WO2021207954A1 - 一种目标识别的方法和装置 - Google Patents

一种目标识别的方法和装置 Download PDF

Info

Publication number
WO2021207954A1
WO2021207954A1 PCT/CN2020/084818 CN2020084818W WO2021207954A1 WO 2021207954 A1 WO2021207954 A1 WO 2021207954A1 CN 2020084818 W CN2020084818 W CN 2020084818W WO 2021207954 A1 WO2021207954 A1 WO 2021207954A1
Authority
WO
WIPO (PCT)
Prior art keywords
point cloud
point
effective
points
grid
Prior art date
Application number
PCT/CN2020/084818
Other languages
English (en)
French (fr)
Inventor
邵政涵
李晨鸣
彭学明
Original Assignee
华为技术有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 华为技术有限公司 filed Critical 华为技术有限公司
Priority to EP20930782.6A priority Critical patent/EP4130798A4/en
Priority to PCT/CN2020/084818 priority patent/WO2021207954A1/zh
Priority to CN202080004293.4A priority patent/CN112513679B/zh
Publication of WO2021207954A1 publication Critical patent/WO2021207954A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
    • G01S17/00—Systems using the reflection or reradiation of electromagnetic waves other than radio waves, e.g. lidar systems
    • G01S17/88—Lidar systems specially adapted for specific applications
    • G01S17/93—Lidar systems specially adapted for specific applications for anti-collision purposes
    • G01S17/931—Lidar systems specially adapted for specific applications for anti-collision purposes of land vehicles
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
    • G01S17/00—Systems using the reflection or reradiation of electromagnetic waves other than radio waves, e.g. lidar systems
    • G01S17/88—Lidar systems specially adapted for specific applications
    • G01S17/89—Lidar systems specially adapted for specific applications for mapping or imaging
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
    • G01S7/00—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00
    • G01S7/48—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S17/00
    • G01S7/4802—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S17/00 using analysis of echo signal for target characterisation; Target signature; Target cross-section
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/20—Image preprocessing
    • G06V10/30—Noise filtering
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/50—Context or environment of the image
    • G06V20/56—Context or environment of the image exterior to a vehicle by using sensors mounted on the vehicle
    • G06V20/58—Recognition of moving objects or obstacles, e.g. vehicles or pedestrians; Recognition of traffic objects, e.g. traffic signs, traffic lights or roads
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V2201/00—Indexing scheme relating to image or video recognition or understanding
    • G06V2201/12—Acquisition of 3D measurements of objects

