement point to "2", which indicates that the target measurement point is the first candidate point for the spatial stable point Ps. In this case, the spatial stable point detection block 74 determines whether the target measurement point is a spatial stable point Ps.

Then, the temporally stable point detection block 73 judges whether or not the processing of steps S11 to S19 has been performed for all the measurement points (step S20). If the temporally stable point detection block 73 has not performed the processing of steps S11 to S19 for all the measurement points (step S20; No), it selects one unprocessed measurement point as the measurement point to be processed next, and returns to the processing of step S11. On the other hand, if the temporally stable point detection block 73 has performed the processing of steps S11 to S19 for all the measurement points (step S20; Yes), it ends the processing of the flowchart. Thereafter, when point cloud information for the next frame is again supplied from the point cloud information generation block 72, the temporally stable point detection block 73 executes the processing of the flowchart of FIG. 3 for the point cloud information.

Here, we provide additional explanation on the calculation process of distance variance V in step S16.

The time-stable point detection block 73 calculates the distance variance V, which indicates the temporal variation in distance, for each measurement point, based on the point cloud information generated by the point cloud information generation block 72. At this time, the time-stable point detection block 73 refers to the number of calculation frames Nf included in the first parameter group 81, and calculates the distance variance V for each measurement point based on the point cloud information for the number of calculation frames Nf.

Here, a specific example of calculating distance variance V will be described with reference to FIG. 4. FIGS. 4(A) to (C) are diagrams showing measurement points measured in each measurement period of three frames as viewed from the rider 100. Here, the number of calculation frames Nf is set to "3", and the measurement point for which distance variance V is calculated is "Ptag".

In this case, the temporally stable point detection block 73 obtains the point cloud information of the current frame (i.e., obtained by the most recent scan) from the point cloud information generation block 72, and also references the point cloud information of the previous frame and the frame before that from the memory 8. Then, the temporally stable point detection block 73 obtains the distances (here, "x1", "x2", and "x3") based on this point cloud information for the measurement points Ptag for which the distance variance V is to be calculated. Then, the temporally stable point detection block 73 calculates the distance variance V from these distances x1, x2, and x3.

Next, we provide additional explanation on the point cloud information generated by the point cloud information generation block 72.

The point cloud information generation block 72 measures the time of flight from when the irradiated light is emitted until the receiving unit 2 detects the reflected light for each measurement point, and calculates the distance based on the flight time. Figure 4 is a graph showing the change in reflected light intensity over time indicated by the detection signal S2 output by the receiving unit 2 at a predetermined time after the emission of the irradiated light for a certain measurement point. As shown in Figure 4, the point cloud information generation block 72 acquires a time waveform of the reflected light intensity for each measurement point based on the detection signal S2 supplied from the receiving unit 2. In this case, the point cloud information generation block 72 regards the length of time from when the irradiated light is emitted until the reflected light intensity is at its maximum as the flight time of light ("t1" in Figure 4), and calculates the distance to the target measurement point.

In this way, the point cloud information generation block 72 calculates the distance for each measurement point based on the timing at which the reflected light intensity is at its maximum. Here, as shown in FIG. 4, due to background light, noise, etc., the reflected light intensity transitions to a value greater than 0 even at times other than when the receiver 2 receives the reflected light. Therefore, even if the irradiated object is not within the range of the measurable distance and no reflected light is generated, the distance is calculated at each measurement point. The measurement point that indicates the calculated distance when no reflected light is generated is the noise generation point.

[Spatially stable point detection process]
Next, the process of detecting the spatially stable point Ps by the spatially stable point detection block 74 will be described.

The spatially stable point detection block 74 performs three main processes (first process to third process) for detecting the spatially stable point Ps. As the first process, the spatially stable point detection block 74 judges whether the measurement point for which the first valid point flag F1 is "2" (first candidate point) is an isolated point far from the surrounding points, and determines the measurement point that is not an isolated point as the second candidate point. As the second process, the spatially stable point detection block 74 groups the second candidate point and the temporally stable point Pt by attaching the same label (also called "cluster label") to the surrounding points with a small distance difference. As the third process, the spatially stable point detection block 74 calculates the number of group measurement points Ng based on the cluster label, and judges for each group whether the second candidate point is a spatially stable point Ps based on the number of group measurement points Ng.

