A point cloud data filtering method and device, a terminal device, and a storage medium

Abstract

The application is suitable for the automatic driving technical field, and provides a point cloud data filtering method and device, a terminal equipment and a storage medium. The point cloud data filtering method comprises the following steps: receiving a plurality of groups of point cloud data; wherein each group of point cloud data comprises point cloud data collected under a first view angle and point cloud data collected under a second view angle; acquiring a point cloud coincidence range, and according to the point cloud coincidence range, a group of target point cloud data is respectively cut out from each group of point cloud data; extracting point cloud features of the target point cloud data; if it is confirmed that ghosting exists in a corresponding group of point cloud data according to the point cloud features of the target point cloud data, the corresponding group of point cloud data is filtered out. The above scheme effectively improves the efficiency and accuracy of point cloud data filtering, improves the quality of point cloud data, and is beneficial to more accurately reconstructing an automatic driving scene.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of automatic driving, and particularly relates to a point cloud data filtering method and device, a terminal device and a storage medium. BACKGROUND

[0002] With the continuous development of computer hardware and software technology, more and more enterprises begin to invest in the research and development of automatic driving systems. Among them, point cloud data is an important data in the automatic driving system. Compared with conventional image data, point cloud data can contain more entity information, thereby being more conducive to reconstructing an automatic driving scene.

[0003] At present, a Lidar (Light Detection and Ranging) is generally arranged on an automatic driving device to collect point cloud data to realize the reconstruction of an automatic driving scene. In order to make the scene display more comprehensive, more than one Lidar is often arranged on the automatic driving device, and then the point cloud data collected by the Lidars is sequentially subjected to coordinate system conversion and splicing to obtain a more complete point cloud data for three-dimensional reconstruction.

[0004] However, due to the error of the internal and external parameters of the Lidar, ghosting problems are likely to occur in the splicing process, affecting the rendering effect of the automatic driving scene. SUMMARY

[0005] The embodiments of the present application provide a point cloud data filtering method, device, terminal device and storage medium, which can solve the above technical problems.

[0006] In a first aspect, the embodiments of the present application provide a point cloud data filtering method, comprising: receiving a plurality of groups of point cloud data; wherein each group of point cloud data comprises point cloud data collected at a first view angle and point cloud data collected at a second view angle; obtaining a point cloud overlap range, and according to the point cloud overlap range, a group of target point cloud data is respectively cut out from each group of point cloud data; extracting point cloud features of the target point cloud data; if it is confirmed that ghosting exists in a corresponding group of point cloud data according to the point cloud features of the target point cloud data, the corresponding group of point cloud data is filtered out.

[0007] Further, the obtaining of the point cloud overlap range comprises: obtaining a data collection range at the first view angle, a data collection range at the second view angle and a calibration parameter of a data collection device; and according to the data collection range at the first view angle, the data collection range at the second view angle and the calibration parameter of the data collection device, the point cloud overlap range is obtained.

[0008] Further, the target point cloud data includes first target point cloud data and second target point cloud data, and the extracting the point cloud feature of the target point cloud data includes: extracting a first mean value feature and a first variance feature from the first target point cloud data; extracting a second mean value feature and a second variance feature from the second target point cloud data; and extracting a centroid distance feature according to a first centroid of the first target point cloud data and a second centroid of the second target point cloud data.

[0009] Further, before the filtering out the corresponding group of point cloud data when it is confirmed that the corresponding group of point cloud data has ghosting according to the point cloud feature of the target point cloud data, the method further includes: fusing the point cloud feature of the target point cloud data to obtain a point cloud fusion feature; and obtaining a point cloud fusion feature threshold, and confirming that the corresponding group of point cloud data has ghosting when the point cloud fusion feature exceeds the point cloud fusion feature threshold; wherein the point cloud fusion feature threshold is determined according to a standard deviation of a normal distribution, and the normal distribution is a distribution of a plurality of point cloud fusion features.