Definitions

  • This application relates to the field of automatic driving, and more specifically, to a method and device for target recognition.
  • lidar is one of the main visual sensors on vehicles.
  • Lidar is an active sensor used to detect the position of a target by emitting a laser beam.
  • the distance information of the target can be obtained from the laser point cloud reflected from the target, so as to realize the detection, tracking and identification of the target.
  • lidar adopts infrared laser ranging method, it is affected by road dust, vehicle exhaust, and water mist reflection from watery road surface. It is easy to produce a large number of false reflection points. These false reflection points will cause recognition errors and affect the follow-up Tracking path planning, etc.
  • the existing methods for removing false reflection points in the laser point cloud do not have a very suitable solution.
  • most of the existing methods are more complicated and costly to implement, resulting in more false targets in the laser point cloud, which seriously reduces the accuracy and safety of automatic driving.
  • the present application provides a method and device for target recognition, which can effectively reduce the false detection rate of laser point clouds and increase the detection rate of false targets in laser point clouds.
  • the method is simple to implement and easy to implement.
  • a target recognition method is provided, and the execution subject of the method may be a target recognition device integrated on a mobile device.
  • the execution body of the method may also be a chip, a chip system, or an integrated circuit on a mobile device, or the execution body of the method may also be a lidar.
  • the method includes: receiving a target laser point cloud and generating a point cloud grid corresponding to the target laser point cloud, wherein a point in the target laser point cloud corresponds to an effective point in the point cloud grid; according to the point cloud The geometric features of the grid to determine whether the target laser point cloud is a noisy point cloud.
  • the geometric features of the point cloud grid include: the microscopic roughness of the point cloud grid, and/or the point cloud in the point cloud grid
  • the discontinuity of the distribution may include vehicles, airplanes, unmanned aerial vehicles, ships, and other devices that can move spatial positions or change spatial shapes through human operations.
  • the method of target recognition provided by the first aspect converts the laser point cloud into a grid by using the laser point cloud obtained by a single laser radar single frame detection, and uses the geometric characteristics of the point cloud grid instead of the intensity characteristics of the lidar reflection points Identifying whether the laser point cloud is a false target caused by splashing water, road dust or automobile exhaust caused by road surface water can effectively reduce the false detection rate and increase the detection rate of false targets.
  • the method is simple to implement and easy to implement. Moreover, since a single laser radar is used for single frame detection, the calculation delay can be reduced.
  • the micro-roughness of the point cloud grid is: the ratio of the total number of mutation points in the point cloud grid to the number of effective points included in the point cloud grid, where the The total number of mutation points in the point cloud grid is the sum of the number of mutation points in each row of the point cloud grid.
  • the micro-roughness of the point cloud grid can be determined relatively quickly and simply, which is easy to implement, has high accuracy, and low complexity.
  • the distribution discontinuity of the point cloud in the point cloud grid is: the sum of the discontinuities of each column of the point cloud in the point cloud grid.
  • the discontinuity of the point cloud distribution in the point cloud grid can be determined relatively quickly and simply, which is easy to implement, has high accuracy, and low complexity.
  • judging (or determining) whether the target laser point cloud is a noise point cloud includes:
  • the target laser point cloud is a noise point cloud
  • the target laser point cloud is a noise point cloud.
  • the method further includes:
  • the number of effective points in the i-th row of the point cloud in the point cloud grid is less than or equal to 2, it is determined that the number of mutation points in the i-th row in the point cloud grid is 0, and i is a positive integer;
  • the number of mutation points of each row of point clouds is determined by the above method, which is easy to implement, and the result is higher in accuracy, which can better reflect the mutation situation of each row of point clouds.
  • the second valid point is determined The effective point is the mutation point;
  • the second effective point is not a sudden change point.
  • the three consecutive effective points in the i-th row of point cloud are the first effective point P, the second effective point Q, and the third effective point S, respectively, and O is the origin of the coordinate.
  • the reference point corresponding to the second effective point Q is Q 1 ,
  • the difference in length is greater than or equal to the preset threshold, then the second effective point Q is a sudden change point, if and The difference in length is less than the preset threshold, then the second effective point Q is not a sudden change point.
  • the method further includes:
  • the third effective point is determined to be the first sudden change point, and the first effective point and the second effective point are divided on the plane.
  • the effective points other than the third effective point and the third effective point respectively determine the second line between the first effective point and the third effective point, and the second line between the second effective point and the third effective point.
  • the fourth effective point and the fifth effective point with the furthest distance between the three lines if the distance between the fourth effective point and the second line is greater than or equal to the third threshold, then the sudden change point in the point cloud of the i-th row The number increases by 1.
  • the number of mutation points in the i-th row of point cloud increases by one; repeat the above process until Until no new straight line is generated on the plane, the total number of mutation points in the line of point cloud can be obtained by accumulation.
  • the total number of mutation points in the i-th row of point cloud is zero.
  • the number of mutation points of each row of point clouds is determined by the above method, which is convenient for implementation, and the result is more accurate, which can better reflect the mutation situation of each row of point clouds.
  • the discontinuity of each column of the point cloud in the point cloud grid is determined according to the number of valid points included in each column of the point cloud.
  • the method further includes:
  • the discontinuity of the j-th column of point cloud is determined to be a, j is a positive integer, and a is a positive number;
  • the discontinuity of the jth column of point cloud is determined Is 0, j is a positive integer;
  • the jth column is determined
  • the discontinuity of the point cloud is b, j is an integer greater than 1, and b is a positive number;
  • the number of valid points in the j-th column of point cloud in the point cloud grid is greater than 1, it is determined whether the row coordinates corresponding to the valid points in the j-th column of point cloud are continuous. The degree of continuity is 0. If it is not continuous, the discontinuity of the point cloud in the j-th column is determined to be c, where j is an integer greater than 1, and c is a positive number.
  • the discontinuity of each column of point clouds is determined by the above method, which is easy to implement, has low complexity, and the accuracy of the result is higher, which can better reflect the discontinuity of each column of point clouds.
  • the target laser point cloud is obtained by single-frame detection of a single laser radar.
  • a single laser radar single frame point cloud is used to obtain the point cloud grid, which reduces the calculation delay, is simple to calibrate, and is easy to implement.
  • the false target laser spot is caused by one or more of water splashes caused by road surface water, automobile exhaust, or road dust.
  • a device for target recognition includes a unit for executing each step in the above first aspect or any possible implementation of the first aspect.
  • a device for target recognition includes at least one processor and a memory, and the at least one processor is configured to execute the above first aspect or the method in any possible implementation of the first aspect.
  • a device for target recognition includes at least one processor and an interface circuit, and the at least one processor is configured to execute the above first aspect or the method in any possible implementation of the first aspect.
  • a lidar is provided, and the lidar includes the target recognition device provided in the third, fourth, or fifth aspect.
  • the lidar can be installed on a mobile device.
  • mobile devices may include vehicles, airplanes, unmanned aerial vehicles, ships, and other devices that can move spatial positions or change spatial shapes through human operations.
  • a computer program product includes a computer program.
  • the computer program product includes a computer program.
  • the computer program is executed by a processor, the computer program is used to execute the method in the first aspect or any possible implementation of the first aspect.
  • a computer-readable storage medium stores a computer program, and when the computer program is executed, it is used to execute the first aspect or any possible implementation of the first aspect , Or execute the method in the second aspect or any possible implementation of the second aspect.
  • a chip or integrated circuit in an eighth aspect, includes a processor, which is used to call and run a computer program from a memory, so that a device installed with the chip or integrated circuit executes the first aspect or the first aspect.
  • the method in any possible implementation of one aspect.
  • a chip system which is applied to lidar; the chip system includes one or more interface circuits and one or more processors; the interface circuit and the processor are interconnected by wires; the processing The device is used to execute the method in the first aspect or any possible implementation of the first aspect.
  • the method and device for target recognition convert the laser point cloud into a grid by using the laser point cloud obtained by a single laser radar single frame detection, and use the geometric characteristics of the point cloud grid instead of the lidar reflection points
  • the intensity characteristics of the laser point cloud determine whether the laser point cloud is a false target caused by splashing water, road dust or automobile exhaust caused by road surface water, which can effectively reduce the false detection rate and increase the detection rate of false targets.
  • the applicability and compatibility of the brand's lidar are relatively good, and the method is simple to implement and easy to implement. Moreover, since a single laser radar is used for single frame detection, the calculation delay can be reduced.
  • FIG. 1 is a schematic diagram of the basic working principle of lidar.
  • Fig. 2 is a schematic flowchart of an example of a target recognition method provided by an embodiment of the present application.
  • Fig. 3 is a schematic diagram of an example of a laser radar wire harness provided by an embodiment of the present application.
  • FIG. 4 is a schematic diagram of an example of converting a target laser point cloud into a point cloud grid provided by an embodiment of the present application.
  • Fig. 5 is a schematic diagram of an example of a point cloud grid provided by an embodiment of the present application.
  • Fig. 6 is a schematic diagram of converting three consecutive effective points in a row of a point cloud grid to an x-o-y plane provided by an embodiment of the present application.
  • Fig. 7 is a schematic diagram of transforming all valid points in a row of a point cloud grid to an x-o-y plane provided by an embodiment of the present application.
  • Fig. 8 is a schematic diagram of another example of a point cloud grid provided by an embodiment of the present application.
  • Fig. 9 is a schematic block diagram of a device for target recognition provided in an embodiment of the present application.
  • FIG. 10 is a schematic block diagram of another example of a device for target recognition according to an embodiment of the present application.
  • lidar is one of the indispensable sensors.
  • Lidar is an active sensor that emits a laser beam to detect the position of a target. Its working principle is to send a detection signal (or laser beam) to the target, and then combine the time stamp of the received signal reflected from the target (or target echo) with the time stamp of the transmitted signal Compare. Based on the speed of light, the distance information of the target can be obtained, so as to realize the detection, tracking and identification of the target.
  • lidar is composed of a laser transmitter, an optical receiver, a turntable, and an information processing system.
  • the laser transmitter turns the electric pulse into a light pulse and emits it
  • the optical receiver restores the light pulse reflected from the target into an electric pulse and sends it to the information processing system.
  • lidar usually undertakes tasks such as target detection and target tracking.
  • Lidar scans to obtain the original laser point cloud.
  • the ground processing is first required to remove the ground points, and then the ground point cloud is clustered.
  • the point cloud can be understood as: a collection of point data on the appearance surface of a product obtained by a measuring instrument, which is called a point cloud.
  • the laser point cloud can be understood as: when a beam of laser light hits the surface of an object, the reflected laser light will carry information such as orientation and distance.
  • the laser point cloud is a three-dimensional description of the surrounding environment obtained by the laser radar by scanning the surrounding environment.
  • lidar uses infrared laser ranging method, it is affected by road dust, vehicle exhaust, and water mist reflection from water pavement. It is easy to produce a large number of false reflection points. After clustering, a large number of false reflection points are generated. Fake target point cloud affects follow-up tracking and path planning.
  • the above method uses the intensity information of the lidar.
  • the intensity of the lidar is a measurement index that reflects the intensity of the lidar pulse echo generated at a certain point (collected for each laser reflection point).
  • the intensity value is to a certain extent
  • the upper is equal to the reflectivity of the object scanned by the lidar pulse, which is a function of the wavelength used by the lidar (usually in the near-infrared band). Since the intensity values of lidars of different manufacturers are different, the applicability of lidars of different manufacturers is limited.
  • the algorithm uses continuous frames, so there is a delay in raindrop detection, and the false targets caused by raindrops cannot be removed in the first time, so the adverse effects on subsequent tracking cannot be completely avoided.
  • this method requires a separate establishment of the data structure of the grayscale image, and there is a waste of computing power. Moreover, because the intensity information of the lidar is used, only raindrops with extremely low reflection intensity can be removed, but false targets caused by automobile exhaust and fly dust on the road cannot be removed.
  • the algorithm requires two lidars, the hardware cost is relatively high.
  • the point cloud data of the two lidars requires precise time synchronization and spatial consistency, so the algorithm is relatively complicated, and the calibration accuracy and the measurement accuracy of each lidar are extremely demanding.
  • the algorithm only mentions the removal of false reflection points caused by raindrops, and it cannot effectively remove the false targets caused by road dust and automobile exhaust.
  • this application provides a method of target recognition, by converting the laser point cloud obtained by a single laser radar single frame detection into a grid, and using the geometric characteristics of the grid instead of the intensity characteristics of the laser point cloud to determine the laser Whether the point cloud is a false target caused by splashing water, road dust or automobile exhaust caused by road surface water, it can effectively reduce the false detection rate and increase the detection rate of false targets. Moreover, it can be used for different models and brands of lidars. The applicability and compatibility are relatively good, and the method is simple to implement and easy to implement.
  • the target recognition method provided in this application can be applied in the field of autonomous driving.
  • self-driving mobile devices, etc. mobile devices may include vehicles, airplanes, unmanned aerial vehicles, ships, and other devices that can move spatial positions or change spatial shapes through human operations.
  • This application does not limit the specific form of the mobile device.
  • FIG. 2 is a schematic flowchart of a target recognition method 200 according to an embodiment of the present application.
  • the execution subject of the target recognition method provided in this application may be a target recognition device integrated on a mobile device.
  • it may also be a chip, a chip system, or an integrated circuit on a mobile device, or it may be a laser radar, which includes the above-mentioned chip, chip system, or integrated circuit.
  • This application is not restricted here.
  • the method 200 includes:
  • S210 Receive a target laser point cloud, and generate a point cloud grid corresponding to the target laser point cloud, where a point in the target laser point cloud corresponds to an effective point in the point cloud grid.
  • S220 Determine whether the target laser point cloud is a noisy point cloud according to the geometric characteristics of the point cloud grid.
  • the geometric characteristics of the point cloud grid include: the microscopic roughness of the point cloud grid, and/or the point The discontinuity of the point cloud distribution in the cloud grid.
  • the lidar first scans the original laser point cloud.
  • the original laser point cloud is obtained by a single laser radar single frame detection.
  • the ground points can be removed first, and then the point cloud after going to the ground can be clustered to obtain the target laser point cloud (or it can also become the target point cloud).
  • the target point cloud may include false target point clouds caused by splashing water, road dust, or automobile exhaust caused by road surface water, it is necessary to remove these unreasonable false target point clouds for the target point cloud.
  • a point cloud mesh corresponding to the target laser point cloud can be generated.
  • a point in the target laser point cloud corresponds to a valid point in the point cloud grid.
  • the effective point in the point cloud grid can be understood as a return point, that is, the point reflected from the target received by the lidar.
  • the targets here include: false targets caused by car exhaust on the road, splashing water, flying dust, and false targets with clutter on the roadside bushes, etc., as well as obstacles next to the vehicle (such as pedestrians, roadside buildings, etc.), etc. It can be effectively used as an effective target for vehicle follow-up tracking and path planning.
  • the noise point cloud refers to the laser point cloud of the target laser point cloud that does not represent an effective target during the driving of the vehicle, and the noise point cloud is not processed in the subsequent tracking and path planning of the vehicle.
  • Noise point cloud can also be called false target laser point cloud.
  • Noise point cloud can be understood as the target caused by splashing water, road dust or car exhaust, and roadside bushes and other surface messy false targets caused by road surface water. Point cloud.
  • the laser point cloud caused by these false targets will affect the follow-up tracking and path planning, etc. Therefore, the noise point cloud caused by these false targets needs to be removed or marked as invalid. In the subsequent processing, the noise point cloud will not be processed .
  • the geometric characteristics of the point cloud grid include: the microscopic roughness of the point cloud grid, and/or the discontinuity of the point cloud distribution in the point cloud grid.
  • the micro-roughness of the point cloud grid can be understood as: the degree of sudden change or the size of the difference between the effective points in the point cloud grid.
  • the discontinuity of the point cloud distribution in the point cloud grid can be understood as: the degree of discontinuity between the effective points in the point cloud grid, and further, the geometric characteristics of the point cloud grid can be used to identify or determine the target laser Whether the point cloud is a noise point cloud. For example, the greater the microscopic roughness of the point cloud grid, the more likely the laser point cloud is to be a noisy point cloud. The greater the discontinuity of the point cloud distribution in the point cloud grid, the more likely the laser point cloud is to be a noisy point cloud.
  • step S220 is to identify the target laser point cloud that is not suitable for subsequent processing in the subsequent tracking and path planning of the vehicle. Therefore, the following alternative solutions are also within the scope of step S220:
  • Marking the target laser point cloud as invalid means that the target point cloud will not enter the subsequent tracking and path planning processing of the vehicle.
  • the target laser point cloud is marked as invalid; and/or,
  • the target laser point cloud is marked as invalid.
  • the target recognition method provided in this application converts the laser point cloud into a grid by using the laser point cloud obtained from a single laser radar single frame detection, and uses the geometric characteristics of the point cloud grid instead of the intensity characteristics of the lidar reflection points to determine Whether the laser point cloud is a false target caused by splashing water, road dust or automobile exhaust caused by road surface water can effectively reduce the false detection rate and increase the false target detection rate.
  • the method is simple to implement and easy to implement. Moreover, since a single laser radar is used for single frame detection, the calculation delay can be reduced.
  • Fig. 3 shows a schematic diagram of the wire harness of the lidar.
  • the lidar used in autonomous driving generally has multiple laser beams (Beam). For example, there are 16 lines, 32 lines, 64 lines and so on.
  • Each wire harness has a serial number, from bottom to top, from 0 to wire number -1.
  • the sequence numbers can also be remapped when the point cloud grid is generated to meet the order. All the wire beams will be scanned synchronously around the vertical axis (Z axis) of the lidar.
  • the swept angle can be called the azimuth angle ⁇ , and the sample is sampled every time a certain angle is swept, and the certain angle swept can be called the horizontal angle.
  • Resolution ⁇ h Resolution
  • Figure a in Figure 4 shows an example of a schematic diagram of a laser point cloud in an XYZ coordinate system. Take the black dots shown in Figure a in Figure 4 as an example. According to its coordinates on the XOY plane, the azimuth angle ⁇ can be calculated, divided by the horizontal angular resolution ⁇ h , and then rounded up to obtain the column coordinates of the point in the grid (for example, 12). If the point is obtained by scanning the line beam with serial number 6, then its row coordinate in the grid is 6. If the row coordinates of the points in the target point cloud are between 5 and 8, and the column coordinates are between 11-15, the target point cloud can be transformed into a grid with 4 rows and 5 columns.
  • Figure b in Figure 4 shows a schematic diagram of the point cloud grid obtained from Figure a.
  • the column coordinates in the grid corresponding to any point in the laser point cloud can be calculated by formula (1), and the row coordinates (row) are the laser beam channel serial numbers of the point.
  • ⁇ is the azimuth angle of the point
  • ⁇ h is the horizontal angular resolution of the point.
  • the target laser point cloud can be transformed into a point cloud grid.
  • the point cloud grid shown in figure b in Figure 4 includes valid points and invalid points.
  • the valid points can also be called return points.
  • the invalid points indicate that the laser beam emitted by the laser radar has not received reflection from the target.
  • the returned point is equivalent to the point in the point cloud grid without the laser point cloud, but only occupies a position in the point cloud grid.
  • determining whether the target laser point cloud is a noise point cloud may specifically include:
  • micro-roughness of the point cloud grid is greater than or equal to the first threshold, judge or determine that the target laser point cloud is a noise point cloud; and/or,
  • the target laser point cloud is a noise point cloud.
  • the following three possible methods can be used to judge or determine whether the target laser point cloud is a noise point cloud according to the geometric characteristics of the point cloud grid.
  • One possible implementation is to compare the micro-roughness of the point cloud grid with a preset threshold (first threshold), for example, if the micro-roughness of the point cloud grid is greater than or equal to the first threshold ( Or greater than the first threshold), it is determined that the target laser point cloud is a noise point cloud. If the microscopic roughness of the point cloud grid is less than the first threshold (or, less than or the first threshold), it is determined that the target laser point cloud is not a noise point cloud, that is, the target laser point cloud can be used for subsequent vehicle tracking And the target point cloud for path planning.
  • first threshold a preset threshold
  • Another possible implementation is to compare the distribution discontinuity of the point cloud in the point cloud grid with a preset threshold (the second threshold). For example, if the distribution discontinuity of the point cloud in the point cloud grid is greater than or equal to the second threshold (or greater than the second threshold), it is determined that the target laser point cloud is a noise point cloud. If the discontinuity of the point cloud distribution in the point cloud grid is less than the second threshold (or, less than or equal to the second threshold), it is determined that the target laser point cloud is not a noise point cloud, that is, the target laser point cloud can be used for Target point cloud for vehicle follow-up tracking and path planning.
  • the second threshold a preset threshold
  • Another possible implementation is to combine the micro-roughness and distribution discontinuity of the point cloud grid to make judgments. For example, if the micro-roughness of the point cloud grid is greater than or equal to the first threshold (or greater than the first threshold), and the distribution discontinuity of the point cloud in the point cloud grid is greater than or equal to the second threshold (or greater than The second threshold), it is determined that the target laser point cloud is a noise point cloud. If, if the micro-roughness of the point cloud grid is less than the first threshold (or, less than or the first threshold), or the discontinuity of the point cloud distribution in the point cloud grid is less than the second threshold (or, less than or Equal to the second threshold), it is determined that the target laser point cloud is not a noise point cloud.
  • the above three methods are only exemplary to illustrate the use of the geometric characteristics of the point cloud grid in the embodiment of the present application to determine whether the target laser point cloud is a noise point cloud.
  • the geometric features of the point cloud grid may also include geometric features of other dimensions or types.
  • the point cloud grid can also be used according to other methods or rules. The microscopic roughness, and/or, the discontinuity of the distribution is judged.
  • the embodiments of the application are not limited here.
  • the micro-roughness of the point cloud grid may be: The ratio of the total number of mutation points to the effective points included in the point cloud grid.
  • the total number of mutation points in the point cloud grid is the sum of the number of mutation points in each row of the point cloud grid.
  • the mutation points in the point cloud grid can be understood as part of the effective points included in the point cloud grid, and the mutation degree or difference between this part of the effective points and other effective points is relatively large. In other words, the mutation point can be a subset of the valid points in the point cloud grid.
  • the total number of mutation points in the point cloud grid may be the sum of the number of mutation points in each row of the point cloud grid.
  • the number of mutation points in the point cloud grid can be evaluated row by row, and finally the total number of mutation points in the point cloud grid can be obtained.
  • the distribution discontinuity of the point cloud in the point cloud grid is: the sum of the discontinuities of each column of the point cloud in the point cloud grid .
  • the discontinuity of the effective point cloud in each column of the point cloud grid can be evaluated column by column, and the discontinuities of all the columns can be added together to finally get the discontinuity of the point cloud distribution in the point cloud grid Spend.
  • the micro-roughness of the point cloud grid may also be determined in other ways. For example, evaluate the micro-roughness of the effective point cloud in each column of the point cloud grid column by column (for example, determine the number of abrupt points in each column in the point cloud grid), and compare the effective point cloud in each column The total number of mutation points in the point cloud grid is added to obtain the total number of mutation points in the point cloud grid, and finally the ratio of the total number of mutation points in the point cloud grid to the effective points included in the point cloud grid is obtained, and the The micro-roughness of the point cloud grid.
  • the discontinuity in the point cloud grid is determined row by row, and the discontinuities of each row are added together to finally get the midpoint of the point cloud grid.
  • Discontinuity of cloud distribution can also be used to determine the micro-roughness and distribution discontinuity of the point cloud grid. This application does not limit the specific method or algorithm for determining the micro-roughness and distribution discontinuity of the point cloud grid.
  • the total number of mutation points in the point cloud grid may be the sum of the number of mutation points in each row of the point cloud grid.
  • the number of effective points in the first row of point cloud is equal to 3, then the three effective points are three consecutive effective points by default, assuming these three consecutive The effective points are the first effective point, the second effective point, and the third effective point. If the distance between the first effective point and the third effective point and the second effective point is greater than or equal to the preset threshold, then It is determined that the second effective point is a sudden change point; if the distance between the line connecting the first effective point and the third effective point and the second effective point is less than the preset threshold, the second effective point is not a sudden change point. The following will be explained in detail with examples.