Figure 6 is an example of a flowchart showing the procedure of the first process for detecting the spatially stable point Ps. The spatially stable point detection block 74 repeatedly executes the process shown in Figure 6.

First, the spatially stable point detection block 74 determines the measurement point to be processed, and sets the initial value of the second valid point flag F2 of the measurement point to "0" indicating invalidity (step S21). Next, the spatially stable point detection block 74 determines whether the first valid point flag F1 of the target measurement point indicates "2" (i.e., the first candidate point) (step S22). If the first valid point flag F1 of the target measurement point indicates "2" (step S22; Yes), the spatially stable point detection block 74 proceeds to step S23. On the other hand, if the first valid point flag F1 of the target measurement point is other than "2" (step S22; No), the spatially stable point detection block 74 proceeds to step S25. In this case, since the target measurement point is not the first candidate point of the spatially stable point Ps, the spatially stable point detection block 74 leaves the second valid point flag F2 at "0" indicating invalidity.

Next, the spatially stable point detection block 74 determines whether or not there are surrounding points whose distance difference is equal to or less than the candidate determination threshold dDth1 for the measurement point whose first valid point flag F1 is "2" (step S23). A specific example of the determination in step S23 will be described later with reference to FIG. 7.

If there is a surrounding point with a distance difference less than or equal to the candidate determination threshold dDth1 for the target measurement point (step S23; Yes), the spatially stable point detection block 74 sets the second valid point flag F2 of the target measurement point to "1" indicating that it is valid (step S24). In this case, the spatially stable point detection block 74 determines that the target measurement point is not an isolated point but a second candidate point for the spatially stable point Ps, and sets the second valid point flag F2 of the target measurement point to "1" indicating that it is valid. The second valid point flag F2 indicating "1" is an example of a "flag" in this invention.

On the other hand, if there are no surrounding points whose distance difference with respect to the target measurement point is equal to or less than the candidate determination threshold dDth1 (step S23; No), the spatially stable point detection block 74 proceeds to step S25. In this case, since the target measurement point is an isolated point and not a spatially stable point Ps, the spatially stable point detection block 74 leaves the second valid point flag F2 at "0", indicating that it is invalid.

The spatially stable point detection block 74 then determines whether or not the processing of steps S21 to S24 has been performed for all measurement points (step S25). If the spatially stable point detection block 74 has not performed the processing of steps S21 to S24 for all measurement points (step S25; No), it selects one unprocessed measurement point as the measurement point to be processed and returns to step S21. On the other hand, if the spatially stable point detection block 74 has performed the processing of steps S21 to S24 for all measurement points (step S25; Yes), it ends the processing of the flowchart. After that, when point cloud information for the next frame is supplied from the temporally stable point detection block 73, the spatially stable point detection block 74 performs the processing of the flowchart in FIG. 6 on the point cloud information.

Here, we provide additional explanation of the processing in step S23 with reference to Figure 7.

Figure 7 is a distribution diagram of measurement points that clearly shows the surrounding points for measurement point Ptag when each measurement point is viewed from the rider 100. In Figure 7, the surrounding points for measurement point Ptag are hatched.

In the example of FIG. 7, the spatially stable point detection block 74 determines two measurement points vertically and horizontally among the measurement points that have been measured at the time when the measurement point Ptag is measured, as surrounding points of the target measurement point Ptag. In other words, when the point cloud information is considered as an image, the spatially stable point detection block 74 determines the measured measurement points that correspond to pixels within two pixels above, below, left, and right from the pixel position of the measurement point Ptag as surrounding points. Note that, when the distances to all measurement points of the current frame have already been measured, the spatially stable point detection block 74 may also determine measurement points located below the measurement point Ptag (i.e., measurement points excluded from the surrounding points as unmeasured measurement points) as surrounding points.