[0010] Further, the point cloud feature includes a first mean value feature, a first variance feature, a second mean value feature, a second variance feature and a centroid distance feature, and the fusing the point cloud feature of the target point cloud data to obtain a point cloud fusion feature includes: obtaining a mean value difference feature between the first mean value feature and the second mean value feature, and a variance difference feature between the first variance feature and the second variance feature; and obtaining the point cloud fusion feature according to the mean value difference feature, a preset mean value fusion parameter, the variance difference feature, a preset variance fusion parameter, the centroid distance feature and a preset centroid distance fusion parameter.

[0011] Further, the group of point cloud data is stored in a PCD format file, and the extracting a group of target point cloud data from each group of point cloud data according to the point cloud coincidence range includes: extracting a group of target point cloud data from each PCD format file according to the point cloud coincidence range; and the filtering out the corresponding group of point cloud data includes: filtering out a corresponding PCD format file.

[0012] In a second aspect, the embodiments of the present application provide a point cloud data filtering device, comprising: a receiving unit configured to receive a plurality of groups of point cloud data; wherein each group of the point cloud data comprises point cloud data corresponding to a first view angle and point cloud data corresponding to a second view angle; a cutting unit configured to obtain a point cloud overlap range, and cut a group of target point cloud data from each group of the point cloud data according to the point cloud overlap range; an extracting unit configured to extract point cloud features of the target point cloud data; and a filtering unit configured to filter out a corresponding group of point cloud data if it is confirmed that there is a ghosting in the corresponding group of point cloud data according to the point cloud features of the target point cloud data.

[0013] Further, the cutting unit comprises: a first obtaining unit configured to obtain a data acquisition range under a first view angle, a data acquisition range under a second view angle, and calibration parameters of a data acquisition device; and a second obtaining unit configured to obtain the point cloud overlap range according to the data acquisition range under the first view angle, the data acquisition range under the second view angle, and the calibration parameters of the data acquisition device.

[0014] Further, each group of the target point cloud data comprises first target point cloud data and second target point cloud data, the extracting unit comprises: a first extracting unit configured to extract a first mean feature and a first variance feature from the first target point cloud data; a second extracting unit configured to extract a second mean feature and a second variance feature from the second target point cloud data; and a third extracting unit configured to extract a centroid distance feature according to a first centroid of the first target point cloud data and a second centroid of the second target point cloud data.

[0015] Further, the point cloud data filtering device further comprises: a fusion unit configured to fuse the point cloud features of the target point cloud data to obtain point cloud fusion features; and a judging unit configured to obtain a point cloud fusion feature threshold, and confirm that there is a ghosting in the corresponding group of point cloud data if the point cloud fusion features exceed the point cloud fusion feature threshold; wherein the point cloud fusion feature threshold is determined according to a standard deviation of a normal distribution, and the normal distribution is a distribution presented by a plurality of point cloud fusion features.

[0016] Further, the point cloud features comprise a first mean feature, a first variance feature, a second mean feature, a second variance feature, and a centroid distance feature, and the fusion unit is specifically configured to: obtain a mean difference feature between the first mean feature and the second mean feature, and a variance difference feature between the first variance feature and the second variance feature; and obtain the point cloud fusion features according to the mean difference feature, a preset mean fusion parameter, the variance difference feature, a preset variance fusion parameter, the centroid distance feature, and a preset centroid distance fusion parameter.

[0017] Further, the intercepting unit is specifically configured to: according to the point cloud coincidence range, take out a group of the target point cloud data from each of the PCD format files; and the filtering unit is specifically configured to: filter out the corresponding one of the PCD format files.

[0018] In a third aspect, an embodiment of the present application provides a terminal device, which comprises a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor implements the method in the first aspect when executing the computer program.

[0019] In a fourth aspect, an embodiment of the present application provides a computer readable storage medium, which stores a computer program, and the computer program is executable on a processor to implement the method in the first aspect.