  • the second effective point Q is a sudden change point, if and The difference in length is less than the preset threshold, then the second effective point Q is not a sudden change point.
  • the effective point Q of the second point in the first row is a mutation point, if it is, the number of mutation points in the first row is 1, if not, the number of mutation points in the first row is 0 . Since there are only three valid points in the first row, the number of mutations in the first row is at most one.
  • the third row (the row coordinate in the point cloud grid is 6), since the number of valid points in the third row is 4, take three each time and use the same method as the first row for judgment. For example, suppose that the 4 valid points in the third row are the fourth valid point (the coordinates in the point cloud grid are (6,11)) and the fifth valid point (the coordinates in the point cloud grid are (6,12)), the sixth valid point (the coordinate in the point cloud grid is (6,13)), the seventh valid point (the coordinate in the point cloud grid is (6,15)). First take the fourth effective point, the fifth effective point, and the sixth effective point. Use the same method as the first row to determine whether the fifth effective point is a mutation point.
  • the number of mutation points in the third row can be determined. Since there are only four effective points in the third row, the number of mutation points in the third row may be: 2 (the fifth effective point and the sixth effective point), 1 (the fifth effective point or the first Six effective points), 0.
  • the second valid point of the three valid points is judged according to the same method as the first row (The coordinates in the point cloud grid are (5,13) whether it is a sudden change point, if it is, the number of sudden change points in the fourth row is 1, if not, the number of sudden changes in the fourth row is 0. Because of the first There are only three valid points in the four rows, so the number of mutations in the fourth row is at most one.
  • the number of valid points in the i-th row of the point cloud in the point cloud grid is less than or equal to 2, it is determined that the number of mutation points in the i-th row in the point cloud grid is 0, and i is a positive integer; Determine whether the second effective point of every three consecutive effective points in the i-th row of point cloud is a sudden change point, and determine the number of sudden changes in the i-th row of the point cloud grid.
  • the number of mutation points in each row can be determined, and then the number of mutation points in each row can be added to obtain the total number of mutation points in the point cloud grid, and then the total number of mutation points in the point cloud grid can be calculated as the total number of mutation points in the point cloud grid.
  • the ratio of all the effective points included is the micro-roughness of the point cloud grid.
  • the total number of mutation points in the point cloud grid can be the sum of the number of mutation points in each row of the point cloud grid.
  • all valid points in the row can be projected onto the X-O-Y plane. Determine the third effective point with the longest distance from the first line according to the first line connecting the first effective point with the largest azimuth angle and the second effective point with the smallest azimuth angle above the X-O-Y plane;
  • the third effective point is determined to be the first sudden change point, and the first effective point, Among the effective points other than the second effective point and the third effective point, the second line to the first effective point and the third effective point, and the second line to the second effective point and the third effective point are respectively determined.
  • the fourth effective point and the fifth effective point with the furthest distance between the third link if the distance between the fourth effective point and the second link is greater than or equal to the third threshold (or other preset thresholds), Then the number of mutation points in the line of point cloud increases by one, and if the distance between the fifth effective point and the third line is greater than or equal to the third threshold (or other preset thresholds), then the line of points
  • the number of mutation points in the cloud increases by one, that is, every time a new effective point is determined, two new straight lines are added. Repeat the above process until the remaining valid points that are not used in the point cloud grid are 0, that is, no new straight lines are generated. Accumulation can get the total number of mutation points in the line of point cloud.
  • the total number of mutation points in the row of point cloud is zero.
  • the first line can determine the first line The effective point C with the farthest distance. Since the effective point C is determined based on the effective point A and the effective point B, the effective point C can determine a straight line with the effective point A, and the effective point C can determine a straight line with the effective point B. ,
  • the two straight lines can be equivalent to the first line mentioned above.
  • a method similar to determining the effective point C can be used in the remaining effective points that are not used in the point cloud grid, and a new Valid points and new straight lines until the remaining valid points that are not used in the point cloud grid are 0, that is, no new straight lines are generated.
  • the distance between the newly determined effective point and the straight line generating the newly determined effective point is compared with a preset threshold to determine whether the newly determined effective point is a mutation point,
  • the total number of mutation points in the row of point cloud can be obtained by accumulating the number of mutation points.
  • the sequence of the straight lines generated in sequence can be as shown in Table 1.
  • the sequence number of the generated lines indicates the sequence of the generated lines, and the smaller the number, the earliest generated. For example, the number 1 is generated first, and the number is 2 second. That is, every time a new effective point is determined among the unused effective points, two new straight lines are added.
  • FIG. 7 shows a schematic diagram of projecting all valid points in a certain row of a point cloud grid onto the XOY plane.
  • select the point A with the largest azimuth angle ( ⁇ max ) and the point H with the smallest azimuth angle ( ⁇ min ) connect AH to form a direct line, and then determine the line with the largest distance AH
  • Point C determine the distance from point C to the straight line AH (assumed to be L 1 ), if L 1 is greater than or equal to the preset threshold, it is proved that point C is a sudden change point. If L 1 is less than the preset threshold, it is proved that point C is not a sudden change point, and it can be determined that there is no sudden change point in the point cloud grid.
  • L 1 is greater than or equal to the preset threshold, further, connect AC to form a straight line, and connect CH to form a straight line.
  • the effective points on the XOY plane except for point A, point C, and point H determine and AC The point with the farthest distance from the straight line and the point with the farthest distance from the straight line CH.
  • the point farthest from the AC straight line (assuming L 2 ) is point G
  • the point farthest from the straight line distance of CH (assuming L 3 ) is point D
  • L 3 Whether it exceeds the preset threshold. If both are exceeded, it proves that the G and D points are mutation points.
  • the threshold for comparison with L 2 and the threshold for comparison with L 3 may be the same or different.
  • the threshold value compared with L 2 may be the same as or different from the threshold value compared with L 1.
  • L 2 exceeds the preset threshold and L 3 does not exceed the preset threshold, it proves that point G is also a mutation point, and point D is not a mutation point. Further, continue to connect AG to form a straight line AG, and connect CG to form a straight line CG. Among the effective points on the XOY plane except for points A, C, H, G, and D, continue to determine the distance to the straight line AG. The farthest point and the point farthest from the straight line CG, and then compare the distance with the preset threshold respectively. Repeat the process until all the remaining valid points in the line of point cloud on the plane are covered to stop, and the total number of mutation points in the line of point cloud can be obtained by accumulation.
  • L 2 does not exceed the preset threshold and L 3 exceeds the preset threshold, it proves that point G is not a mutation point, and point D is a mutation point.
  • connect CD to form a straight line CD
  • connect HD to form a straight line HD
  • the farthest point and the point farthest from the straight line HD and then compare the distance with the preset threshold respectively. Repeat the process until all the remaining valid points in the line of point cloud on the plane are covered to stop, and the total number of mutation points in the line of point cloud can be obtained by accumulation.
  • the thresholds used may also be different, or may also be the same.
  • the number of mutation points in each row of the point cloud grid can be determined, and then the number of mutation points in each row can be added to obtain the total number of mutation points in the point cloud grid, and then the point cloud grid can be calculated The ratio of the total number of mutation points to the effective number of points included in the point cloud grid, and the micro-roughness of the point cloud grid is obtained.
  • the discontinuity of the point cloud distribution in the point cloud grid can be the sum of the discontinuities of each column of the point cloud in the point cloud grid.
  • the discontinuity of each column of the point cloud in the point cloud grid may be determined according to the number of valid points included in each column of the point cloud.
  • the discontinuity of the jth column of point cloud is determined to be a, j is a positive integer, and a is a positive number;
  • the discontinuity of the jth column of point cloud is determined Is 0, j is a positive integer;
  • the jth column is determined
  • the discontinuity of the point cloud is b, j is an integer greater than 1, and b is a positive number;
  • the number of valid points in the j-th column of point cloud in the point cloud grid is greater than 1, it is determined whether the row coordinates corresponding to the valid points in the j-th column of point cloud are continuous. The continuity is 0. If it is not continuous, it is determined that the discontinuity of the point cloud in the j-th column is c, where j is an integer greater than 1, and c is a positive number.
  • a is greater than b, and b is greater than c.
  • the effective number of points is 1, and the first column is the first column in the point cloud grid Point cloud, the discontinuity of the first column of point cloud is determined to be 0.
  • the number of effective points is 4, which is greater than 1. It is necessary to further determine the row coordinates of the effective point cloud on the second column (ie the serial number of the laser beam) Whether it is continuous. For the 4 effective points in the second column of point cloud, their row coordinates are 5, 6, 7, 8, respectively, that is, the serial number of the laser beam is continuous, and the discontinuity of the second column of point cloud is 0.
  • the discontinuity is determined to be b , That is, for the third column of point cloud, its discontinuity is b.
  • the discontinuity of the fourth column of point cloud is determined to be a.
  • the number of effective points is 3, and it is necessary to further determine whether the row coordinates (ie, the serial number of the laser beam) of the effective point cloud on the fifth column are continuous.
  • their row coordinates are 5, 6, and 8, respectively, that is, the serial number of the laser beam is discontinuous, and the discontinuity of the fifth column of point cloud is c.
  • the discontinuity of each column of point clouds can be determined separately, and then the discontinuities of all columns are added together, and finally the distribution discontinuity of the point cloud in the point cloud grid is obtained.
  • the discontinuity of the point cloud distribution in the point cloud grid and the micro roughness of the point cloud grid can be determined. Thereby, it is determined whether the target point cloud is a fake target point cloud, and the method is simple to implement and easy to implement.
  • the geometric characteristics of the point cloud grid are used, and the intensity characteristics of the reflected signal of the lidar are not used, and the applicability and compatibility of lidars of different models and brands are better.
  • Using a single laser radar single frame point cloud to obtain the point cloud grid can reduce the calculation delay, and the false detection of false targets in the point cloud is low. For splashing water, road dust or car exhaust caused by road surface water, etc. The false targets caused can be accurately detected, which improves the detection rate of false targets.
  • pre-set and pre-defined can be achieved by pre-saving corresponding codes, tables, or other methods that can be used to indicate related information in devices (for example, including terminals and network devices). To achieve, this application does not limit its specific implementation.
  • FIG. 9 shows a schematic block diagram of a device 300 for target recognition according to an embodiment of the present application.
  • the device 300 may correspond to the lidar described in the above method 200, or may be a chip, component, integrated circuit, or chip applied to the lidar. Chips in in-vehicle processors, etc.
  • each module or unit in the device 300 is respectively used to execute each action or processing procedure performed in the above-mentioned method 200.
  • the device 300 includes a transceiving unit 310 and a processing unit 320, and the transceiving unit 310 is configured to perform specific signal transceiving under the driving of the processing unit 320.
  • the transceiver unit 310 is configured to receive the target laser point cloud
  • the processing unit 320 is configured to generate a point cloud grid corresponding to the target laser point cloud, where a point in the target laser point cloud corresponds to an effective point in the point cloud grid.
  • the processing unit 320 is also used for judging or determining whether the target laser point cloud is a noisy point cloud according to the geometric characteristics of the point cloud grid, and the geometric characteristics of the point cloud grid include: the microscopic roughness of the point cloud grid Degree, and/or, the degree of discontinuity of the point cloud distribution in the point cloud grid.
  • the target recognition device converts the laser point cloud into a grid by using the laser point cloud obtained by a single laser radar single frame detection, and uses the geometric characteristics of the point cloud grid instead of the intensity characteristics of the lidar reflection points to determine Whether the laser point cloud is a false target caused by splashing water, road dust or automobile exhaust caused by road surface water can effectively reduce the false detection rate and increase the detection rate of false targets.
  • a single laser radar is used for single frame detection, the applicability and compatibility of laser radars of different models and brands are relatively good, and the method is simple to implement and easy to implement.
  • the micro-roughness of the point cloud grid is: the ratio of the total number of abrupt points in the point cloud grid to the number of effective points included in the point cloud grid, where: The total number of mutation points in the point cloud grid is the sum of the number of mutation points in each row of the point cloud grid.
  • the distribution discontinuity of the point cloud in the point cloud grid is: the sum of the discontinuities of each column of the point cloud in the point cloud grid.
  • the processing unit 320 is further configured to: when the micro-roughness of the point cloud grid is greater than or equal to a first threshold, determine whether the target laser point cloud is noise Point cloud; and/or, in the case where the distribution discontinuity of the point cloud in the point cloud grid is greater than or equal to the second threshold, it is determined that the target laser point cloud is a noise point cloud.
  • the processing unit 320 is further configured to:
  • the number of effective points in the i-th row of the point cloud in the point cloud grid is less than or equal to 2, it is determined that the number of mutation points in the i-th row in the point cloud grid is 0, and i is a positive integer;
  • the number of effective points in the i-th row of point cloud in the point cloud grid is greater than or equal to 3, then three consecutive effective points in the i-th row of point cloud determine the second of the three consecutive effective points Whether the effective point is a mutation point, i is a positive integer;
  • the processing unit 320 is further configured to:
  • the three consecutive effective points in the i-th row of point cloud are the first effective point, the second effective point, and the third effective point. If the distance between the line connecting the first effective point and the third effective point and the second effective point is greater than or equal to a preset threshold, determining that the second effective point is a sudden change point;
  • the second effective point is not a sudden change point.
  • the processing unit 320 is further configured to:
  • the three consecutive effective points in the point cloud of the i-th row are the first effective point P, the second effective point Q, and the third effective point S, and O is the origin of the coordinate.
  • the reference point corresponding to the second effective point is Q 1 ,
  • the processing unit 320 is also used to: and In the case that the difference in length is greater than or equal to the preset threshold, the second effective point Q is determined to be a sudden change point. and When the difference in length is less than the preset threshold, it is determined that the second effective point Q is not a sudden change point.
  • the processing unit 320 is further configured to:
  • the third effective point is determined to be the first sudden change point, and the first effective point and the second effective point are divided on the plane.
  • the effective points other than the third effective point and the third effective point respectively determine the second line between the first effective point and the third effective point, and the second line between the second effective point and the third effective point.
  • the fourth effective point and the fifth effective point with the furthest distance between the three lines if the distance between the fourth effective point and the second line is greater than or equal to the third threshold, then the sudden change point in the i-th row of point cloud The number increases by 1.
  • the number of mutation points in the i-th row of point cloud increases by one; repeat the above process until Until no new straight line is generated on the plane, the total number of mutation points in the line of point cloud can be obtained by accumulation.
  • the number of mutation points in the i-th row of point cloud is zero.
  • the discontinuity of each column of point clouds in the point cloud grid is determined according to the number of valid points included in each column of point clouds.
  • the processing unit 320 is further configured to:
  • the number of valid points in the j-th column of point cloud in the point cloud grid is greater than 1, it is determined whether the row coordinates corresponding to the valid points in the j-th column of point cloud are continuous. The degree of continuity is 0. If it is not continuous, the discontinuity of the point cloud in the j-th column is determined to be c, where j is an integer greater than 1, and c is a positive number.
  • the target laser point cloud is obtained by single-frame detection of a single laser radar.
  • the false target laser spot is caused by one or more of water splash caused by road surface water, automobile exhaust, or road dust.
  • the transceiving unit 310 may include a receiving unit (module) and a sending unit (module), configured to perform the steps of receiving information and sending information in each embodiment of the aforementioned method 200.
  • the device 300 can also be the storage unit (module).
  • the transceiving unit 310 may be a transceiver, an input/output interface, or an interface circuit.
  • the storage module is used to store instructions executed by the transceiver unit 310 and the processing unit 320.
  • the transceiving unit 310, the processing unit 310, and the storage unit are coupled with each other.
  • the storage unit stores instructions, the processing unit 320 is used to execute the instructions stored in the storage unit, and the transceiving unit 310 is used to perform specific signal transceiving under the driving of the processing unit 320.
  • the transceiving unit 310 may be a transceiver, an input/output interface, or an interface circuit.
  • the storage unit may be a memory.
  • the processing unit 320 may be implemented by a processor. As shown in FIG. 10, the device 400 for target recognition may include a processor 410, a memory 420, and a transceiver 430.
  • the device 300 for target recognition shown in FIG. 9 or the device 400 for target recognition shown in FIG. 10 can implement the embodiment in the aforementioned method 200 and the steps performed in the embodiment shown in FIG. 2 to FIG. 8. For similar descriptions, reference can be made to the descriptions in the aforementioned corresponding methods. To avoid repetition, I won’t repeat them here.
  • each unit in the device can be all implemented in the form of software called by processing elements; they can also be all implemented in the form of hardware; part of the units can also be implemented in the form of software called by the processing elements, and some of the units can be implemented in the form of hardware.
  • each unit can be a separate processing element, or it can be integrated in a certain chip of the device for implementation.
  • it can also be stored in the memory in the form of a program, which is called and executed by a certain processing element of the device. Function.
  • the processing element may also be called a processor, and may be an integrated circuit with signal processing capability.
  • each step of the above method or each of the above units may be implemented by an integrated logic circuit of hardware in a processor element or implemented in a form of being called by software through a processing element.
  • the unit in any of the above devices may be one or more integrated circuits configured to implement the above methods, for example: one or more application specific integrated circuits (ASIC), or, one or Multiple digital signal processors (digital signal processors, DSP), or, one or more field programmable gate arrays (FPGA), or a combination of at least two of these integrated circuits.
  • ASIC application specific integrated circuits
  • DSP digital signal processors
  • FPGA field programmable gate arrays
  • the unit in the device can be implemented in the form of a processing element scheduler
  • the processing element can be a general-purpose processor, such as a central processing unit (CPU) or other processors that can call programs.
  • CPU central processing unit
  • these units can be integrated together and implemented in the form of a system-on-a-chip (SOC).
  • the processor may be a central processing unit (CPU), and the processor may also be other general-purpose processors, digital signal processors (digital signal processors, DSPs), and application-specific integrated circuits. (application specific integrated circuit, ASIC), ready-made programmable gate array (field programmable gate array, FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc.
  • the general-purpose processor may be a microprocessor or the processor may also be any conventional processor or the like.
  • the memory in the embodiments of the present application may be volatile memory or non-volatile memory, or may include both volatile and non-volatile memory.
  • the non-volatile memory can be read-only memory (ROM), programmable read-only memory (programmable ROM, PROM), erasable programmable read-only memory (erasable PROM, EPROM), and electrically available Erase programmable read-only memory (electrically EPROM, EEPROM) or flash memory.
  • the volatile memory may be random access memory (RAM), which is used as an external cache.
  • RAM random access memory
  • static random access memory static random access memory
  • DRAM dynamic random access memory
  • DRAM synchronous dynamic random access memory
  • Access memory synchronous DRAM, SDRAM
  • double data rate synchronous dynamic random access memory double data rate SDRAM, DDR SDRAM
  • enhanced synchronous dynamic random access memory enhanced SDRAM, ESDRAM
  • synchronous connection dynamic random access memory Take memory (synchlink DRAM, SLDRAM) and direct memory bus random access memory (direct rambus RAM, DR RAM).
  • the above-mentioned embodiments may be implemented in whole or in part by software, hardware, firmware or any other combination.
  • the above-mentioned embodiments may be implemented in the form of a computer program product in whole or in part.
  • the computer program product includes one or more computer instructions or computer programs.
  • the computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices.
  • the computer instruction may be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium.
  • the computer instruction may be transmitted from a website, computer, server, or data center through a cable (For example, infrared, wireless, microwave, etc.) to transmit to another website, computer, server or data center.
  • the computer-readable storage medium may be any available medium that can be accessed by a computer or a data storage device such as a server or a data center that includes one or more sets of available media.
  • the usable medium may be a magnetic medium (for example, a floppy disk, a hard disk, a magnetic tape), an optical medium (for example, a DVD), or a semiconductor medium.
  • the semiconductor medium may be a solid state drive.
  • the embodiment of the present application also provides a computer-readable medium for storing computer program code, and the computer program includes instructions for executing the target recognition method of the embodiment of the present application in the above method 200.
  • the readable medium may be a read-only memory (ROM) or a random access memory (RAM), which is not limited in the embodiment of the present application.
  • the present application also provides a computer program product.
  • the computer program product includes instructions. When the instructions are executed, the device identified by the target is caused to perform operations corresponding to the foregoing method 200, respectively.
  • the embodiment of the present application also provides a chip system, which is applied to lidar.
  • the chip system includes one or more interface circuits and one or more processors; the interface circuit and the processor are interconnected by wires; the processor is used to execute the target recognition method in the embodiment of the present application in the method 200 described above.
  • An embodiment of the present application also provides a chip, which includes a processor, configured to call and run a computer program from a memory, so that a device installed with the chip executes the target recognition method in the above method 200.
  • any of the device for target recognition provided in the foregoing embodiments of the present application may include the system chip.
  • the computer instructions are stored in a storage unit.
  • the storage unit is a storage unit in the chip, such as a register, a cache, etc.
  • the storage unit can also be a storage unit in the terminal located outside the chip, such as a ROM or other storage units that can store static information and instructions.
  • static storage devices RAM, etc.
  • the processor mentioned in any of the above may be a CPU, a microprocessor, an ASIC, or one or more integrated circuits used to control the program execution of the above-mentioned target recognition method.
  • the processing unit and the storage unit can be decoupled, respectively set on different physical devices, and connected in a wired or wireless manner to realize the respective functions of the processing unit and the storage unit, so as to support the system chip to implement the above-mentioned embodiments Various functions in.
  • the processing unit and the memory may also be coupled to the same device.
  • the application also provides a laser radar, which includes any one of the target recognition devices provided in the embodiments of the application.
  • the lidar includes any chip system or chip provided in this application.
  • the memory in the embodiments of the present application may be a volatile memory or a non-volatile memory, or may include both volatile and non-volatile memory.
  • the non-volatile memory can be read-only memory (ROM), programmable read-only memory (programmable ROM, PROM), erasable programmable read-only memory (erasable PROM, EPROM), and electrically available Erase programmable read-only memory (electrically EPROM, EEPROM) or flash memory.
  • the volatile memory may be random access memory (RAM), which is used as an external cache.
  • RAM random access memory
  • static random access memory static random access memory
  • DRAM dynamic random access memory
  • DRAM synchronous dynamic random access memory
  • Access memory synchronous DRAM, SDRAM
  • double data rate synchronous dynamic random access memory double data rate SDRAM, DDR SDRAM
  • enhanced synchronous dynamic random access memory enhanced SDRAM, ESDRAM
  • synchronous connection dynamic random access memory Take memory (synchlink DRAM, SLDRAM) and direct memory bus random access memory (direct rambus RAM, DR RAM).
  • the disclosed system, device, and method can be implemented in other ways.
  • the device embodiments described above are only illustrative.
  • the division of the unit is only a logical function division, and there may be other divisions in actual implementation, for example, multiple units or components may be combined or may be Integrate into another system, or some features can be ignored or not implemented.
  • the displayed or discussed mutual coupling or direct coupling or communication connection may be indirect coupling or communication connection through some interfaces, devices or units, and may be in electrical, mechanical or other forms.
  • the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the objectives of the solutions of the embodiments.
  • the functional units in the various embodiments of the present application may be integrated into one processing unit, or each unit may exist alone physically, or two or more units may be integrated into one unit.
  • the function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer readable storage medium.
  • the technical solution of the present application essentially or the part that contributes to the existing technology or the part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including Several instructions are used to make a computer device (which may be a personal computer, a server, or a network device, etc.) execute all or part of the steps of the methods described in the various embodiments of the present application.
  • the aforementioned storage media include: U disk, mobile hard disk, read-only memory (ROM), and random access.