The spatially stable point detection block 74 then calculates the distance difference between the distance indicated by each surrounding point and the distance indicated by the measurement point Ptag, and if there is a surrounding point for which the distance difference is equal to or less than the candidate determination threshold dDth1, it considers the measurement point Ptag to be the second candidate point. In this case, the spatially stable point detection block 74 sets the second valid point flag F2 of the measurement point Ptag to "1", indicating that it is valid. By performing this process for all measurement points, the spatially stable point detection block 74 can preferably identify a first candidate point that is not an isolated point as a second candidate point.

Figure 8 is an example of a flowchart showing the procedure of the second process for detecting the spatially stable point Ps. The spatially stable point detection block 74 executes the process shown in Figure 8 after executing the first process shown in Figure 6.

First, the spatially stable point detection block 74 resets the cluster label assigned to each measurement point (step S31). For example, the spatially stable point detection block 74 sets the cluster label assigned to each measurement point to a predetermined initial value indicating that grouping has not been processed.

Next, the spatially stable point detection block 74 determines the measurement point to be processed and determines whether the first valid point flag F1 or the second valid point flag F2 of the measurement point is "1" indicating that the measurement point is valid (step S32). That is, in this case, the spatially stable point detection block 74 determines whether the measurement point of interest is a time-stable point Pt or corresponds to one of the second candidate points. Then, if the first valid point flag F1 or the second valid point flag F2 of the measurement point of interest is "1" (step S32; Yes), the spatially stable point detection block 74 proceeds to step S33. On the other hand, if neither the first valid point flag F1 nor the second valid point flag F2 of the measurement point of interest is "1" (i.e., not valid) (step S32; No), the spatially stable point detection block 74 considers that the measurement point of interest is neither a time-stable point Pt nor a second candidate point. In this case, the spatially stable point detection block 74 leaves the cluster label of the measurement point of interest at its initial value. In this case, the target measurement point is excluded from grouping by cluster label.

Next, the spatially stable point detection block 74 determines whether or not there is a surrounding point for which the first valid point flag F1 or the second valid point flag F2 is "1" and the distance difference between the surrounding point and the target measurement point is less than the clump determination distance threshold dDth2 (step S33). A specific example of the process of step S33 will be described later with reference to FIG. 9.

Then, if the first valid point flag F1 or the second valid point flag F2 is "1" and there is a surrounding point whose distance difference from the target measurement point is less than the clump determination distance threshold dDth2 (step S33; Yes), the spatially stable point detection block 74 assigns the same cluster label to the combination of the target measurement point and the corresponding surrounding point (step S34). In this way, the spatially stable point detection block 74 can assign the same cluster label to a combination of a temporally stable point Pt and/or a second candidate point whose measurement direction from the lidar 100 is similar and whose distance difference is small, thereby assigning the same cluster label to a group of measurement points indicating distances to the same object.

On the other hand, if there are no surrounding points for which the first valid point flag F1 or the second valid point flag F2 is "1", or if the distance difference between any of these surrounding points and the target measurement point is equal to or greater than the clump determination distance threshold dDth2 (step S33; No), the spatially stable point detection block 74 assigns a new cluster label to the target measurement point (step S35). In this case, the spatially stable point detection block 74 assigns a cluster label indicating a value that has not yet been used to the target measurement point.

Then, the spatially stable point detection block 74 determines whether or not the processing of steps S32 to S35 has been performed for all measurement points (step S36). If the processing of steps S32 to S35 has not been performed for all measurement points (step S36; No), the spatially stable point detection block 74 determines one of the unprocessed measurement points as the measurement point to be processed, and returns to step S32. On the other hand, if the spatially stable point detection block 74 has performed the processing of steps S31 to S35 for all measurement points (step S36; Yes), it ends the processing of the flowchart. After that, if the spatially stable point detection block 74 has performed the first processing shown in FIG. 6 again, it executes the processing of the flowchart shown in FIG. 8.

Here, a specific example of the second process is described with reference to FIG.