[0020] In the embodiment of the present application, the device receives a plurality of groups of point cloud data, wherein each group of point cloud data comprises point cloud data acquired at a first view angle and point cloud data acquired at a second view angle; a point cloud coincidence range is obtained, and according to the point cloud coincidence range, a group of target point cloud data is intercepted from each group of point cloud data; point cloud features of the target point cloud data are extracted; if it is confirmed that ghosting exists in the corresponding group of point cloud data according to the point cloud features of the target point cloud data, the corresponding group of point cloud data is filtered out. The above method can intercept data within the point cloud coincidence range from the point cloud data acquired at the first view angle and the point cloud data acquired at the second view angle, so as to obtain a group of target point cloud data, and by extracting the point cloud features of the group of target point cloud data, it is determined whether ghosting exists in the corresponding group of point cloud data, and then the point cloud data with ghosting is filtered out, which effectively improves the efficiency and accuracy of point cloud data filtering, improves the quality of point cloud data, and is more conducive to accurately reconstructing an autonomous driving scene. BRIEF DESCRIPTION OF DRAWINGS

[0021] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can also be obtained by those skilled in the art without any creative labor.

[0022] Figure 1 is a schematic flow chart of a point cloud data filtering method provided by the first embodiment of the present application;

[0023] Figure 2 is a schematic flow chart of S102 in a point cloud data filtering method provided by the first embodiment of the present application;

[0024] Figure 3is a schematic flowchart of S103 in a point cloud data filtering method provided by a first embodiment of the present application;

[0025] Figure 4 is another schematic flowchart of a point cloud data filtering method provided by the first embodiment of the present application;

[0026] Figure 5 is a schematic flowchart of S105 in a point cloud data filtering method provided by the first embodiment of the present application;

[0027] Figure 6 is a schematic diagram of a point cloud data filtering apparatus provided by a second embodiment of the present application;

[0028] Figure 7 is a schematic diagram of a terminal device provided by a third embodiment of the present application. DETAILED DESCRIPTION

[0029] In the following description, specific details are set forth, such as particular system configurations, techniques, etc., in order to provide a thorough understanding of the embodiments of the present application. However, persons having ordinary skill in the art will readily understand that the present application can be practiced without these specific details. In other instances, well-known structures, devices, circuits, and materials have not been described in detail in order to avoid obscuring the application.

[0030] It should be understood that the term "comprises / comprising" when used in this specification and the appended claims specifies the presence of stated features, integers, steps, operations, elements, and / or components, but does not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof.

[0031] It should also be understood that the term "and / or" when used in this specification and the appended claims, such as in the phrases "A and / or B" and "A and / or B and / or C", means any combination of one or more of the associated listed items and can be interpreted as "one of A or B" or "one of A, B, or C" or "one of A, B, and C" or "one of A, B, C, and / or D" and so on.

[0032] As used in this specification and the appended claims, the term "if' can be construed to mean "when" or "upon" or "in response to determining" or "in response to detecting", depending on the context. Similarly, the phrase "if it is determined" or "if [a described condition or event] is detected" can be construed to mean "upon determining" or "in response to determining" or "upon detecting [the described condition or event]" or "in response to detecting [the described condition or event]", depending on the context.

[0033] In addition, in the description of the specification and the appended claims of the present application, the terms "first", "second", "third", etc. are used only to distinguish descriptions, and cannot be understood as indicating or implying relative importance.

[0034] Reference within the specification to "one embodiment" or "some embodiments" means that a particular feature, structure, or characteristic described in connection with the embodiment is included in at least one embodiment. The appearances of the phrase "in one embodiment" or "in some embodiments" in various places within specified

[0035] See Figure 1 , Figure 1 is a schematic flow chart of a point cloud data filtering method provided by the first embodiment of the present application. In the embodiment, the execution subject of the point cloud data filtering method is a device with a point cloud data filtering function. The device can be an autonomous driving device (for example, a car), a personal computer, a server, or the like, and can also be a processor or a microprocessor. The embodiments of the present application take the device as the execution subject of the point cloud data filtering method for the following explanation and description, and do not specifically limit the device. As shown in the point cloud data filtering method, the method can include the following steps. Figure 1

[0036] S101: receiving a plurality of groups of point cloud data; wherein each group of point cloud data includes point cloud data collected at a first view angle and point cloud data collected at a second view angle.

[0037] The device receives a plurality of groups of point cloud data.