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Radar, Positioning & Navigation (AREA)
  • Remote Sensing (AREA)
  • Electromagnetism (AREA)
  • Multimedia (AREA)
  • Theoretical Computer Science (AREA)
  • Optical Radar Systems And Details Thereof (AREA)

Abstract

一种目标识别的方法和装置，可以应用在自动驾驶或者智能驾驶领域。方法包括：接收目标激光点，生成目标激光点云对应的网格，目标激光点云中的一个点对应点云网格中的一个有效点（S210）；根据点云网格的几何特征，判断目标激光点云是否为噪声点云，点云网格的几何特征包括：点云网格的微观粗糙度和/或点云网格中点云的分布不连续度（S220）。提供的目标识别的方法，可以确定激光点云是否是由于路面积水导致飞溅水花、路面飞尘或者汽车尾气等造成的假目标，有效的降低误检率，提高假目标的检出率，方法实现简单，便于实施。

Description

一种目标识别的方法和装置 技术领域
本申请涉及自动驾驶领域，更为具体的，涉及一种目标识别的方法和装置。
背景技术
对于自动驾驶技术而言，激光雷达是车辆上主要配备的视觉传感器之一。激光雷达，是通过发射激光束，用以探测目标位置的主动传感器，根据从目标反射回来的激光点云可获得目标的距离信息，从而实现对目标进行探测、跟踪和识别。由于激光雷达采用红外激光测距法，受到路面飞尘、车辆尾气以及积水路面溅起的水雾反射的影响，容易产生大量的假反射点，这些假反射点会造成识别误差，从而影响后续的跟踪路径规划等。
目前，对于由于对路面飞尘和汽车尾气等造成的假发射点(或者可以称为假目标点云)，现有的去除激光点云中的假反射点的方法中还没有非常适合的解决方案，并且，现有的方法大多比较复杂，实施的成本较高，导致激光点云还存在较多的假目标，严重的降低了自动驾驶的精度和安全性。
发明内容
本申请提供了一种目标识别的方法和装置，可以有效的降低激光点云的误检率，提高激光点云中假目标的检出率，方法实现简单，便于实施。
第一方面，提供了一种目标识别的方法，该方法的执行主体可以是集成在移动设备上的目标识别的装置。或者，该方法的执行主体还可以是移动设备上的芯片、芯片系统或者集成电路等，或者，该方法的执行主体还可以是激光雷达。该方法包括：接收目标激光点云，生成该目标激光点云对应的点云网格，其中，该目标激光点云中的一个点对应该点云网格中的一个有效点；根据该点云网格的几何特征，判断该目标激光点云是否为噪声点云，该点云网格的几何特征包括：该点云网格的微观粗糙度，和/或，该点云网格中点云的分布不连续度。其中，移动设备可以包括车辆、飞机、无人机、轮船等可以通过人的操作移动空间位置或改变空间形状的设备。
第一方面提供的目标识别的方法，通过利用单个激光雷达单帧检测获得的激光点云，将激光点云转化为网格，利用点云网格的几何特征而非激光雷达反射点的强度特征识别该激光点云是否是由于路面积水导致飞溅水花、路面飞尘或者汽车尾气等造成的假目标，可以有效的降低误检率，提高假目标的检出率，方法实现简单，便于实施。并且，由于是利用单个激光雷达单帧检测，可以降低了计算延迟。
在第一方面一种可能的实现方式中，该点云网格的微观粗糙度为：该点云网格中的突变点总数占该点云网格中包括的有效点数的比例，其中，该点云网格中的突变点总数为该点云网格中每一行中的突变点数之和。在该实现方式中，可以较为快速简单的确定点云网格的微观粗糙度，容易实现，精度较高，复杂度较低。
在第一方面一种可能的实现方式中，该点云网格中点云的分布不连续度为：该点云网 格中每一列点云的不连续度之和。在该实现方式中，可以较为快速简单的确定点云网格中点云的分布不连续度，容易实现，精度较高，复杂度较低。
在第一方面一种可能的实现方式中，判断(或者确定)该目标激光点云是否为噪声点云，包括：
若该点云网格的微观粗糙度大于或者等于第一阈值，判断该目标激光点云是为噪声点云；和/或，
若该点云网格中点云的分布不连续度大于或者等于第二阈值，判断该目标激光点云是为噪声点云。
在该实现方式中，通过将点云网格的微观粗糙度和预设的比较，和/或，将该点云网格中点云的分布不连续度与预设的阈值进行比较，判断该激光点云是否为噪声点云，便于实现，复杂度低。
在第一方面一种可能的实现方式中，该方法还包括：
若该点云网格中第i行点云中有效点数小于或者等于2，则确定该点云网格中第i行的突变点数为0，i为正整数；
若该点云网格中第i行点云中有效点数大于或者等于3，则依次取该第i行点云中连续的三个有效点确定该连续的三个有效点中的第二有效点是否为突变点，i为正整数；
根据该第i行点云中每连续的三个有效点中的第二个有效点是否为突变点，确定该点云网格中第i行的突变点数。
在该实现方式中，通过上述方法确定每一行点云的突变点数，便于实现，并且，结果的精度较高，可以更好的反映出每一行点云的突变情况。
在第一方面一种可能的实现方式中，若该点云网格中第i行点云中有效点数大于或者等于3，该第i行点云中连续的三个有效点分别为第一有效点、第二有效点、第三有效点，若该第一有效点与该第三有效点的连线与该第二有效点之间的距离大于或者等于预设的阈值，则确定该第二有效点为突变点；
若该第一有效点与该第三有效点的连线与该第二有效点之间的距离小于预设的阈值，则确定该第二有效点不是突变点。
在第一方面一种可能的实现方式中，该第i行点云中连续的三个有效点分别为第一有效点P、第二有效点Q、第三有效点S，O为坐标原点，
Figure PCTCN2020084818-appb-000001
与
Figure PCTCN2020084818-appb-000002
的夹角为β，
Figure PCTCN2020084818-appb-000003
与
Figure PCTCN2020084818-appb-000004
的夹角为θ，
Figure PCTCN2020084818-appb-000005
与
Figure PCTCN2020084818-appb-000006
的夹角为α，α＝180 °-θ-β，
Figure PCTCN2020084818-appb-000007
与第二有效点Q对应的参考点为Q 1，
Figure PCTCN2020084818-appb-000008
若
Figure PCTCN2020084818-appb-000009
与
Figure PCTCN2020084818-appb-000010
长度的差值大于或者等于预设的阈值，则该第二有效点Q为突变点，若
Figure PCTCN2020084818-appb-000011
与
Figure PCTCN2020084818-appb-000012
长度的差值小于预设的阈值，则该第二有效点Q不是突变点。
在第一方面一种可能的实现方式中，该方法还包括：
将该第i行点云中有效点投影到平面上；
根据在该平面上方位角最大的第一有效点和方位角最小的第二有效点的第一连线确定与该第一连线距离最远的第三有效点；
若该第三有效点与该第一直线的距离大于或者等于第三阈值，则确定与该第三有效点为第一突变点，在该平面上除过该第一有效点、第二有效点、第三有效点之外的有效点中，分别确定与第一有效点和第三有效点的之间的第二连线、以及与第二有效点和第三有效点的之间的第三连线距离最远的第四有效点和第五有效点，若该第四有效点与该第二连线的距离大于或者等于该第三阈值，则该第i行点云中的突变点个数增加1个，若该第五有效点与该第三连线的距离大于或者等于该第三阈值，则该第i行点云中的突变点个数增加1个；重复上述过程，直到在所述平面上没有新的直线产生为止，累加可以得到该行点云中的突变点的总个数。
若该第三有效点与该第一直线的距离小于第三阈值，则该第i行点云中的突变点总个数为0。
在该实现方式中，通过上述方法确定每一行点云的突变点数，便于实现，并且，结果的精确度较高，可以更好的反映出每一行点云的突变情况。
在第一方面一种可能的实现方式中，该点云网格中每一列点云的不连续度是根据该每一列点云中包括的有效点个数确定的。
在第一方面一种可能的实现方式中，该方法还包括：
若该点云网格中第j列点云中有效点数为0，则确定该第j列点云的不连续度为a，j为正整数，a为正数；
若该点云网格中第j列点云中有效点数为1，且该第j列为该点云网格中的最后一列或者第一列，则确定该第j列点云的不连续度为0，j为正整数；
若该点云网格中第j列点云中有效点数为1，且该点云网格中第j+1列或者第j-1列点云中有效点数大于1，则确定该第j列点云的不连续度为b，j为大于1的整数，b为正数；
若该点云网格中第j列点云中有效点数为大于1，则确定该第j列点云中有效点对应的行坐标是否连续，如果连续，则确定该第j列点云的不连续度为0，如果不连续，则确定该第j列点云的不连续度为c，j为大于1的整数，c为正数。
在该实现方式中，通过上述方法确定每一列点云的不连续度，便于实现，复杂度低，并且，结果的精度较高，可以更好的反映出每一列点云的不连续度情况。
在第一方面一种可能的实现方式中，该目标激光点云由单个激光雷达单帧检测获得。在该实现方式中，采用单个激光雷达单帧点云获取点云网格，降低了计算延迟，标定简单，便于实施。
在第一方面一种可能的实现方式中，该假目标激光点是由路面积水导致飞溅水花、汽车尾气、或者路面飞尘中的一种或者多种造成的。
第二方面，提供了一种目标识别的装置，该装置包括用于执行以上第一方面或第一方面的任意可能的实现方式中的各个步骤的单元。
第三方面，提供了一种目标识别的装置，该装置包括至少一个处理器和存储器，该至少一个处理器用于执行以上第一方面或第一方面的任意可能的实现方式中的方法。
第四方面，提供了一种目标识别的装置，该装置包括至少一个处理器和接口电路，该至少一个处理器用于执行以上第一方面或第一方面的任意可能的实现方式中的方法。
第五方面，提供了一种激光雷达，该激光雷达包括：上述第三方面、第四方面或者第五方面提供的目标识别的装置。
可选的，该激光雷达可以安装在移动设备上。可选的，移动设备可以包括车辆、飞机、无人机、轮船等可以通过人的操作移动空间位置或改变空间形状的设备。
第六方面，提供了一种计算机程序产品，该计算机程序产品包括计算机程序，该计算机程序在被处理器执行时，用于执行第一方面或第一方面的任意可能的实现方式中的方法。
第七方面，提供一种计算机可读存储介质，该计算机可读存储介质中存储有计算机程序，当该计算机程序被执行时，用于执行第一方面或第一方面的任意可能的实现方式中的方法，或者执行第二方面或第二方面的任意可能的实现方式中的方法。
第八方面，提供了一种芯片或者集成电路，该芯片或者集成电路包括：处理器，用于从存储器中调用并运行计算机程序，使得安装有该芯片或者集成电路的设备执行第一方面或第一方面的任意可能的实现方式中的方法。
第九方面，提供了一种芯片系统，该芯片系统应用于激光雷达；该芯片系统包括一个或多个接口电路和一个或多个处理器；该接口电路和该处理器通过线路互联；该处理器用于执行第一方面或第一方面的任意可能的实现方式中的方法。
本申请实施例提供的目标识别的方法和装置，通过利用单个激光雷达单帧检测获得的激光点云，将激光点云转化为网格，利用点云网格的几何特征而非激光雷达反射点的强度特征确定该激光点云是否是由于路面积水导致飞溅水花、路面飞尘或者汽车尾气等造成的假目标，可以有效的降低误检率，提高假目标的检出率，对不同型号和品牌的激光雷达的适用和兼容性比较好，方法实现简单，便于实施。并且，由于是利用单个激光雷达单帧检测，可以降低了计算延迟。
附图说明
图1是激光雷达基本工作原理示意图。
图2是本申请实施例提供的一例目标识别的方法的示意性流程图。
图3是本申请实施例提供的一例激光雷达的线束的示意图。
图4是本申请实施例提供的一例将目标激光点云转化为点云网格的示意图。
图5是本申请实施例提供的一例点云网格的示意图。
图6是本申请实施例提供的将点云网格中的一行中的连续三个有效点转化到在x-o-y平面上的示意图。
图7是本申请实施例提供的将点云网格中的一行中的所有有效点转化到在x-o-y平面上的示意图。
图8是本申请实施例提供的另一例点云网格的示意图。
图9是本申请实施例提供的目标识别的装置的示意性框图。
图10是本申请实施例提供的另一例目标识别的装置的示意性框图。
具体实施方式
下面将结合附图，对本申请中的技术方案进行描述。
近年来，随着人工智能技术的持续发展，各大高科技企业、研究所、学者对自动驾驶这一领域的研究投入也不断增大。目前对于自动驾驶而言，业内普遍认为激光雷达是不可缺少的传感器之一。激光雷达，是以发射激光束探测目标的位置的主动传感器。其工作原理是向目标发射探测信号(或者也可以称为激光束)，然后将接收到的从目标反射回来的信号(或者也可以称为目标回波)的时间戳与发射信号的时间戳进行比较。基于光速，就可获得目标的距离信息，从而实现对目标进行探测、跟踪和识别。一般情况下，激光雷达由激光发射机、光学接收机、转台和信息处理系统等组成。其中，激光发射机将电脉冲变成光脉冲发射出去，光接收机再把从目标反射回来的光脉冲还原成电脉冲，送到信息处理系统。目前在自动驾驶领域，例如在支持自动驾驶的汽车上，激光雷达通常承担的任务包括目标检测、目标跟踪等。
激光雷达的基本工作原理如图1所示，激光雷达扫描得到原始激光点云，对应原始激光点云，首先需要进行去地处理以去除地面点，然后对去地后的点云进行聚类处理，得到目标点云，在将目标点云送入目标跟踪模块之前，需要对不合理的假目标点云进行滤除，防止对跟踪以及后续的规划控制模块造成不利影响。其中，点云可以理解为：通过测量仪器得到的产品外观表面的点数据集合，称之为点云。激光点云可以理解为：当一束激光照射到物体表面时，所反射的激光会携带方位、距离等信息。若将激光束按照某种轨迹进行扫描，便会边扫描边记录到反射的激光点信息，由于扫描极为精细，则能够得到大量的激光点，因而就可形成激光点云。激光点云是激光雷达通过扫描周围环境得到的对周围环境的三维描述。
道路交通场景下，由于激光雷达采用红外激光测距法，故受到路面飞尘、车辆尾气以及积水路面溅起的水雾反射的影响，容易产生大量的假反射点，经过聚类后产生大量假目标点云，影响后续的跟踪和路径规划等。
目前，有学者提出了一种下雨环境下的激光雷达数据抗干扰处理方法及其实验装置。该方法对模拟下雨环境中的激光雷达的数据进行采样，并根据测量的距离和激光反射点的强度，将激光数据转化为图像，经滤波后，采用连续帧比较算法进行下雨干扰点的剔除。该方法对二维激光雷达在下雨环境下数据的处理方法提供了一种解决方案。
但是，上述方法利用了激光雷达的强度信息，激光雷达的强度是反映生成某点的激光雷达脉冲回波强度的一种测量指标(针对每个激光反射点而采集)，该强度值在一定程度上等于被激光雷达脉冲扫到的对象的反射率，反射率是激光雷达所用波长(通常是在近红外波段)的函数。由于不同厂家的激光雷达强度值不同，因此限制了对不同厂家的激光雷达的适用性。并且，该算法利用了连续帧，因此雨滴检测存在延迟，不能在第一时间去除雨滴造成的假目标，故无法完全避免对后续跟踪的不利影响。进一步的，该方法需要单独建立灰度图这一数据结构，存在算力浪费。并且，由于利用了激光雷达的强度信息，因此只能对反射强度极低的雨滴进行去除，而不能去除路面上的汽车尾气、飞尘造成的假目标。
此外，还有其他学者提供了另外一种方基于双激光雷达的雨滴反射点去除算法。该方法用两个激光雷达同时扫描，经过处理后，未被两个激光雷达都扫到的点归结于雨滴造成的假反射点。
但是，由于该算法需要两个激光雷达，因此硬件的成本较高。两个激光雷达的点云数据需要精确的时间同步和空间一致性，因此算法相对复杂，对标定精度和每个激光雷达的 测量精度要求极高。并且，该算法只提及对雨滴造成的假反射点进行去除，并不能对路面飞尘和汽车尾气造成的假目标实现有效去除。
可见，目前提供的去除激光点云中的假目标方法均不能有效的去除由于对路面飞尘和汽车尾气造成的假目标，并且，方法均比较复杂，实施的成本较高，导致激光点云还存在较多的假目标，严重的降低了自动驾驶的精度和安全性。
有鉴于此，本申请提供了一种目标识别的方法，通过将单个激光雷达单帧检测获得的激光点云转化为网格，利用网格的几何特征而非激光点云的强度特征确定该激光点云是否是由于路面积水导致飞溅水花、路面飞尘或者汽车尾气等造成的假目标，可以有效的降低误检率，提高假目标的检出率，并且，对不同型号和品牌的激光雷达的适用和兼容性比较好，方法实现简单，便于实施。
本申请提供的目标识别的方法，可以应用在自动驾驶领域。例如，自动驾驶移动设备等，移动设备可以包括车辆、飞机、无人机、轮船等可以通过人的操作移动空间位置或改变空间形状的设备。本申请对于移动设备的具体形态不作限制。
下面结合图2详细说明本申请提供的目标识别的方法，图2是本申请一个实施例的目标识别的方法200的示意性流程图，
应理解，本申请提供的目标识别的方法的执行主体可以是集成在移动设备上的目标识别的装置。或者，还可以是移动设备上的芯片、芯片系统或者集成电路等，或者，还可以是激光雷达，该激光雷达包括上述的芯片、芯片系统或者集成电路。本申请在此不做限制。
如图2所示的，该方法200包括：
S210，接收目标激光点云，生成该目标激光点云对应的点云网格，其中，该目标激光点云中的一个点对应该点云网格中的一个有效点。
S220，根据该点云网格的几何特征，判断该目标激光点云是否为噪声点云，该点云网格的几何特征包括：该点云网格的微观粗糙度，和/或，该点云网格中点云的分布不连续度。
在S210中，激光雷达首先会扫描得到原始激光点云。应理解，在本申请实施例中，该原始激光点云由单个激光雷达单帧检测获得。对于接收到的原始激光点云，可以先进行去除地面点，然后对去地后的点云进行聚类处理，得到目标激光点云(或者也可以成为目标点云)。由于目标点云中可能包括由于路面积水导致飞溅水花、路面飞尘或者汽车尾气等造成的假目标点云，因此需要对该目标点云需要进行这些不合理的假目标点云的去除。在本申请实施例中，可以生成该目标激光点云对应的点云网格(mesh)。其中，该目标激光点云中的一个点对应该点云网格中的一个有效点。点云网格中有效点可以理解为有返回点，即激光雷达接收到的从目标反射回来的点。例如，这里的目标包括：路面上汽车尾气、飞溅水花、飞尘造成的假目标以及路边灌木丛等表面杂乱的假目标等，以及车辆旁边的障碍物(例如行人、路边建筑等)等可以有效的用于车辆后续的跟踪和路径规划的有效目标。
在S220中，根据该点云网格的几何特征，识别(或者确定)该目标激光点云是否为噪声点云。在本申请实施例中，噪声点云指目标激光点云中不代表车辆驾驶过程中有效目标的激光点云，噪声点云在车辆后续的跟踪和路径规划中不做处理。噪声点云也可以称为假目标激光点云，噪声点云可以理解为：由于路面积水导致飞溅水花、路面飞尘或者汽车尾气、以及路边灌木丛等表面杂乱的假目标等造成的目标点云。这些假目标造成的激光点 云会影响后续的跟踪和路径规划等，因此，需要将这些假目标造成的噪声点云去除或者标识为无效，在后续的处理中，将不会处理该噪声点云。
该点云网格的几何特征包括：该点云网格的微观粗糙度，和/或，该点云网格中点云的分布不连续度。点云网格的微观粗糙度可以理解为：点云网格中的有效点之间的突变程度或者差异性的大小。点云网格中点云的分布不连续度可以理解为：点云网格中的有效点之间的不连续程度，进一步的，可以通过该点云网格的几何特征识别或者确定该目标激光点云是否为噪声点云。例如，点云网格的微观粗糙度越大，该激光点云越有可能为噪声点云。该点云网格中点云的分布不连续度越大，该激光点云越有可能为噪声点云。
应理解的，步骤S220的实质是识别出不适合后续在车辆后续的跟踪和路径规划中进行处理的目标激光点云。因此，如下替代方案也在步骤S220的范围之内：
根据该点云网格的几何特征，将该目标激光点云标识为无效；或者