Figure 9 shows a planar distribution diagram of measurement points that clearly shows the temporally stable points Pt and the second candidate points. Figure 9(A) clearly shows the temporally stable points Pt, the second candidate points, and other measurement points in a distinguishable manner.

In this case, the spatially stable point detection block 74 performs the second process on all the measurement points, thereby attaching the cluster label "L1" to the time-stable point Pt and the second candidate point in the dashed frame 20, and attaching the cluster label "L2" to the second candidate point in the dashed frame 21. In this case, the difference between the distances indicated by the time-stable point Pt and the second candidate point in the dashed frame 20 and the distances indicated by the surrounding points in the dashed frame 20 is less than the clump determination distance threshold dDth2. Similarly, the difference between the distances indicated by the time-stable point Pt and the second candidate point in the dashed frame 21 and the distances indicated by the surrounding points in the dashed frame 21 is less than the clump determination distance threshold dDth2. Therefore, it is estimated that the measurement points in the dashed frame 20 are measurement points that measure the distance to the same object. Similarly, it is estimated that the measurement points in the dashed frame 21 are measurement points that measure the distance to the same object. Therefore, by using the second process, the spatially stable point detection block 74 can appropriately assign the same cluster label to the temporally stable point Pt and the cluster of second candidate points that measure the distance to the same object.

Figure 10 is an example of a flowchart showing the procedure of the third process for detecting the spatially stable point Ps. The spatially stable point detection block 74 executes the process shown in Figure 10 after executing the second process shown in Figure 8.

First, the spatially stable point detection block 74 calculates the number of group measurement points Ng for each group divided by the cluster label (step 41).

Then, the spatially stable point detection block 74 determines whether the group measurement point count Ng for the target group is less than the clump determination point count threshold Ngth (step S42). If the group measurement point count Ng is less than the clump determination point count threshold Ngth (step S42; Yes), the spatially stable point detection block 74 sets the second valid point flag F2 of the measurement point belonging to the target group to "0" indicating invalidity (step S43). In this case, the spatially stable point detection block 74 determines that the number of measurement points belonging to the target group is small, and that the measurement points belonging to the target group are not spatially stable points Ps (i.e., are noise generation points).

On the other hand, if the group measurement point count Ng is equal to or greater than the clump determination point count threshold Ngth (step S42; No), the spatially stable point detection block 74 does not change the second valid point flag F2 of the measurement point belonging to the target group. In this case, if the target measurement point is the second candidate point, the second valid point flag F2 remains at "1", and if the target measurement point is the temporally stable point Pt, the second valid point flag F2 remains at "0". Therefore, in this case, the spatially stable point detection block 74 can assign the second valid point flag F2 to "1", indicating validity, to the second candidate point belonging to the group in which the group measurement point count Ng is equal to or greater than the clump determination point count threshold Ngth.

The spatial stable point detection block 74 then determines whether or not the determination process of step S42 has been performed for all groups (step S44). If the spatial stable point detection block 74 has performed the determination process of step S42 for all groups (step S44; Yes), it ends the processing of the flowchart. On the other hand, if there is a group for which the determination process of step S42 has not been performed (step S44; No), the spatial stable point detection block 74 again performs the processing of step S42 for the group for which the determination process has not been performed.

After the third process is executed, the spatially stable point detection block 74 supplies the point cloud information to which the first valid point flag F1 and the second valid point flag F2 are added to the output block 75. In this case, the output block 75 regards a measurement point where neither the first valid point flag F1 nor the second valid point flag F2 is "1" as a noise generation point, and generates noise-removed point cloud information S5 by removing the measurement data indicating the distance and direction of the measurement point from the point cloud information. In this case, the noise-removed point cloud information S5 includes measurement data corresponding to any of the temporally stable points Pt, the spatially stable points Ps, or measurement points indicating a distance between the minimum distance Dth1 and the stable distance Dth2. The output block 75 then supplies the noise-removed point cloud information S5 to the driving assistance device.