[0038] In an optional embodiment, the point cloud data can refer to visual point cloud data. In another optional embodiment, the point cloud data can also refer to laser point cloud data.

[0039] In the embodiments of the present application, each group of point cloud data includes point cloud data collected at a first view angle and point cloud data collected at a second view angle.

[0040] Specifically, in an autonomous driving scenario, if the point cloud data is laser point cloud data, two laser radars can be arranged on the autonomous driving device, and the device establishes a data connection with the two laser radars. One laser radar collects point cloud data at a first view angle, and the other laser radar collects point cloud data at a second view angle, so as to obtain a plurality of groups of point cloud data through the laser radars.

[0041] ​In one embodiment, one laser radar can be arranged at the left side of the autonomous driving device, thus the first view angle can refer to the left side view angle, and another laser radar can be arranged at the right side of the autonomous driving device, thus the second view angle can refer to the right side view angle.

[0042] It can be understood that the number of arranged laser radars can be reasonably adjusted based on the data acquisition range of each laser radar and the required driving environment range to be acquired, and then the point cloud data in each group can include point cloud data acquired under several view angles.

[0043] It can also be understood that if the point cloud data is visual point cloud data, the laser radars can be replaced by depth cameras, and the specific arrangement manner is the same as the foregoing.

[0044] In the embodiments of the present application, each group of point cloud data is stored in a PCD file, and each PCD file includes a file header which can be used to indicate information related to the point cloud data, specifically including: PCD file version, dimensions that points can have and field names corresponding to each dimension, data type corresponding to each dimension (such as float type, double float type, etc.), number of elements contained in each dimension, size of point cloud data (ordered point cloud contains width of point cloud and height of point cloud, etc., unordered point cloud contains number of points in point cloud), acquisition view point of point cloud data, and data type of stored point cloud data, etc.

[0045] S102: Acquire a point cloud coincidence range, and according to the point cloud coincidence range, respectively cut out a group of target point cloud data from each group of point cloud data.

[0046] The device acquires a point cloud coincidence range, and according to the point cloud coincidence range, respectively cut out a group of target point cloud data from each group of point cloud data.

[0047] The point cloud coincidence range refers to a range in which point cloud coincidence occurs between point cloud data under the first view angle and point cloud data under the second view angle.

[0048] Optionally, please refer to Figure 2 , Figure 2 is a schematic flowchart of S102 in a point cloud data filtering method provided by the first embodiment of the present application, and S102 includes:

[0049] S1021: Acquire a data acquisition range under a first view angle, a data acquisition range under a second view angle, and calibration parameters of a data acquisition device.

[0050] S1022: According to the data acquisition range under the first view angle, the data acquisition range under the second view angle, and the calibration parameters of the data acquisition device, acquire the point cloud coincidence range.

[0051] The device obtains a data acquisition range in the first perspective, a data acquisition range in the second perspective, and a calibration parameter of the data acquisition device.

[0052] For example, one laser radar (laser radar A) acquires point cloud data in the first perspective, and another laser radar (laser radar B) acquires point cloud data in the second perspective, and the data acquisition ranges of both are within 150 m in front and within 40 m left and right.

[0053] When acquiring laser point cloud, the data acquisition device is the laser radar, and the calibration parameter of the data acquisition device is the calibration parameter of the laser radar, which specifically includes an intrinsic matrix and an extrinsic matrix of the laser radar.

[0054] In this example, the calibration parameter of the laser radar A includes an intrinsic matrix and an extrinsic matrix of the laser radar A, and the calibration parameter of the laser radar B includes an intrinsic matrix and an extrinsic matrix of the laser radar B.

[0055] The device can perform coordinate system transformation on the data acquisition range in the first perspective according to the data acquisition range in the first perspective and the intrinsic matrix and the extrinsic matrix of the laser radar A, and perform coordinate system transformation on the data acquisition range in the second perspective according to the data acquisition range in the second perspective and the intrinsic matrix and the extrinsic matrix of the laser radar B, so as to convert the data acquisition ranges in the two perspectives to the same coordinate system.