根据该点云网格的几何特征，确认该目标激光点云不做为车联后续处理的输入。
将目标激光点云标识为无效则意味着该目标点云不会进入车辆后续的跟踪和路径规划处理。
具体可以包括：
若该点云网格的微观粗糙度大于或者等于第一阈值，则将该目标激光点云标识为无效；和/或，
若该点云网格中点云的分布不连续度大于或者等于第二阈值，则将该目标激光点云标识为无效。
本申请提供的目标识别的方法，通过利用单个激光雷达单帧检测获得的激光点云，将激光点云转化为网格，利用点云网格的几何特征而非激光雷达反射点的强度特征确定该激光点云是否是由于路面积水导致飞溅水花、路面飞尘或者汽车尾气等造成的假目标，可以有效的降低误检率，提高假目标的检出率，方法实现简单，便于实施。并且，由于是利用单个激光雷达单帧检测，可以降低了计算延迟。
下面简单说明生成目标激光点云对应的点云网格的过程。
如图3所示，图3所示的为激光雷达的线束的示意图。自动驾驶利用的激光雷达一般有多条激光线束(Beam)。例如，常见的有16线、32线、64线等。每条线束有一个序号，从下到上分别是0到线数-1。可选的，如果激光雷达原始线束编号不是按照该顺序，也可在生成点云网格时对序号进行重映射，使其满足该顺序。所有的线束会同步的绕激光雷达垂直轴(Z轴)扫描，扫过的角度可以称为方位角γ,每扫过一定的角度就采样一次，每扫过的一定的角度可以称为水平角分辨率α h。
图4中的a图所示的为一例激光点云在X-Y-Z坐标系中的示意图。以图4中的a图所示的黑色点为例。根据其在X-O-Y平面上的坐标可以计算得到方位角γ，除以水平角分辨率α h，再向上取整便得到该点在网格中的列坐标(例如是12)。若该点是序号为6的线束扫描得到，则其在网格中的行坐标是6。若该目标点云中的点行坐标在5到8之间，列坐标在11-15之间，则该目标点云可以转化为4行5列的网格。图4中的b图所示的为由a图得到的点云网格的示意图。
其中，激光点云中任意一个点对应的网格中的列坐标可以由公式(1)计算得到，行坐标(row)为该点的激光雷达的线束(laser beam channel)序号。
Figure PCTCN2020084818-appb-000013
公式(1)中，γ为该点的方位角，α h为该点的水平角分辨率。
利用上述的方法，可以将目标激光点云转化为点云网格。其中，图4中的b图所示点云网格中，包括有效点和无效点，有效点也可以称为有返回点，无效点表示该激光雷达的线束发射的激光没有收到从目标反射回来的点，相当于点云网格中的该点没有激光点云，仅仅是占了点云网格中的一个位置而已。
可选的，在本申请实施例中，作为一种可能的实现方式，在S220中，判断该目标激光点云是否为噪声点云，具体可以包括：
若该点云网格的微观粗糙度大于或者等于第一阈值，判断或者确定该目标激光点云是为噪声点云；和/或，
若该点云网格中点云的分布不连续度大于或者等于第二阈值，判断或者确定该目标激光点云是为噪声点云。
具体而在，可以利用如下三种可能的方式，根据该点云网格的几何特征判断或者确定该目标激光点云是否为噪声点云。
一种可能的实现方式为：利用该点云网格的微观粗糙度与预设的阈值(第一阈值)进行比较，例如，若该点云网格的微观粗糙度大于或者等于第一阈值(或者大于第一阈值)，则判断该目标激光点云是为噪声点云。若该点云网格的微观粗糙度小于第一阈值(或者，小于或者第一阈值)，则判断该目标激光点云不是噪声点云，即该目标激光点云为可以用于车辆后续的跟踪和路径规划的目标点云。
另一种可能的实现方式为：利用点云网格中点云的分布不连续度与预设的阈值(第二阈值)进行比较。例如，若该点云网格中点云的分布不连续度大于或者等于第二阈值(或者大于第二阈值)，判断该目标激光点云是为噪声点云。若该点云网格中点云的分布不连续度小于第二阈值(或者，小于或者等于第二阈值)，判断该目标激光点云不是噪声点云，即该目标激光点云为可以用于车辆后续的跟踪和路径规划的目标点云。
再一种可能的实现方式为：结合该点云网格的微观粗糙度以及分布不连续度进行判断。例如，若该点云网格的微观粗糙度大于或者等于第一阈值(或者大于第一阈值)，并且，该点云网格中点云的分布不连续度大于或者等于第二阈值(或者大于第二阈值)，判断该目标激光点云是为噪声点云。如果，若该点云网格的微观粗糙度小于第一阈值(或者，小于或者第一阈值)，或者，该点云网格中点云的分布不连续度小于第二阈值(或者，小于或者等于第二阈值)，则判断该目标激光点云不是噪声点云。
应理解，上述三种方式只是示例性的说明本申请实施例中利用该点云网格的几何特征确定该目标激光点云是否为噪声点云。应该理解的时，在本申请实施例中，该点云网格的几何特征还可以包括其他维度或者类型的几何特征。并且，在根据点云网格的微观粗糙度，和/或，分布不连续度进行判断目标激光点云是否为噪声点云的过程中，还可以根据其他方式或者规则，利用点云网格的微观粗糙度，和/或，分布不连续度进行判断。本申请实施例在此不做限制。
可选的，作为一种可能的实现方式，在本申请实施例中，该点云网格的微观粗糙度(或者点云网格的微观粗糙度的值)可以为：该点云网格中的突变点总数占该点云网格中包括 的有效点数的比例。其中，该点云网格中的突变点总数为该点云网格中每一行中的突变点数之和。点云网格中的突变点可以理解为点云网格中包括的部分有效点，这部分有效点与其他有效点的突变程度或者差异性比较大。换句话说，突变点可以为点云网格中的有效点的子集。可选的，该点云网格中的突变点总数可以为该点云网格中每一行中的突变点数之和。也就是说，可以逐行评价点云网格中的突变点数，最终得到该点云网格中的突变点总数。
可选的，作为一种可能的实现方式，在本申请实施例中，该点云网格中点云的分布不连续度为：该点云网格中每一列点云的不连续度之和。也就是说，可以逐列评价点云网格中的每一列中的有效点云的不连续度，将所有列的不连续度相加，最终得到该点云网格中点云的分布不连续度。
应理解，在本申请实施例中，对于该点云网格的微观粗糙度，还可以利用其他方式确定。例如，逐列评价点云网格中的每一列中的有效点云的微观粗糙度(例如确定点云网格中的每一列中的突变点的个数)，将每一列中的有效点云的突变点的个数的相加，得到该点云网格中的突变点总数，最终得到该点云网格中的突变点总数占该点云网格中包括的有效点数的比例，得到该点云网格的微观粗糙度。类似的，对于点云网格中点云的分布不连续度，也逐行确定点云网格中的不连续度，将每一行的不连续度相加，最终得到该点云网格中点云的分布不连续度。或者，还可以利用其他计算方法确定该点云网格的微观粗糙度和分布不连续度。本申请对于确定该点云网格的微观粗糙度和分布不连续度的具体方式或者算法不做限制。
下面将具体说明点云网格的微观粗糙度和布不连续度的计算过程。
首先介绍点云网格的微观粗糙度的计算过程。以图5所示的目标点云网格为例进行说明。该点云网格中的突变点总数可以为该点云网格中每一行中的突变点数之和。对于图5所示的目标点云网格而言：
对于第一行(在点云网格中的行坐标为8)，第一行点云中有效点数等于3，则默认该三个有效点为连续的三个有效点，假设这连续的三个有效点分别为第一有效点、第二有效点、第三有效点，若第一有效点与第三有效点的连线与第二有效点之间的距离大于或者等于预设的阈值，则确定第二有效点为突变点；若第一有效点与第三有效点的连线与第二有效点之间的距离小于该预设的阈值，则第二有效点不是突变点。下面将结合例子具体说明。
取第一行点云中连续的三个有效点确定该连续的三个有效点中的第二有效点是否为突变点。假设第一行中的三个有效点分别为第一有效点P(在点云网格中的坐标为(8,12))、第二有效点Q在点云网格中的坐标为(8,13)、第三有效点S在点云网格中的坐标为(8,15)。将这三个有效点投影到x-o-y平面上，得到图5所示的这三个有效点在x-o-y平面上坐标图，如图6所示的，假设点P、点S所处的微小区域非常平滑，则点Q应该在点P和点S连接成的直线上。
由于点P、点S的坐标已知，根据向量点成公式，如图6所示的，O为坐标原点，
Figure PCTCN2020084818-appb-000014
与
Figure PCTCN2020084818-appb-000015
的夹角为β，
Figure PCTCN2020084818-appb-000016
与
Figure PCTCN2020084818-appb-000017
的夹角为θ，
Figure PCTCN2020084818-appb-000018
与
Figure PCTCN2020084818-appb-000019
的夹角为α，其中，α＝180°-θ-β，
则有：
Figure PCTCN2020084818-appb-000020
Figure PCTCN2020084818-appb-000021
假设与第二有效点Q对应的参考点为Q 1，根据正弦定理:
则
Figure PCTCN2020084818-appb-000022
进一步的进行判断：若
Figure PCTCN2020084818-appb-000023
与
Figure PCTCN2020084818-appb-000024
长度的差值大于或者等于预设的阈值，则第二有效点Q为突变点，若
Figure PCTCN2020084818-appb-000025
与
Figure PCTCN2020084818-appb-000026
长度的差值小于预设的阈值，则第二有效点Q不是突变点。
根据上述的方法，可以确定第一行中的第二点有效点Q是否为突变点，如果是，则第一行中的突变点数为1，如果不是，则第一行中的突变点数为0。由于第一行中只有三个有效点，因此，第一行中的突变数最多为一个。
对于第二行(在点云网格中的行坐标为7)，由于第二行中没有有效点，则点云网格中第二行的突变点数为0。
对于第三行(在点云网格中的行坐标为6)，由于第三行中的有效点数为4个，则每次取三个利用上第一行相同的方法进行判断。例如，假设第三行中的4个有效点以依次为第四有效点(在点云网格中的坐标为(6,11))、第五有效点(在点云网格中的坐标为(6,12))、第六有效点(在点云网格中的坐标为(6,13))、第七有效点(在点云网格中的坐标为(6,15))。首先取第四有效点、第五有效点、第六有效点，按照和第一行相同的方法进行判断第五有效点是否为突变点，在确定第五有效点是否为突变点之后，再取第五有效点、第六有效点、第七有效点，按照和第一行相同的方法进行判断第六有效点是否为突变点。这样便可以确定第三行中的突变点数。由于第三行中只有四个有效点，因此，第三行中的突变点的个数可能为：2个(第五有效点和第六有效点)、1个(第五有效点或者和第六有效点)、0个。
对于第四行(在点云网格中的行坐标为5)，由于第四行中有三个有效点，按照和第一行相同的方法进行判断这三个有效点中的第二个有效点(在点云网格中的坐标为(5,13)是否为突变点，如果是，则第四行中的突变点数为1，如果不是，则第四行中的突变点数为0。由于第四行中只有三个有效点，因此，第四行中的突变数最多为一个。
通过上述的方法，即：若该点云网格中第i行点云中有效点数小于或者等于2，则确定该点云网格中第i行的突变点数为0，i为正整数；根据该第i行点云中每连续的三个有效点中的第二个有效点是否为突变点，确定该点云网格中第i行的突变点数。这样可以确定每一行的突变点数，然后将每一行的突变点数相加，得到该点云网格中的突变点总数，然后计算该点云网格中的突变点总数占该点云网格中包括的所有有效点数的比例，得到该点云网格的微观粗糙度。
可选的，作为另外一种可能的实现方式，对于点云网格的微观粗糙度的计算，还可以利用另外的一种方法。对于任意一个点云网格，该点云网格中的突变点总数可以为该点云 网格中每一行中的突变点数之和。对于该点云网格中每一行中的突变点数，可以将该行中的所有有效点投影到X-O-Y平面上。根据在X-O-Y平面上方位角最大的第一有效点和方位角最小的第二有效点的第一连线确定与第一连线距离最远的第三有效点；
若第三有效点与第一直线的距离大于或者等于预设的阈值(第三阈值)，则确定与第三有效点为第一突变点，在该X-O-Y平面上除过第一有效点、第二有效点、第三有效点之外的有效点中，分别确定与第一有效点和第三有效点的之间的第二连线、以及与第二有效点和第三有效点的之间的第三连线距离最远的第四有效点和第五有效点，若第四有效点与第二连线的距离大于或者等于该第三阈值(或者可以为其他预设的阈值)，则该行点云中的突变点个数增加1个，若第五有效点与第三连线的距离大于或者等于所述第三阈值(或者可以为其他预设的阈值)，则该行点云中的突变点个数增加1个，即每确定出一个新的有效点，便新增加两条直线。重复上述过程，直到点云网格中没有使用的剩余的有效点为0，即没有新的直线产生。累加可以得到该行点云中的突变点的总个数。
若第三有效点与第一直线的距离小于第三阈值，则该行点云中的突变点的总个数为0。
换句话说，在由有效点A(方位角最大的有效点)和有效点B(方位角最小的有效点)确定出第一连线后，由第一连线可以确定出与第一连线距离最远的有效点C，由于有效点C是根据有效点A和有效点B确定出来的，因此，有效点C可以和有效点A确定一条直线，有效点C可以和有效点B确定一条直线，这两条直线分别可以相当于上述的第一连线，对于每条直线，在点云网格中没有使用的剩余的有效点中利用和确定有效点C类似的方法，又可以确定出新的有效点和新的直线，直到点云网格中没有使用的剩余的有效点为0，即没有新的直线产生。对于每一个新确定的有效点，将该新确定的有效点与产生该新确定的有效点的直线之间的距离与预设的阈值进行比较，确定该新确定的有效点是否为突变点，将突变点数累加便可以得到该行点云中的突变点的总个数。
例如，假设在X-O-Y平面上有9个有效点，有效点的编号分别为1、2、3…..9。假设1号有效点和2号有效点分别为方位角最大的有效点和方位角最小的有效点，则依次产生的直线的顺序可以为表1所示的。
表1
Figure PCTCN2020084818-appb-000027
其中，表1中，产生的直线的次序编号表示产生的直线的先后顺序，编号越小，产生的最早。例如，编号为1最先产生，编号为2次之。即在未使用中的有效点中每确定出一个新的有效点，便新增加两条直线。
下面结合具体的例子进行说明。
例如，假设图7所示的为将一个点云网格中某一行中的所有有效点投影到X-O-Y平面上的示意图。如图7所示的，在X-O-Y平面上，选择方位角最大(θ max)的点A和方位角最小(θ min)的点H，连接AH形成一条直接，然后确定到该直线AH距离最大的点C，确定点C到该直线AH的距离(假设为L 1),如果L 1大于或者等于预设的阈值则证明点C为突变点。如果L 1小于预设的阈值则证明点C不是突变点，并且，可以确定该点云网格中没有突变点。
如果L 1大于或者等于预设的阈值，进一步的，连接AC形成直线，连接CH形成直线，在该X-O-Y平面上除过A点、C点、H点之外的有效点中，分别确定与AC直线距离最远的点以及与直线CH距离最远的点。假设与AC直线距离(假设为L 2)最远的点为G点，与CH直线距离(假设为L 3)最远的点为D点，继续判断L 2是否超过预设的阈值，L 3是否超过预设的阈值。如果两者均超过，则证明G点和D点均为突变点。如果两者均没有超过，则证明G点和D点均不是突变点，也就是说，该点云网格中只有一个突变点C。应该理解，与L 2进行比较的阈值和与L 3进行比较的阈值可以相同，也可以不同。可选的，与L 2进行比较的阈值可以和与L 1进行比较的阈值相同或者不同。
如果L 2超过预设的阈值，L 3没有超过预设的阈值，证明G点也为突变点，D点不是突变点。进一步的，继续连接AG形成直线AG，连接CG形成直线CG，在该X-O-Y平面上除过A点、C点、H点、G点、D点之外的有效点中，继续确定与直线AG距离最远的点以及与直线CG最远的点，然后分别将距离与预设的阈值进行比较。重复该过程，直到将该平面上的该行点云中所有的剩余的有效点均覆盖到则停止，累加可以得到该行点云中的突变点的总个数。
如果L 2没有超过预设的阈值，L 3超过预设的阈值，证明G点不是突变点，而D点是突变点。进一步的，继续连接CD形成直线CD，连接HD形成直线HD，在该X-O-Y平面上除过A点、C点、H点、G点、D点之外的有效点中，继续确定与直线CD距离最远的点以及与直线HD最远的点，然后分别将距离与预设的阈值进行比较。重复该过程，直到将该平面上的该行点云中所有的剩余的有效点均覆盖到则停止，累加可以得到该行点云中的突变点的总个数。
应理解，在上述的过程，由于每一次都需要和预设的阈值进行比较，与不同直线之间的距离进行比较时，所利用的阈值也可以是不同的，或者，也可以是相同的。
利用上述的方法，可以确定该点云网格中每一行中的突变点数然后将每一行的突变点数相加，得到该点云网格中的突变点总数，然后计算该点云网格中的突变点总数占该点云网格中包括的有效点数的比例，得到该点云网格的微观粗糙度。
应理解，在本申请实施例中，除了利用上述的方式确定点云网格中每一行中的突变点数之外，还可以利用其他的方式或者算法确定点云网格中每一行中的突变点数。本申请实施例在此不作限制。
对于点云网格中点云的分布不连续度，可以为点云网格中每一列点云的不连续度之和。
可选的，点云网格中每一列点云的不连续度可以是根据每一列点云中包括的有效点个数确定的。
例如：若该点云网格中第j列点云中有效点数为0，则确定该第j列点云的不连续度为a，j为正整数，a为正数；
若该点云网格中第j列点云中有效点数为1，且该第j列为该点云网格中的最后一列或者第一列，则确定该第j列点云的不连续度为0，j为正整数；
若该点云网格中第j列点云中有效点数为1，且该点云网格中第j+1列或者第j-1列点云中有效点数大于1，则确定该第j列点云的不连续度为b，j为大于1的整数，b为正数；
若该点云网格中第j列点云中有效点数为大于1，则确定该第j列点云中有效点对应的行坐标是否连续，如果连续，则确定该第j列点云的不连续度为0。如果不连续，则确定该第j列点云的不连续度为c，j为大于1的整数，c为正数。
可选的，a大于b，b大于c。
下面将结合图8所示的点云网格为例，具体说明确定点云网格中的每一列中的有效点云的不连续度的具体方式。
对于图8所示的点云网格，对于第一列点云(在点云网格中的列坐标为11)，有效点数为1，且第一列为点云网格中的第一列点云，则确定第一列点云的不连续度为0。
对于第二列点云(在点云网格中的列坐标为12)，有效点数为4个，大于1，需要进一步的判第二列上的有效点云的行坐标(即激光线束序号)是否连续。对于第二列点云的4个有效点，其行坐标分别为5、6、7、8，即激光线束序号是连续的，则第二列点云的不连续度为0。
对于第三列点云(在网格中的列坐标为13)，有效点数为1个，且为中间列，左相邻的第二列的有效点数大于1，则确定其不连续度为b，即对于第三列点云而言，其不连续度为b。
对于第四列点云(在网格中的列坐标为14)，有效点数为0个，则确定该第四列点云的不连续度为a。
对于第五列点云(在网格中的列坐标为15)，有效点数为3个，需要进一步的判第五列上的有效点云的行坐标(即激光线束序号)是否连续。对于第五列点云的3个有效点，其行坐标分别为5、6、8，即激光线束序号是不连续的，则第五列点云的不连续度为c。
通过上述的方法，可以分别确定每一列点云的不连续度，然后将所有列的不连续度相加，最终得到该点云网格中点云的分布不连续度。
应理解，在本申请实施例中，除了利用上述的方式确定点云网格中每一列点云的分布不连续度之外，还可以利用其他的方式或者算法确定点云网格中每一列点云的分布不连续度。本申请实施例在此不作限制。
通过上述的方法，可以确定点云网格中点云的分布不连续度以及该点云网格的微观粗糙度。从而确定该目标点云是否为假目标点云，方法实现简单，便于实施。并且，仅采用点云网格的几何特征，并不利用激光雷达反射信号的强度特征，对不同型号和品牌的激光雷达的适用和兼容性比较好。采用单个激光雷达单帧点云获取点云网格，可以降低了计算延迟，并且，点云中假目标的误检漏检低，对于由路面积水导致飞溅水花、路面飞尘或者汽车尾气等造成的假目标均可以准确的检测出，提高了假目标的检出率。