Figure 11 shows an example of setting the second valid point flag F2 based on the third process in the distribution of measurement points shown in Figure 9. In the example of Figure 11, the lump determination score threshold Ngth is set to "8".

In this case, the output block 75 recognizes that the group measurement point count Ng of the group to which the cluster label L1 is assigned is "14", and the group measurement point count Ng of the group to which the cluster label L2 is assigned is "6". Then, since the group measurement point count Ng of the group to which the cluster label L1 is assigned is equal to or greater than the cluster determination point threshold Ngth (14>8), the output block 75 leaves the second valid point flag F2 of the second candidate points P1-P4 to which the cluster label L1 is assigned at "1". In this case, the second candidate points P1-P4 are regarded by the output block 75 as spatially stable points Ps (i.e. valid points), and the measurement data of the second candidate points P1-P4 is not removed from the point cloud information.

On the other hand, because the group measurement point count Ng of the group to which the cluster label L2 is assigned is less than the clump determination point count threshold Ngth (6<8), the output block 75 sets the second valid point flag F2 of the second candidate points P5 to P10 to which the cluster label L1 is assigned to "0." In this case, the second candidate points P5 to P10 are regarded as noise generating points by the output block 75, and the measurement data of the second candidate points P5 to P10 is removed from the point cloud information.

As described above, the spatially stable point detection block 74 of the control unit 7 of the lidar 100 according to this embodiment assigns the second valid point flag F2 indicating validity to a measurement point where there are other measurement points within the candidate determination threshold dDth1 for a plurality of measurement points acquired for each frame, each of which indicates a distance from a reference point. Then, the spatially stable point detection block 74 groups the measurement points to which the second valid point flag F2 indicating validity is assigned into measurement point groups in which the distance between the measurement points is within the cluster determination distance threshold dDth2. Then, the spatially stable point detection block 74 judges for each group whether the measurement points to which the second valid point flag F2 is assigned are noise generation points or not, based on the group measurement point number Ng, which is the number of measurement points for each group. This allows the spatially stable point detection block 74 to accurately judge whether the measurement point is a spatially stable point Ps or a noise generation point.

[Modification]
Preferred modifications of the above-described embodiment will be described below. The following modifications may be applied to the above-described embodiment in any combination.

(Variation 1)
When the lidar 100 is mounted on a moving object such as a vehicle, the time-stable point detection block 73 may determine the correspondence between the measurement points for each frame by taking into account the speed of the moving object, and calculate the distance variance V for each corresponding measurement point.

Figures 12 (A) to (C) are diagrams showing measurement points measured during each measurement period of three frames as viewed from the lidar 100 when the mobile body carrying the lidar 100 is moving. Here, the number of calculation frames Nf is set to "3", and the measurement point for which distance variance V is calculated is "Ptag1".

In this case, the time-stable point detection block 73 acquires the point cloud information of the current frame (i.e., obtained by the latest scan) from the point cloud information generation block 72, and also refers to the point cloud information of the previous frame and the point cloud information of the frame before the previous frame from the memory 8. When the time-stable point detection block 73 identifies the measurement points Ptag2 and Ptag3 corresponding to the measurement point Ptag1 in the current frame in the previous frame and the frame before the previous frame, it identifies them based on the speed of the moving body and the frame difference from the current frame. For example, when the moving body is within a predetermined speed range, the time-stable point detection block 73 shifts the measurement points Ptag2 and Ptag3 corresponding to the measurement point Ptag1 by one to the right or left (here, right) for each frame shift. In this case, the memory 8 stores map information that corresponds to the speed (and moving direction) of the moving body and the number and direction of the measurement points to be shifted for each frame difference. Then, the time-stable point detection block 73 refers to this map information and identifies the measurement points Ptag2 and Ptag3 of the past frame that correspond to the measurement point Ptag1 of the current frame from the information on the speed of the moving object output by the sensor attached to the moving object.The time-stable point detection block 73 then calculates the distance variance V from the distances "y1", "y2", and "y3" indicated by these measurement points Ptag1, Ptag2, and Ptag3, respectively.