[0056] Further, the device can obtain a point cloud overlap range between the point cloud data acquired in the first perspective and the point cloud data acquired in the second perspective after the laser radar A and the laser radar B acquire a group of point cloud data.

[0057] Suppose the obtained point cloud overlap range is x∈(10, 50), y∈(-5, 5), that is, the x-axis coordinate value of the point is limited to ∈(10, 50) in the x-axis direction, and the y-axis coordinate value of the point is limited to ∈(-5, 5) in the y-axis direction. The counting unit of the point cloud overlap unit is not limited here, such as m (meter).

[0058] It can be confirmed that the point cloud overlap range is unchanged after the arrangement positions and data acquisition perspectives of the two laser radars are fixed. Therefore, the device can extract a group of target point cloud data from each group of point cloud data according to the point cloud overlap range.

[0059] As described above, a group of point cloud data is stored in a PCD format file, so the device can extract a group of target point cloud data from each group of point cloud data according to the point cloud overlap range, which can be understood as the device extracting a group of target point cloud data from each PCD format file according to the point cloud overlap range.

[0060] In the embodiment, the target point cloud data set includes first target point cloud data and second target point cloud data. Then, the first target point cloud data can be composed of points in the point cloud overlap range in the point cloud data collected at the first view angle, and the second target point cloud data can be composed of points in the point cloud overlap range in the point cloud data collected at the second view angle.

[0061] S103: Extract the point cloud feature of the target point cloud data.

[0062] The device extracts the point cloud feature in the target point cloud data.

[0063] In an optional implementation, the device can extract the point cloud feature in the target point cloud data through a pre-trained feature extraction network.

[0064] In another optional implementation, please refer to Figure 3 , Figure 3 is a schematic flowchart of S103 in a point cloud data filtering method provided by the first embodiment of the application, and S103 includes:

[0065] S1031: Extract the first mean feature and the first variance feature from the first target point cloud data.

[0066] The device extracts the first mean feature and the first variance feature from the first target point cloud data.

[0067] The first mean feature includes the mean of the values of all points in the first target point cloud data in each dimension, and the first variance feature includes the variance of the values of all points in the first target point cloud data in each dimension.

[0068] S1032: Extract the second mean feature and the second variance feature from the second target point cloud data.

[0069] The device extracts the second mean feature and the second variance feature from the second target point cloud data.

[0070] The second mean feature includes the mean of the values of all points in the second target point cloud data in each dimension, and the second variance feature includes the variance of the values of all points in the second target point cloud data in each dimension.

[0071] S1033: Extract the centroid distance feature according to the first centroid of the first target point cloud data and the second centroid of the second target point cloud data.

[0072] The device extracts the centroid distance feature according to the first centroid of the first target point cloud data and the second centroid of the second target point cloud data.

[0073] The first centroid of the first target point cloud data can refer to a point corresponding to the first mean feature (the point does not necessarily exist in the first target point cloud data), the second centroid of the second target point cloud data can refer to a point corresponding to the second mean feature (the point does not necessarily exist in the second target point cloud data), and the centroid distance feature refers to the distance between the first centroid and the second centroid.

[0074] The first centroid of the first target point cloud data and the second centroid of the second target point cloud data can also be obtained by other existing centroid calculation methods.

[0075] If it is confirmed according to the point cloud feature of the target point cloud data that there is a ghost in the corresponding group of point cloud data, the corresponding group of point cloud data is filtered out.

[0076] If it is confirmed according to the point cloud feature of the target point cloud data that there is a ghost in the corresponding group of point cloud data, the corresponding group of point cloud data is filtered out.

[0077] If the point cloud feature of the target point cloud data is extracted according to the pre-trained feature extraction network, the point cloud feature of the target point cloud data can be input into the pre-trained ghost judgment network to obtain a judgment result output by the pre-trained ghost judgment network. If the judgment result indicates that there is a ghost in the corresponding group of point cloud data, the corresponding group of point cloud data is filtered out.