应理解，上述只是为了帮助本领域技术人员更好地理解本申请实施例，而非要限制本申请实施例的范围。本领域技术人员根据所给出的上述示例，显然可以进行各种等价的修 改或变化，例如，上述方法200的各个实施例中某些步骤可以是不必须的，或者可以新加入某些步骤等。或者上述任意两种或者任意多种实施例的组合。这样的修改、变化或者组合后的方案也落入本申请实施例的范围内。
还应理解，上文对本申请实施例的描述着重于强调各个实施例之间的不同之处，未提到的相同或相似之处可以互相参考，为了简洁，这里不再赘述。
还应理解，上述各过程的序号的大小并不意味着执行顺序的先后，各过程的执行顺序应以其功能和内在逻辑确定，而不应对本申请实施例的实施过程构成任何限定。
还应理解，本申请实施例中，“预先设定”、“预先定义”可以通过在设备(例如，包括终端和网络设备)中预先保存相应的代码、表格或其他可用于指示相关信息的方式来实现，本申请对于其具体的实现方式不做限定。
还应理解，本申请实施例中的方式、情况、类别以及实施例的划分仅是为了描述的方便，不应构成特别的限定，各种方式、类别、情况以及实施例中的特征在不矛盾的情况下可以相结合。
还应理解，在本申请的各个实施例中，如果没有特殊说明以及逻辑冲突，不同的实施例之间的术语和/或描述具有一致性、且可以相互引用，不同的实施例中的技术特征根据其内在的逻辑关系可以组合形成新的实施例。
以上结合图1至图8对本申请实施例的目标识别的方法做了详细说明。以下，结合图9至图10对本申请实施例目标识别的装置进行详细说明。
图9示出了本申请实施例的目标识别的装置300的示意性框图，该装置300可以对应上述方法200中描述的激光雷达，也可以是应用于激光雷达上的芯片、组件、集成电路、车载处理器等中的芯片等。并且，该装置300中各模块或单元分别用于执行上述方法200中所执行的各动作或处理过程。
如图9所示，该装置300包括收发单元310和处理单元320，收发单元310用于在处理单元320的驱动下执行具体的信号收发。
收发单元310，用于接收目标激光点云；
处理单元320：用于生成该目标激光点云对应的点云网格，其中，该目标激光点云中的一个点对应该点云网格中的一个有效点。
该处理单元320，还用于根据该点云网格的几何特征，判断或者确定该目标激光点云是否为噪声点云，该点云网格的几何特征包括：该点云网格的微观粗糙度，和/或，该点云网格中点云的分布不连续度。
本申请提供的目标识别的装置，通过利用单个激光雷达单帧检测获得的激光点云，将激光点云转化为网格，利用点云网格的几何特征而非激光雷达反射点的强度特征确定该激光点云是否是由于路面积水导致飞溅水花、路面飞尘或者汽车尾气等造成的假目标，可以有效的降低误检率，提高假目标的检出率。并且，由于是利用单个激光雷达单帧检测，对不同型号和品牌的激光雷达的适用和兼容性比较好，方法实现简单，便于实施。
可选的，在本申请的一些实施例中，该点云网格的微观粗糙度为:该点云网格中的突变点总数占该点云网格中包括的有效点数的比例，其中，该点云网格中的突变点总数为该点云网格中每一行中的突变点数之和。
可选的，在本申请的一些实施例中，该点云网格中点云的分布不连续度为:该点云网 格中每一列点云的不连续度之和。
可选的，在本申请的一些实施例中，该处理单元320还用于：在该点云网格的微观粗糙度大于或者等于第一阈值的情况下，判断该目标激光点云是为噪声点云；和/或，在该点云网格中点云的分布不连续度大于或者等于第二阈值的情况下，判断该目标激光点云是为噪声点云。
可选的，在本申请的一些实施例中，该处理单元320还用于：
在该点云网格中第i行点云中有效点数小于或者等于2的情况下，则确定该点云网格中第i行的突变点数为0，i为正整数；
在该点云网格中第i行点云中有效点数大于或者等于3的情况下，则在该第i行点云中连续的三个有效点确定该连续的三个有效点中的第二有效点是否为突变点，i为正整数；
根据该第i行点云中每连续的三个有效点中的第二个有效点是否为突变点，确定该点云网格中第i行的突变点数。
可选的，在本申请的一些实施例中，该处理单元320还用于：
若该点云网格中第i行点云中有效点数大于或者等于3，该第i行点云中连续的三个有效点分别为第一有效点、第二有效点、第三有效点，若该第一有效点与该第三有效点的连线与该第二有效点之间的距离大于或者等于预设的阈值，则确定该第二有效点为突变点；
若该第一有效点与该第三有效点的连线与该第二有效点之间的距离小于预设的阈值，则确定该第二有效点不是突变点。
可选的，在本申请的一些实施例中，该处理单元320还用于：
该第i行点云中连续的三个有效点分别为第一有效点P、第二有效点Q、第三有效点S，O为坐标原点，
Figure PCTCN2020084818-appb-000028
与
Figure PCTCN2020084818-appb-000029
的夹角为β，
Figure PCTCN2020084818-appb-000030
与
Figure PCTCN2020084818-appb-000031
的夹角为θ，
Figure PCTCN2020084818-appb-000032
与
Figure PCTCN2020084818-appb-000033
的夹角为α,α＝180 °-θ-β，
Figure PCTCN2020084818-appb-000034
与第二有效点对应的参考点为Q 1，
Figure PCTCN2020084818-appb-000035
该处理单元320还用于：在
Figure PCTCN2020084818-appb-000036
与
Figure PCTCN2020084818-appb-000037
长度的差值大于或者等于预设的阈值的情况下，确定该第二有效点Q为突变点，在
Figure PCTCN2020084818-appb-000038
与
Figure PCTCN2020084818-appb-000039
长度的差值小于预设的阈值的情况下，确定该第二有效点Q不是突变点。
可选的，在本申请的一些实施例中，该处理单元320还用于：
将该第i行点云中有效点投影到平面上；
根据在该平面上方位角最大的第一有效点和方位角最小的第二有效点的第一连线确定与该第一连线距离最远的第三有效点；
若该第三有效点与该第一直线的距离大于或者等于第三阈值，则确定与该第三有效点为第一突变点，在该平面上除过该第一有效点、第二有效点、第三有效点之外的有效点中，分别确定与第一有效点和第三有效点的之间的第二连线、以及与第二有效点和第三有效点 的之间的第三连线距离最远的第四有效点和第五有效点，若该第四有效点与该第二连线的距离大于或者等于该第三阈值，则该第i行点云中的突变点个数增加1个，若该第五有效点与该第三连线的距离大于或者等于该第三阈值，则该第i行点云中的突变点个数增加1个；重复上述过程，直到在所述平面上没有新的直线产生为止，累加可以得到该行点云中的突变点的总个数。
若该第三有效点与该第一直线的距离小于第三阈值，则该第i行点云中的突变点个数为0。
可选的，在本申请的一些实施例中，该点云网格中每一列点云的不连续度是根据每一列点云中包括的有效点个数确定的。
可选的，在本申请的一些实施例中，该处理单元320还用于：
在该点云网格中第j列点云中有效点数为0的情况下，确定该第j列点云的不连续度为a，j为正整数，a为正数；
在该点云网格中第j列点云中有效点数为1，且该第j列为该点云网格中的最后一列或者第一列的情况下，确定该第j列点云的不连续度为0，j为正整数；
在该点云网格中第j列点云中有效点数为1，且该点云网格中第j+1列或者第j-1列点云中有效点数大于1的情况下，确定该第j列点云的不连续度为b，j为大于1的整数，b为正数；
若该点云网格中第j列点云中有效点数为大于1，则确定该第j列点云中有效点对应的行坐标是否连续，如果连续，则确定该第j列点云的不连续度为0，如果不连续，则确定该第j列点云的不连续度为c，j为大于1的整数，c为正数。
可选的，在本申请的一些实施例中，该目标激光点云由单个激光雷达单帧检测获得。
可选的，在本申请的一些实施例中，该假目标激光点是由路面积水导致飞溅水花、汽车尾气、或者路面飞尘中的一种或者多种造成的。
应理解，该装置300中各模块(单元)执行上述相应步骤的具体过程请参照前文中结合方法200以及图2至图8中相关实施例中的描述，为了简洁，这里不加赘述。
可选的，收发单元310可以包括接收单元(模块)和发送单元(模块)，用于执行前述方法200中的各个实施例接收信息和发送信息的步骤。
进一步的，该装置300还可以该存储单元(模块)。收发单元310可以是收发器、输入/输出接口或接口电路。存储模块用于存储收发单元310和处理单元320执行的指令。收发单元310、处理单元310和存储单元相互耦合，存储单元存储指令，处理单元320用于执行存储单元存储的指令，收发单元310用于在处理单元320的驱动下执行具体的信号收发。
应理解，收发单元310可以是收发器、输入/输出接口或接口电路。存储单元可以是存储器。处理单元320可由处理器实现。如图10所示，目标识别的装置400可以包括处理器410、存储器420和收发器430。
图9所示的目标识别的装置300或图10所示的目标识别的装置400能够实现前述方法200中的实施例以及图2至图8所示的实施例中执行的步骤。类似的描述可以参考前述对应的方法中的描述。为避免重复，这里不再赘述。
应理解，以上装置中单元的划分仅仅是一种逻辑功能的划分，实际实现时可以全部或 部分集成到一个物理实体上，也可以物理上分开。且装置中的单元可以全部以软件通过处理元件调用的形式实现；也可以全部以硬件的形式实现；还可以部分单元以软件通过处理元件调用的形式实现，部分单元以硬件的形式实现。例如，各个单元可以为单独设立的处理元件，也可以集成在装置的某一个芯片中实现，此外，也可以以程序的形式存储于存储器中，由装置的某一个处理元件调用并执行该单元的功能。这里该处理元件又可以称为处理器，可以是一种具有信号处理能力的集成电路。在实现过程中，上述方法的各步骤或以上各个单元可以通过处理器元件中的硬件的集成逻辑电路实现或者以软件通过处理元件调用的形式实现。
在一个例子中，以上任一装置中的单元可以是被配置成实施以上方法的一个或多个集成电路，例如：一个或多个专用集成电路(application specific integrated circuit，ASIC)，或，一个或多个数字信号处理器(digital signal processor，DSP)，或，一个或者多个现场可编程门阵列(field programmable gate array，FPGA)，或这些集成电路形式中至少两种的组合。再如，当装置中的单元可以通过处理元件调度程序的形式实现时，该处理元件可以是通用处理器，例如中央处理器(central processing unit，CPU)或其它可以调用程序的处理器。再如，这些单元可以集成在一起，以片上系统(system-on-a-chip，SOC)的形式实现。
应理解，本申请实施例中，处理器可以为中央处理单元(central processing unit，CPU)，该处理器还可以是其他通用处理器、数字信号处理器(digital signal processor，DSP)、专用集成电路(application specific integrated circuit，ASIC)、现成可编程门阵列(field programmable gate array，FPGA)或者其他可编程逻辑器件、分立门或者晶体管逻辑器件、分立硬件组件等。通用处理器可以是微处理器或者该处理器也可以是任何常规的处理器等。
还应理解，本申请实施例中的存储器可以是易失性存储器或非易失性存储器，或可包括易失性和非易失性存储器两者。其中，非易失性存储器可以是只读存储器(read-only memory，ROM)、可编程只读存储器(programmable ROM，PROM)、可擦除可编程只读存储器(erasable PROM，EPROM)、电可擦除可编程只读存储器(electrically EPROM，EEPROM)或闪存。易失性存储器可以是随机存取存储器(random access memory，RAM)，其用作外部高速缓存。通过示例性但不是限制性说明，许多形式的随机存取存储器(random access memory，RAM)可用，例如静态随机存取存储器(static RAM，SRAM)、动态随机存取存储器(DRAM)、同步动态随机存取存储器(synchronous DRAM，SDRAM)、双倍数据速率同步动态随机存取存储器(double data rate SDRAM，DDR SDRAM)、增强型同步动态随机存取存储器(enhanced SDRAM，ESDRAM)、同步连接动态随机存取存储器(synchlink DRAM，SLDRAM)和直接内存总线随机存取存储器(direct rambus RAM，DR RAM)。
上述实施例，可以全部或部分地通过软件、硬件、固件或其他任意组合来实现。当使用软件实现时，上述实施例可以全部或部分地以计算机程序产品的形式实现。该计算机程序产品包括一个或多个计算机指令或计算机程序。在计算机上加载或执行该计算机指令或计算机程序时，全部或部分地产生按照本申请实施例的流程或功能。该计算机可以为通用计算机、专用计算机、计算机网络、或者其他可编程装置。该计算机指令可以存储在计算 机可读存储介质中，或者从一个计算机可读存储介质向另一个计算机可读存储介质传输，例如，该计算机指令可以从一个网站站点、计算机、服务器或数据中心通过有线(例如红外、无线、微波等)方式向另一个网站站点、计算机、服务器或数据中心进行传输。该计算机可读存储介质可以是计算机能够存取的任何可用介质或者是包含一个或多个可用介质集合的服务器、数据中心等数据存储设备。该可用介质可以是磁性介质(例如，软盘、硬盘、磁带)、光介质(例如，DVD)、或者半导体介质。半导体介质可以是固态硬盘。
本申请实施例还提供了一种计算机可读介质，用于存储计算机程序代码，该计算机程序包括用于执行上述方法200中本申请实施例的目标识别的方法的指令。该可读介质可以是只读存储器(read-only memory，ROM)或随机存取存储器(random access memory，RAM)，本申请实施例对此不做限制。
本申请还提供了一种计算机程序产品，该计算机程序产品包括指令，当该指令被执行时，使得该目标识别的装置分别执行对应于上述方法200中的操作。
本申请实施例还提供了一种芯片系统，该芯片系统应用于激光雷达。该芯片系统包括一个或多个接口电路和一个或多个处理器；该接口电路和该处理器通过线路互联；该处理器用于执行上述方法200中本申请实施例的目标识别的方法。
本申请实施例还提供了一种芯片，该芯片包括：处理器，用于从存储器中调用并运行计算机程序，使得安装有该芯片的设备执行上述方法200中的目标识别的方法。
可选地，上述本申请实施例中提供的任意一种目标识别的装置可以包括该系统芯片。
可选地，该计算机指令被存储在存储单元中。
可选地，该存储单元为该芯片内的存储单元，如寄存器、缓存等，该存储单元还可以是该终端内的位于该芯片外部的存储单元，如ROM或可存储静态信息和指令的其他类型的静态存储设备，RAM等。其中，上述任一处提到的处理器，可以是一个CPU，微处理器，ASIC，或一个或多个用于控制上述的目标识别的方法的程序执行的集成电路。该处理单元和该存储单元可以解耦，分别设置在不同的物理设备上，通过有线或者无线的方式连接来实现该处理单元和该存储单元的各自的功能，以支持该系统芯片实现上述实施例中的各种功能。或者，该处理单元和该存储器也可以耦合在同一个设备上。
本申请还提供了一种激光雷达，该激光雷达包括本申请实施例提供的任意一种的目标识别的装置。或者，该激光雷达包括本申请提供的任意一种芯片系统或者芯片。
可以理解，本申请实施例中的存储器可以是易失性存储器或非易失性存储器，或可包括易失性和非易失性存储器两者。其中，非易失性存储器可以是只读存储器(read-only memory，ROM)、可编程只读存储器(programmable ROM，PROM)、可擦除可编程只读存储器(erasable PROM，EPROM)、电可擦除可编程只读存储器(electrically EPROM，EEPROM)或闪存。易失性存储器可以是随机存取存储器(random access memory，RAM)，其用作外部高速缓存。通过示例性但不是限制性说明，许多形式的随机存取存储器(random access memory，RAM)可用，例如静态随机存取存储器(static RAM，SRAM)、动态随机存取存储器(DRAM)、同步动态随机存取存储器(synchronous DRAM，SDRAM)、双倍数据速率同步动态随机存取存储器(double data rate SDRAM，DDR SDRAM)、增强型同步动态随机存取存储器(enhanced SDRAM，ESDRAM)、同步连接动态随机存取存储器(synchlink DRAM，SLDRAM)和直接内存总线随机存取存储器(direct rambus RAM， DR RAM)。
在本申请中可能出现的对各种消息/信息/设备/网元/系统/装置/动作/操作/流程/概念等各类客体进行了赋名，可以理解的是，这些具体的名称并不构成对相关客体的限定，所赋名称可随着场景，语境或者使用习惯等因素而变更，对本申请中技术术语的技术含义的理解，应主要从其在技术方案中所体现/执行的功能和技术效果来确定。
本领域普通技术人员可以意识到，结合本文中所公开的实施例描述的各示例的单元及算法步骤，能够以电子硬件、或者计算机软件和电子硬件的结合来实现。这些功能究竟以硬件还是软件方式来执行，取决于技术方案的特定应用和设计约束条件。专业技术人员可以对每个特定的应用来使用不同方法来实现所描述的功能，但是这种实现不应认为超出本申请的范围。
所属领域的技术人员可以清楚地了解到，为描述的方便和简洁，上述描述的系统、装置和单元的具体工作过程，可以参考前述方法实施例中的对应过程，在此不再赘述。
在本申请所提供的几个实施例中，应该理解到，所揭露的系统、装置和方法，可以通过其它的方式实现。例如，以上所描述的装置实施例仅仅是示意性的，例如，该单元的划分，仅仅为一种逻辑功能划分，实际实现时可以有另外的划分方式，例如多个单元或组件可以结合或者可以集成到另一个系统，或一些特征可以忽略，或不执行。另一点，所显示或讨论的相互之间的耦合或直接耦合或通信连接可以是通过一些接口，装置或单元的间接耦合或通信连接，可以是电性，机械或其它的形式。
所述作为分离部件说明的单元可以是或者也可以不是物理上分开的，作为单元显示的部件可以是或者也可以不是物理单元，即可以位于一个地方，或者也可以分布到多个网络单元上。可以根据实际的需要选择其中的部分或者全部单元来实现本实施例方案的目的。
另外，在本申请各个实施例中的各功能单元可以集成在一个处理单元中，也可以是各个单元单独物理存在，也可以两个或两个以上单元集成在一个单元中。
所述功能如果以软件功能单元的形式实现并作为独立的产品销售或使用时，可以存储在一个计算机可读取存储介质中。基于这样的理解，本申请的技术方案本质上或者说对现有技术做出贡献的部分或者该技术方案的部分可以以软件产品的形式体现出来，该计算机软件产品存储在一个存储介质中，包括若干指令用以使得一台计算机设备(可以是个人计算机，服务器，或者网络设备等)执行本申请各个实施例所述方法的全部或部分步骤。而前述的存储介质包括：U盘、移动硬盘、只读存储器(read-only memory，ROM)、随机存取。
以上所述，仅为本申请的具体实施方式，但本申请的保护范围并不局限于此，任何熟悉本技术领域的技术人员在本申请揭露的技术范围内，可轻易想到变化或替换，都应涵盖在本申请的保护范围之内。因此，本申请的保护范围应以所述权利要求的保护范围为准。