In another example, the temporally stable point detection block 73 converts each measurement point indicated by the point cloud information into an absolute coordinate system used in maps, etc., based on information such as the position and orientation of the moving body and the installation position and orientation of the lidar 100 on the moving body. Then, the temporally stable point detection block 73 considers the measurement point of the current frame indicating a point that is stably measured in the absolute coordinate system in each frame of the calculation frame number Nf to be the temporally stable point Pt.

In this way, according to variant example 1, the time-stable point Pt can be suitably detected even when the rider 100 is moving.

(Variation 2)
The configuration of the LIDAR 100 is not limited to the configuration shown in Fig. 1. For example, functions corresponding to the time stable point detection block 73 and the spatial stable point detection block 74 of the control unit 7 may be provided in a device separate from the LIDAR 100.

Figure 13 shows a configuration diagram of the lidar 100X and the noise removal device 200 according to the second modification. In this case, the lidar 100X supplies the point cloud information generated by the point cloud information generation block 72 to the noise removal device 200.

The noise removal device 200 has a control unit 7A and a memory 8A. A first parameter group 81 and a second parameter group 82 are stored in the memory 8A. The control unit 7A functionally has a time stable point detection block 73, a spatial stable point detection block 74, and an output block 75. The time stable point detection block 73 receives point cloud information generated by the point cloud information generation block 72 of the lidar 100X, and performs a time stable point detection process on the received point cloud information by referring to the first parameter group 81. The spatial stable point detection block 74 performs a spatial stable point detection process on the point cloud information to which the time stable point detection block 73 has attached the first valid point flag F1 by referring to the second parameter group 82. The output block 75 then generates noise-removed point cloud information S5 based on the point cloud information with the first valid point flag F1 and the second valid point flag F2 attached, supplied from the spatially stable point detection block 74, and outputs the noise-removed point cloud information S5.

The noise removal device 200 may be realized by a driving assistance device. The first parameter group 81 and the second parameter group 82 may be stored in another device having a memory that the noise removal device 200 can refer to. In the second modification, the noise removal device 200 is an example of an "information processing device" according to the present invention.

(Variation 3)
In addition to, or instead of, determining whether each measurement point is a temporally stable point Pt based on the distance variance V, the temporally stable point detection block 73 may determine whether each measurement point is a temporally stable point Pt based on the reflected light intensity for the number of calculation frames Nf.

For example, the temporally stable point detection block 73 acquires the maximum reflected light intensity (in the case of FIG. 5, the reflected light intensity at time t1) among the reflected light intensity waveforms (see FIG. 5) for the target measurement point for each frame of the most recent calculation frame number Nf. Then, if the variance of the reflected light intensity (also called "intensity variance") for the calculation frame number Nf is less than a predetermined threshold, the temporally stable point detection block 73 determines that the target measurement point is a temporally stable point Pt. On the other hand, if the intensity variance is equal to or greater than the predetermined threshold, the temporally stable point detection block 73 determines that the target measurement point is not a temporally stable point Pt.

For example, consider an example where the number of calculation frames Nf is "6" and the intensity distribution is calculated based on the reflected light intensity acquired at times "T1" to "T6" for each frame. Figures 14 (A) and (B) show the reflected light intensity (maximum value) measured at times T1 to T6 at different measurement points. For the measurement points where the reflected light intensity shown in Figure 14 (A) was measured, the variation in reflected light intensity at times T1 to T6 is small, so the temporally stable point detection block 73 determines that the intensity variance is less than a predetermined threshold value. In this case, the temporally stable point detection block 73 considers the measurement points where the reflected light intensity shown in Figure 14 (A) was measured to be the temporally stable points Pt.

On the other hand, for the measurement point where the reflected light intensity shown in FIG. 14(B) was measured, the variation in reflected light intensity at times T1 to T6 is large, and the joint variance is equal to or greater than a predetermined threshold. In this case, the temporally stable point detection block 73 considers the measurement point where the reflected light intensity shown in FIG. 14(A) was measured not to be a temporally stable point Pt.