[0078] The following describes how to confirm whether there is a ghost when the point cloud feature of the target point cloud data refers to the first mean feature, the first variance feature, the second mean feature, the second variance feature, and the centroid distance feature. Please refer to Figure 4 , Figure 4 is another schematic flowchart of a point cloud data filtering method provided by the first embodiment of the present application. Before S104, it includes:

[0079] S105: Fusing the point cloud feature of the target point cloud data to obtain a point cloud fusion feature.

[0080] The device fuses the point cloud feature of the target point cloud data to obtain a point cloud fusion feature.

[0081] The point cloud feature of the target point cloud data refers to the first mean feature, the first variance feature, the second mean feature, the second variance feature, and the centroid distance feature. The specific meanings are not repeated.

[0082] In an optional embodiment, please refer to Figure 5 , Figure 5 is a schematic flowchart of S105 in a point cloud data filtering method provided by the first embodiment of the present application. S105 includes:

[0083] S1051: Obtain a mean difference feature between the first mean feature and the second mean feature, and a variance difference feature between the first variance feature and the second variance feature.

[0084] The device obtains a mean difference feature between the first mean feature and the second mean feature, and a variance difference feature between the first variance feature and the second variance feature.

[0085] The mean difference feature is an absolute value of a difference between the first mean feature and the second mean feature, and is represented as abs(left_lidar.mean-right_lidar.mean), where abs() is an absolute value function, left_lidar.mean is the first mean feature, and right_lidar.mean is the second mean feature.

[0086] The variance difference feature is an absolute value of a difference between the first variance feature and the second variance feature, and is represented as abs(left_lidar.var-right_lidar.var), where left_lidar.var is the first variance feature, and right_lidar.var is the second variance feature.

[0087] S1052: Obtain the point cloud fusion feature according to the mean difference feature, a preset mean fusion parameter, the variance difference feature, a preset variance fusion parameter, the centroid distance feature, and a preset centroid distance fusion parameter.

[0088] The device obtains the point cloud fusion feature according to the mean difference feature, a preset mean fusion parameter, the variance difference feature, a preset variance fusion parameter, the centroid distance feature, and a preset centroid distance fusion parameter.

[0089] Optionally, the device obtains the point cloud fusion feature by accumulating a product of the mean difference feature and a preset mean fusion parameter, a product of the variance difference feature and a preset variance fusion parameter, and a product of the centroid distance feature and a preset centroid distance fusion parameter.

[0090] The preset mean fusion parameter is represented as a, the preset variance fusion parameter is represented as b, the centroid distance feature is represented as distance(left_lidar.centroid, right_lidar.centroid), left_lidar.centroid is the first centroid, right_lidar.centroid is the second centroid, distance() is a distance function, and the preset centroid distance fusion parameter is c.

[0091] The point cloud fusion feature is represented as M, M = abs (left_lidar.mean - right_lidar.mean) * a + abs (left_lidar.var - right_lidar.var) * b + distance (left_lidar.centroid, right_lidar.centroid) * c.

[0092] Optionally, a is 0.2, b is 0.3, and c is 0.5. This is only an example, and the parameter values can be adjusted.

[0093] S106: Obtain a point cloud fusion feature threshold. If the point cloud fusion feature exceeds the point cloud fusion feature threshold, it is determined that the corresponding group of point cloud data contains a ghosting phenomenon. The point cloud fusion feature threshold is determined according to the standard deviation of a normal distribution, and the normal distribution is the distribution of a plurality of point cloud fusion features.

[0094] The device obtains a point cloud fusion feature threshold. Each group of point cloud data corresponds to a point cloud fusion feature, and a plurality of groups of point cloud data correspond to a plurality of point cloud fusion features.

[0095] The plurality of point cloud fusion features as a whole exhibit a normal distribution, and the standard deviation of the normal distribution can be obtained. The standard deviation of the normal distribution is denoted as σ.

[0096] The point cloud fusion feature threshold is determined based on the standard deviation σ of the normal distribution. For example, the point cloud fusion feature threshold can be 2σ.

[0097] If the point cloud fusion feature corresponding to a group of point cloud data exceeds the point cloud fusion feature threshold, the device determines that the group of point cloud data contains a ghosting phenomenon, and the device filters out the group of point cloud data.