Claims (24)

  1. 一种目标识别方法，其特征在于，包括：
    接收目标激光点云，生成所述目标激光点云对应的点云网格，其中，所述目标激光点云中的一个点对应所述点云网格中的一个有效点；
    根据所述点云网格的几何特征，判断所述目标激光点云是否为噪声点云，所述点云网格的几何特征包括：所述点云网格的微观粗糙度，和/或，所述点云网格中点云的分布不连续度。
  2. 根据权利要求1所述的方法，其特征在于，
    所述点云网格的微观粗糙度，为所述点云网格中的突变点总数占所述点云网格中包括的有效点数的比例；
    其中，所述点云网格中的突变点总数为所述点云网格中每一行中的突变点数之和。
  3. 根据权利要求1或2所述的方法，其特征在于，
    所述点云网格中点云的分布不连续度，为所述点云网格中每一列点云的不连续度之和。
  4. 根据权利要求1至3中任一项所述的方法，其特征在于，
    所述识别所述目标激光点云是否为噪声点云，包括：
    若所述点云网格的微观粗糙度大于或者等于第一阈值，则判断所述目标激光点云为噪声点云；和/或，
    若所述点云网格中点云的分布不连续度大于或者等于第二阈值，则判断所述目标激光点云为噪声点云。
  5. 根据权利要求2至4中任一项所述的方法，其特征在于，所述方法还包括：
    若所述点云网格中第i行点云中有效点数小于或者等于2，则确定所述点云网格中第i行的突变点数为0，i为正整数；
    若所述点云网格中第i行点云中有效点数大于或者等于3，则依次取所述第i行点云中连续的三个有效点确定所述连续的三个有效点中的第二个有效点是否为突变点，i为正整数；
    根据所述第i行点云中每连续的三个有效点中的第二个有效点是否为突变点，确定所述点云网格中第i行的突变点数。
  6. 根据权利要求5所述的方法，其特征在于，
    若所述点云网格中第i行点云中有效点数大于或者等于3，所述第i行点云中连续的三个有效点分别为第一有效点、第二有效点、第三有效点，若所述第一有效点与所述第三有效点的连线与所述第二有效点之间的距离大于或者等于预设的阈值，则确定所述第二有效点为突变点；
    若所述第一有效点与所述第三有效点的连线与所述第二有效点之间的距离小于预设的阈值，则确定所述第二有效点不是突变点。
  7. 根据权利要求2至4中任一项所述的方法，其特征在于，所述方法还包括：
    将所述第i行点云中有效点投影到平面上；
    根据在所述平面上方位角最大的第一有效点和方位角最小的第二有效点的第一连线 确定与所述第一连线距离最远的第三有效点；
    若所述第三有效点与所述第一直线的距离大于或者等于第三阈值，则确定与所述第三有效点为第一突变点，在所述平面上除所述第一有效点、第二有效点、第三有效点之外的有效点中，分别确定与第一有效点和第三有效点之间的第二连线、以及与第二有效点和第三有效点之间的第三连线距离最远的第四有效点和第五有效点，若所述第四有效点与所述第二连线的距离大于或者等于所述第三阈值，则所述第i行点云中的突变点个数增加1个，若所述第五有效点与所述第三连线的距离大于或者等于所述第三阈值，则所述第i行点云中的突变点个数增加1个；重复上述过程，直到在所述平面上没有新的直线产生为止，累加得到所述第i行点云中的突变点个数；
    若所述第三有效点与所述第一直线的距离小于第三阈值，则所述第i行点云中的突变点个数为0。
  8. 根据权利要求3所述的方法，其特征在于，所述点云网格中每一列点云的不连续度是根据所述每一列点云中包括的有效点个数确定的。
  9. 根据权利要求8所述的方法，其特征在于，所述方法还包括：
    若所述点云网格中第j列点云中有效点数为0，则确定所述第j列点云的不连续度为a，j为正整数，a为正数；
    若所述点云网格中第j列点云中有效点数为1，且所述第j列为所述点云网格中的最后一列或者第一列，则确定所述第j列点云的不连续度为0，j为正整数；
    若所述点云网格中第j列点云中有效点数为1，且所述点云网格中第j+1列或者第j-1列点云中有效点数大于1，则确定所述第j列点云的不连续度为b，j为大于1的整数，b为正数；
    若所述点云网格中第j列点云中有效点数为大于1，则确定所述第j列点云中有效点对应的行坐标是否连续，如果连续，则确定所述第j列点云的不连续度为0，如果不连续，则确定所述第j列点云的不连续度为c，j为大于1的整数，c为正数。
  10. 根据权利要求1至8中任一项所述的方法，其特征在于，
    所述目标激光点云由单个激光雷达单帧检测获得。
  11. 一种目标识别装置，其特征在于，包括：
    收发单元，用于接收目标激光点云；
    处理单元，用于生成所述目标激光点云对应的点云网格，其中，所述目标激光点云中的一个点对应所述点云网格中的一个有效点；
    所述处理单元，还用于根据所述点云网格的几何特征，判断所述目标激光点云是否为噪声点云，所述点云网格的几何特征包括：所述点云网格的微观粗糙度，和/或，所述点云网格中点云的分布不连续度。
  12. 根据权利要求11所述的装置，其特征在于，
    所述点云网格的微观粗糙度，为所述点云网格中的突变点总数占所述点云网格中包括的有效点数的比例；
    其中，所述点云网格中的突变点总数，为所述点云网格中每一行中的突变点数之和。
  13. 根据权利要求11或12所述的装置，其特征在于，
    所述点云网格中点云的分布不连续度，为所述点云网格中每一列点云的不连续度之 和。
  14. 根据权利要求11至13中任一项所述的装置，其特征在于，
    所述处理单元还用于：在所述点云网格的微观粗糙度大于或者等于第一阈值的情况下，确定所述目标激光点云为噪声点云；和/或，
    在所述点云网格中点云的分布不连续度大于或者等于第二阈值的情况下，判断所述目标激光点云为噪声点云。
  15. 根据权利要求12至14中任一项所述的装置，其特征在于，所述处理单元还用于：
    在所述点云网格中第i行点云中有效点数小于或者等于2的情况下，判断所述点云网格中第i行的突变点数为0，i为正整数；
    在所述点云网格中第i行点云中有效点数大于或者等于3的情况下，则在所述第i行点云中连续的三个有效点确定所述连续的三个有效点中的第二有效点是否为突变点，i为正整数；
    根据所述第i行点云中每连续的三个有效点中的第二个有效点是否为突变点，确定所述点云网格中第i行的突变点数。
  16. 根据权利要求15所述的装置，所述处理单元还用于：
    若所述点云网格中第i行点云中有效点数大于或者等于3，所述第i行点云中连续的三个有效点分别为第一有效点、第二有效点、第三有效点，若所述第一有效点与所述第三有效点的连线与所述第二有效点之间的距离大于或者等于预设的阈值，则确定所述第二有效点为突变点；
    若所述第一有效点与所述第三有效点的连线与所述第二有效点之间的距离小于预设的阈值，则确定所述第二有效点不是突变点。
  17. 根据权利要求12至14中任一项所述装置，其特征在于，所述处理单元还用于：
    将所述第i行点云中有效点投影到平面上；
    根据在所述平面上方位角最大的第一有效点和方位角最小的第二有效点的第一连线确定与所述第一连线距离最远的第三有效点；
    若所述第三有效点与所述第一直线的距离大于或者等于第三阈值，则确定与所述第三有效点为第一突变点，在所述平面上除过所述第一有效点、第二有效点、第三有效点之外的有效点中，分别确定与第一有效点和第三有效点的之间的第二连线、以及与第二有效点和第三有效点的之间的第三连线距离最远的第四有效点和第五有效点，若所述第四有效点与所述第二连线的距离大于或者等于所述第三阈值，则所述第i行点云中的突变点个数增加1个，若所述第五有效点与所述第三连线的距离大于或者等于所述第三阈值，则所述第i行点云中的突变点个数增加1个；重复上述过程，直到在所述平面上没有新的直线产生为止，累加得到所述第i行点云中的突变点个数；
    若所述第三有效点与所述第一直线的距离小于第三阈值，则所述第i行点云中的突变点个数为0。
  18. 根据权利要求13至17中任一项所述的装置，其特征在于，所述处理单元还用于：
    根据所述每一列点云中包括的有效点个数确定所述点云网格中每一列点云的不连续度。
  19. 根据权利要求18所述的装置，其特征在于，所述处理单元还用于：
    在所述点云网格中第j列点云中有效点数为0的情况下，确定所述第j列点云的不连续度为a，j为正整数，a为正数；
    在所述点云网格中第j列点云中有效点数为1，且所述第j列为所述点云网格中的最后一列或者第一列的情况下，确定所述第j列点云的不连续度为0，j为正整数；
    在所述点云网格中第j列点云中有效点数为1，且所述点云网格中第j+1列或者第j-1列点云中有效点数大于1的情况下，确定所述第j列点云的不连续度为b，j为大于1的整数，b为正数；
    若所述点云网格中第j列点云中有效点数为大于1，则确定所述第j列点云中有效点对应的行坐标是否连续，如果连续，则确定所述第j列点云的不连续度为0，如果不连续，则确定所述第j列点云的不连续度为c，j为大于1的整数，c为正数。
  20. 根据权利要求11至19中任一项所述的装置，其特征在于，
    所述目标激光点云由单个激光雷达单帧检测获得。
  21. 一种目标识别的装置，其特征在于，所述装置包括至少一个处理器，所述至少一个处理器与至少一个存储器耦合：
    所述至少一个处理器，用于执行所述至少一个存储器中存储的计算机程序或指令，以使得所述装置执行如权利要求1至10中任一项所述的目标识别的方法。
  22. 一种激光雷达，其特征在于，包括权利要求11至20中任一项所述的目标识别的装置，或者，包括权利要求21所述的目标识别的装置。
  23. 一种芯片系统，其特征在于，所述芯片系统应用于激光雷达；所述芯片系统包括一个或多个接口电路和一个或多个处理器；所述接口电路和所述处理器通过线路互联；所述处理器用于执行如权利要求1-10中任一项所述的目标识别的方法。
  24. 一种计算机可读存储介质，其特征在于，所述计算机可读存储介质中存储有计算机程序或指令，当计算机读取并执行所述计算机程序或指令时，使得计算机执行如权利要求1至10中任一项所述的目标识别方法。
PCT/CN2020/084818 2020-04-15 2020-04-15 一种目标识别的方法和装置 WO2021207954A1 (zh)