The temporal stable point detection block 73 may consider the target measurement point to be a temporal stable point Pt when the distance variance V is less than the effective variance threshold Vth and the intensity variance is less than a predetermined threshold. In another example, the temporal stable point detection block 73 may consider the target measurement point to be a temporal stable point Pt when the distance variance V is less than the effective variance threshold Vth or the intensity variance is less than a predetermined threshold. For a measurement point that is not considered to be a temporal stable point Pt, the first effective point flag F1 is set to "2" (first candidate point), and a determination is made as to whether or not it is a spatial stable point Ps.

According to this modified example, the lidar 100 can more accurately determine whether each measurement point is a noise generation point or not.

(Variation 4)
When multiple lidars are used, noise determination may be performed taking into account the boundary areas of the measurement ranges of each lidar.

Figure 15 shows a schematic diagram of a lidar system in Modification 4. The lidar system has a first lidar 100Xa and a second lidar Xb corresponding to the lidar 100X described in Modification 2 shown in Figure 13, and a noise removal device 200a corresponding to the noise removal device 200 described in Modification 2. In this configuration, the noise removal device 200a receives point cloud information from the first lidar 100Xa and the second lidar 100Xb, respectively. Hereinafter, the point cloud information generated by the first lidar 100Xa will also be referred to as "first point cloud information," and the point cloud information generated by the second lidar Xb will also be referred to as "second point cloud information."

When detecting a spatially stable point, the noise removal device 200a leaves the second valid point flag at 1 (i.e., considers it to be a spatially stable point Ps) if the total number of group measurement points Ng of the measurement point clouds determined to be measurement points for the same object is equal to or greater than the clump determination point threshold Ngth, even if the group measurement point number Ng of the measurement point clouds measured by a single rider (first rider 100Xa or second rider 100Xb) is less than the clump determination point threshold Ngth. In other words, the noise removal device 200a groups, as the same group, the measurement points determined to be adjacent to each other at the boundary of the measurement range of each rider based on the position and attitude information of each rider for the multiple measurement points measured by each rider.

Specifically, first, the noise removal device 200a assigns cluster labels to the measurement points by executing the first process shown in FIG. 6 and the second process shown in FIG. 8 for each of the first point cloud information and the second point cloud information. In the example of FIG. 15, the same cluster label is assigned to the measurement point groups P21 to P25 belonging to the first point cloud information, and the same cluster label is assigned to the measurement point groups P26 to P29 belonging to the second point cloud information. In this case, the group measurement point counts Ng for each measurement point group are "5" and "4", respectively, which are less than the clump determination score threshold Ngth (here, 8).

Here, the noise removal device 200a determines that the measurement point group P21-P25 belonging to the first point cloud information and the measurement point group P26-P29 belonging to the second point cloud information are close to each other based on the information on the position and attitude of the first rider 100Xa and the second rider 100Xb installed on the vehicles. The information on the position and attitude of the first rider 100Xa and the second rider 100Xb installed on the vehicles is stored in advance in the memory 8A or the like. In this case, the noise removal device 200Aa considers the measurement point group P21-P25 belonging to the first point cloud information and the measurement point group P26-P29 belonging to the second point cloud information to be spatially stable points Ps, respectively, because the sum (9) of these group measurement points Ng (5 and 4) is equal to or greater than the cluster determination point threshold Ngth (set to 8 here).

When the noise removal device 200a determines that the measurement ranges of the first rider 100Xa and the second rider 100Xb overlap or are adjacent based on the position and attitude information installed on the vehicles of the first rider 100Xa and the second rider 100Xb, the noise removal device 200a may integrate the first point cloud information and the second point cloud information into one piece of point cloud information, and then perform the first to third processes on the integrated point cloud information.

In this way, according to this modified example, when there are multiple lidars with overlapping or adjacent measurement ranges, it is possible to effectively prevent the measurement point of an object from being erroneously determined to be a noise generation point even if the object is located on the boundary of the measurement ranges of these lidars.