[0098] For example, if the point cloud fusion feature corresponding to a group of point cloud data exceeds 2σ, the device determines that the group of point cloud data contains a ghosting phenomenon, and the device filters out the group of point cloud data.

[0099] As described above, a group of point cloud data is stored in a PCD format file, so that the filtering out of the corresponding group of point cloud data in S104 includes filtering out a PCD format file corresponding to the group of point cloud data.

[0100] It can be understood that after filtering out a part of PCD format files containing a ghosting phenomenon from a plurality of PCD format files, a plurality of filtered PCD format files are obtained. Based on the plurality of filtered PCD format files, the rendering of the autonomous driving scene is performed, which makes the rendering effect of the autonomous driving scene clearer. In addition, due to the high filtering efficiency and less required computing power, the rendering efficiency of the autonomous driving scene can also be improved.

[0101] In the embodiment of the present application, the device receives a plurality of groups of point cloud data; wherein each group of point cloud data comprises point cloud data collected at a first view angle and point cloud data collected at a second view angle; a point cloud overlap range is obtained, and according to the point cloud overlap range, a group of target point cloud data is respectively cut out from each group of point cloud data; point cloud features of the target point cloud data are extracted; if it is confirmed that ghosting exists in the corresponding group of point cloud data according to the point cloud features of the target point cloud data, the corresponding group of point cloud data is filtered out. The above method can cut out data in the point cloud overlap range from the point cloud data collected at the first view angle and the point cloud data collected at the second view angle, respectively, so as to obtain a group of target point cloud data, and by extracting the point cloud features of the group of target point cloud data, it is determined whether ghosting exists in the corresponding group of point cloud data, and then the point cloud data with ghosting is filtered out, thereby effectively improving the efficiency and accuracy of point cloud data filtering, improving the quality of point cloud data, and being more conducive to accurately reconstructing an autonomous driving scene.

[0102] It should be understood that the size of the serial number of each step in the above embodiment does not mean the order of execution, and the execution order of each process should be determined according to its function and inherent logic, and should not constitute any limitation on the implementation process of the embodiment of the present application.

[0103] Please refer to Figure 6 , Figure 6 is a schematic diagram of a point cloud data filtering device provided by the second embodiment of the present application. Each unit included is used to execute Figure 1 each step in the corresponding embodiment. For details, please refer to Figure 1 the related description in the corresponding embodiment. For the sake of illustration, only the parts related to the present embodiment are shown. Please refer to Figure 6 , the point cloud data filtering device 6 comprises:

[0104] a receiving unit 61, configured to receive a plurality of groups of point cloud data; wherein each group of point cloud data comprises point cloud data corresponding to a first view angle and point cloud data corresponding to a second view angle;

[0105] a cutting unit 62, configured to obtain a point cloud overlap range, and according to the point cloud overlap range, a group of target point cloud data is respectively cut out from each group of point cloud data;

[0106] an extracting unit 63, configured to extract point cloud features of the target point cloud data;

[0107] a filtering unit 64, configured to filter out the corresponding group of point cloud data if it is confirmed that ghosting exists in the corresponding group of point cloud data according to the point cloud features of the target point cloud data.

[0108] Further, the intercepting unit 62 comprises: a first obtaining unit, configured to obtain a data acquisition range under a first view angle, a data acquisition range under a second view angle, and a calibration parameter of a data acquisition device; and a second obtaining unit, configured to obtain the point cloud coincidence range according to the data acquisition range under the first view angle, the data acquisition range under the second view angle, and the calibration parameter of the data acquisition device.

[0109] Further, a set of the target point cloud data comprises first target point cloud data and second target point cloud data, and the extracting unit 63 comprises: a first extracting unit, configured to extract a first mean value feature and a first variance feature from the first target point cloud data; a second extracting unit, configured to extract a second mean value feature and a second variance feature from the second target point cloud data; and a third extracting unit, configured to extract a centroid distance feature according to a first centroid of the first target point cloud data and a second centroid of the second target point cloud data.