Priority Applications (3)

Application Number Priority Date Filing Date Title
EP20930782.6A EP4130798A4 (en) 2020-04-15 2020-04-15 TARGET IDENTIFICATION METHOD AND DEVICE
PCT/CN2020/084818 WO2021207954A1 (zh) 2020-04-15 2020-04-15 一种目标识别的方法和装置
CN202080004293.4A CN112513679B (zh) 2020-04-15 2020-04-15 一种目标识别的方法和装置

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/CN2020/084818 WO2021207954A1 (zh) 2020-04-15 2020-04-15 一种目标识别的方法和装置

Publications (1)

Publication Number Publication Date
WO2021207954A1 true WO2021207954A1 (zh) 2021-10-21

Family

ID=74953149

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2020/084818 WO2021207954A1 (zh) 2020-04-15 2020-04-15 一种目标识别的方法和装置

Country Status (3)

Country Link
EP (1) EP4130798A4 (zh)
CN (1) CN112513679B (zh)
WO (1) WO2021207954A1 (zh)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114384492A (zh) * 2022-03-24 2022-04-22 北京一径科技有限公司 用于激光雷达的点云处理方法及装置、存储介质
CN114488091A (zh) * 2022-01-25 2022-05-13 北京三快在线科技有限公司 一种激光点云的去噪方法、装置、存储介质及电子设备
CN115337430A (zh) * 2022-08-11 2022-11-15 深圳市隆瑞科技有限公司 一种喷雾小车的控制方法和装置

Families Citing this family (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113031010B (zh) * 2021-03-31 2023-04-28 小马易行科技(上海)有限公司 检测天气的方法、装置、计算机可读存储介质与处理器
CN112749504B (zh) * 2021-04-02 2021-06-22 中智行科技有限公司 仿真扫描点的获取方法、装置、电子设备及存储介质
CN113763263A (zh) * 2021-07-27 2021-12-07 华能伊敏煤电有限责任公司 基于点云尾气过滤技术的水雾尾气噪声处理方法
CN115271096B (zh) * 2022-07-27 2023-08-18 阿波罗智能技术(北京)有限公司 点云处理和机器学习模型训练方法、装置及自动驾驶车辆
CN116299254B (zh) * 2022-09-07 2023-10-13 无锡国芯微电子系统有限公司 一种被动雷达寻的器的目标跟踪方法

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104156926A (zh) * 2014-08-19 2014-11-19 武汉海达数云技术有限公司 多场景下车载激光点云噪声点去除方法
CN107818550A (zh) * 2017-10-27 2018-03-20 广东电网有限责任公司机巡作业中心 一种基于LiDAR的点云顶部噪点剔除方法
CN108181635A (zh) * 2017-12-11 2018-06-19 中国南方电网有限责任公司超高压输电公司广州局 一种用于输电线路交叉跨越分析的激光点云分类方法
CN108919295A (zh) * 2018-05-15 2018-11-30 国网通用航空有限公司 机载LiDAR点云道路信息提取方法及装置
US20190178989A1 (en) * 2017-12-11 2019-06-13 Automotive Research & Testing Center Dynamic road surface detecting method based on three-dimensional sensor
CN110458780A (zh) * 2019-08-14 2019-11-15 上海眼控科技股份有限公司 3d点云数据去噪方法、装置、计算机设备和可读存储介质

Family Cites Families (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107742091B (zh) * 2016-08-22 2019-01-29 腾讯科技（深圳）有限公司 一种路肩提取的方法及装置
CN106842231B (zh) * 2016-11-08 2019-03-22 长安大学 一种道路边界检测及跟踪方法
US10241198B2 (en) * 2017-03-30 2019-03-26 Luminar Technologies, Inc. Lidar receiver calibration
WO2019112514A1 (en) * 2017-12-08 2019-06-13 St Engineering Land Systems Ltd. Rain filtering techniques for autonomous vehicle
CN109444847A (zh) * 2018-11-01 2019-03-08 肖湘江 机器人激光雷达避障的噪点滤波方法
CN110988846B (zh) * 2019-04-22 2023-07-18 威力登激光雷达美国有限公司 可用于激光雷达的噪点识别方法以及激光雷达系统
CN110443275B (zh) * 2019-06-28 2022-11-25 炬星科技（深圳）有限公司 去除噪声的方法、设备及存储介质
CN110515054B (zh) * 2019-08-23 2021-07-23 斯坦德机器人（深圳）有限公司 滤波方法和装置、电子设备、计算机存储介质

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104156926A (zh) * 2014-08-19 2014-11-19 武汉海达数云技术有限公司 多场景下车载激光点云噪声点去除方法
CN107818550A (zh) * 2017-10-27 2018-03-20 广东电网有限责任公司机巡作业中心 一种基于LiDAR的点云顶部噪点剔除方法
CN108181635A (zh) * 2017-12-11 2018-06-19 中国南方电网有限责任公司超高压输电公司广州局 一种用于输电线路交叉跨越分析的激光点云分类方法
US20190178989A1 (en) * 2017-12-11 2019-06-13 Automotive Research & Testing Center Dynamic road surface detecting method based on three-dimensional sensor
CN108919295A (zh) * 2018-05-15 2018-11-30 国网通用航空有限公司 机载LiDAR点云道路信息提取方法及装置
CN110458780A (zh) * 2019-08-14 2019-11-15 上海眼控科技股份有限公司 3d点云数据去噪方法、装置、计算机设备和可读存储介质

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
See also references of EP4130798A4 *

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114488091A (zh) * 2022-01-25 2022-05-13 北京三快在线科技有限公司 一种激光点云的去噪方法、装置、存储介质及电子设备
CN114488091B (zh) * 2022-01-25 2024-05-24 北京三快在线科技有限公司 一种激光点云的去噪方法、装置、存储介质及电子设备
CN114384492A (zh) * 2022-03-24 2022-04-22 北京一径科技有限公司 用于激光雷达的点云处理方法及装置、存储介质
CN115337430A (zh) * 2022-08-11 2022-11-15 深圳市隆瑞科技有限公司 一种喷雾小车的控制方法和装置
CN115337430B (zh) * 2022-08-11 2024-06-07 深圳市隆瑞科技有限公司 一种喷雾小车的控制方法和装置

Also Published As

Publication number Publication date
EP4130798A1 (en) 2023-02-08
EP4130798A4 (en) 2023-05-31
CN112513679B (zh) 2021-12-03
CN112513679A (zh) 2021-03-16

Similar Documents

Publication Publication Date Title
WO2021207954A1 (zh) 一种目标识别的方法和装置
WO2022022694A1 (zh) 自动驾驶环境感知方法及系统
CN110764108B (zh) 一种用于港口自动驾驶场景的障碍物检测方法及装置
US20210350149A1 (en) Lane detection method and apparatus,lane detection device,and movable platform
CN110782465B (zh) 一种基于激光雷达的地面分割方法、装置及存储介质
CN111712731A (zh) 目标检测方法、系统及可移动平台
WO2021051281A1 (zh) 点云滤噪的方法、测距装置、系统、存储介质和移动平台
CN112698301B (zh) 基于距离强度关联的激光雷达目标识别方法、装置及设备
WO2022198637A1 (zh) 点云滤噪方法、系统和可移动平台
CN112083441A (zh) 激光雷达和毫米波雷达深度融合的障碍物检测方法及系统
CN113743171A (zh) 目标检测方法及装置
WO2023040737A1 (zh) 目标位置确定方法、装置、电子设备以及存储介质
CN114821526A (zh) 基于4d毫米波雷达点云的障碍物三维边框检测方法
CN114550142A (zh) 基于4d毫米波雷达和图像识别融合的车位检测方法
CN115147333A (zh) 一种目标检测方法及装置
CN113296120B (zh) 一种障碍物检测方法及终端
WO2021179583A1 (zh) 探测方法及探测设备
CN113734176A (zh) 智能驾驶车辆的环境感知系统、方法、车辆及存储介质
CN109917418B (zh) 一种激光雷达无反射区域的测量方法
CN113313654B (zh) 激光点云滤波去噪方法、系统、设备及存储介质
CN114080545A (zh) 数据处理方法、装置、激光雷达和存储介质
TWI843116B (zh) 移動物體檢測方法、裝置、電子設備及存儲介質
CN116228603B (zh) 一种挂车周围障碍物的报警系统及装置
CN116740581B (zh) 确定材质识别模型的方法、回到基站的方法及电子设备
WO2024065685A1 (zh) 一种处理点云的方法和雷达

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 20930782

Country of ref document: EP

Kind code of ref document: A1

ENP Entry into the national phase

Ref document number: 2020930782

Country of ref document: EP

Effective date: 20221102

NENP Non-entry into the national phase

Ref country code: DE