REFERENCE SIGNS LIST 1 Transmitter 2 Receiving unit 3 Beam splitter 5 Scanner 6 Piezo sensor 7, 7A Control unit 8, 8A Memory 100, 100X Lidar 200 Noise removal device

Claims (10)

an assigning unit that assigns a flag to a measurement point that has another measurement point within a first distance difference from a plurality of measurement points that are acquired for each frame and indicate a distance from a reference point, excluding at least the measurement points that indicate a distance within a predetermined range;
a grouping unit that groups the flagged measurement points into groups of measurement points each having a distance difference between the measurement points that is within a second distance difference;
a determination unit that determines, for each group, whether the measurement points to which the flag is assigned are noise or not, based on the number of measurement points for each group;
An information processing device having the above configuration. 2 . The information processing device according to claim 1 , wherein the assignment unit assigns the flag to a measurement point among the plurality of measurement points, the measurement point having a distance difference within the first distance difference from a distance indicated by any one of the measurement points having a measurement direction from the reference point that is similar to the measurement direction. a time-stable point detection unit that detects, as a time-stable point, a measurement point at which a variance of at least one of a distance or a reflected light intensity between a plurality of frames is less than a predetermined value, among the measurement points at which a distance from the reference point is measured;
The information processing apparatus according to claim 1 , wherein the grouping unit performs grouping for each of the measurement points to which the flag is assigned and the temporally stable points, the measurement points each having a distance difference within the second distance difference. 4. The information processing device according to claim 3, wherein the assignment unit assigns the flag to a measurement point where there is another measurement point that is within the first distance difference , for a plurality of measurement points, excluding the temporally stable point and a measurement point that indicates a distance within the predetermined range, among the measurement points at which the distance from the reference point is measured. The information processing device according to claim 3 or 4, wherein the time-stable point detection unit identifies the correspondence between the measurement points for each frame based on the speed of a moving object on which the information processing device is mounted, and calculates the variance of at least one of the distance or the reflected light intensity for each corresponding measurement point. The information processing device according to claim 3 or 4, wherein the grouping unit performs a process for classifying, into the same group, combinations of measurement points whose measurement directions from the reference point are similar and whose distance difference is within the second distance difference, for each measurement point to which the flag is assigned. a point cloud information generating unit that generates point cloud information in which a distance from the reference point calculated based on a timing at which a reflected light intensity detected after each emission of the pulsed light while changing an angle is maximized is associated with an emission direction of the pulsed light;
an output unit that outputs noise-removed point cloud information obtained by removing, from the point cloud information, information on distances in the emission direction corresponding to the measurement points that the determination unit has determined to be noise;
7. The information processing device according to claim 1, further comprising: The information processing device according to any one of claims 1 to 7, wherein the grouping unit performs the grouping of a plurality of measurement points measured by a plurality of measuring devices, into the same group, measurement points that are determined to be adjacent to each other on the boundary of the measurement range of each of the measuring devices based on information on the position and orientation of each of the measuring devices. A control method executed by an information processing device, comprising:
a flag assignment step for assigning a flag to a measurement point for which there exists another measurement point within a first distance difference, the measurement points being acquired for each frame and each of which indicates a distance from a reference point, and at least measurement points indicating distances within a predetermined range being excluded;
a grouping step of grouping the flagged measurement points into groups of measurement points each having a distance difference between the measurement points that is within a second distance difference;
a determination step of determining, for each group, whether or not the measurement points to which the flag is assigned are noise based on the number of measurement points for each group;
The control method includes: an assigning unit that assigns a flag to a measurement point that has another measurement point within a first distance difference from a plurality of measurement points that are acquired for each frame and indicate a distance from a reference point, excluding at least the measurement points that indicate a distance within a predetermined range;
a grouping unit that groups the flagged measurement points into groups of measurement points each having a distance difference between the measurement points that is within a second distance difference;
A program that causes a computer to function as a determination unit that determines, for each group, whether or not the measurement points to which the flag is assigned are noise, based on the number of measurement points in each group.