[0110] Further, the point cloud data filtering device 6 further comprises: a fusion unit, configured to fuse point cloud features of the target point cloud data to obtain point cloud fusion features; and a judging unit, configured to obtain a point cloud fusion feature threshold, and confirm that ghosting exists in the corresponding set of point cloud data if the point cloud fusion features exceed the point cloud fusion feature threshold; wherein the point cloud fusion feature threshold is determined according to a standard deviation of a normal distribution, and the normal distribution is a distribution presented by a plurality of point cloud fusion features.

[0111] Further, the point cloud features comprise a first mean value feature, a first variance feature, a second mean value feature, a second variance feature, and a centroid distance feature, and the fusion unit is specifically configured to: obtain a mean value difference feature between the first mean value feature and the second mean value feature, and a variance difference feature between the first variance feature and the second variance feature; and obtain the point cloud fusion features according to the mean value difference feature, a preset mean value fusion parameter, the variance difference feature, a preset variance fusion parameter, the centroid distance feature, and a preset centroid distance fusion parameter.

[0112] Further, the intercepting unit 62 is specifically configured to: take out a set of the target point cloud data from each PCD format file according to the point cloud coincidence range; and the filtering unit is specifically configured to: filter out a corresponding one of the PCD format files.

[0113] Please refer to Figure 7 , Figure 7 is a schematic diagram of a terminal device provided by the third embodiment of the present application. As shown in Figure 7As shown, the terminal device 7 of this embodiment includes a processor 70, a memory 71, and a computer program 72 stored in the memory 71 and executable on the processor 70, such as a point cloud data filtering program. The processor 70 implements the steps in each of the above point cloud data filtering method embodiments when executing the computer program 72, such as Figure 1 As shown, steps S101 to S104. Alternatively, the processor 70 implements the functions of each module / unit in each of the above device embodiments when executing the computer program 72, such as Figure 6 As shown, the functions of the receiving unit 61 to the filtering unit 64.

[0114] For example, the computer program 72 can be divided into one or more modules / units, which are stored in the memory 71 and executed by the processor 70 to complete the present application. The one or more modules / units can be a series of computer program instruction segments capable of completing a specific function, which are used to describe the execution process of the computer program 72 in the terminal device 7. For example, the computer program 72 can be divided into a receiving unit, a clipping unit, an extracting unit, and a filtering unit, each of which has the following specific functions:

[0115] The receiving unit is configured to receive a plurality of groups of point cloud data; wherein each group of point cloud data includes point cloud data corresponding to a first view angle and point cloud data corresponding to a second view angle;

[0116] The clipping unit is configured to obtain a point cloud overlap range, and clip a group of target point cloud data from each group of point cloud data according to the point cloud overlap range;

[0117] The extracting unit is configured to extract point cloud features of the target point cloud data;

[0118] The filtering unit is configured to filter out a corresponding group of point cloud data if it is confirmed that there is a ghosting in the corresponding group of point cloud data according to the point cloud features of the target point cloud data.

[0119] The terminal device can include, but is not limited to, the processor 70 and the memory 71. Those skilled in the art can understand that, Figure 7 The terminal device 7 is only an example and does not constitute a limitation on the terminal device 7, which can include more or fewer components than shown, or combine certain components, or different components, such as the terminal device can also include an input / output device, a network access device, a bus, etc.

[0120] The processor 70 can be a central processing unit (CPU), and can also be other general-purpose processors, a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor.

[0121] The memory 71 can be an internal storage unit of the terminal device 7, such as a hard disk or a memory of the terminal device 7. The memory 71 can also be an external storage device of the terminal device 7, such as a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, etc. Further, the terminal device 7 can also include both the internal storage unit and the external storage device. The memory 71 is used to store the computer program and other programs and data required by the terminal device. The memory 71 can also be used to temporarily store data that has been output or will be output.

[0122] It should be noted that the information interaction, execution process, etc. between the above apparatuses / units, since based on the same concept as the method embodiments, the specific functions and the brought technical effects can be referred to the method embodiments part, and will not be repeated here.

[0123] The embodiments of the present application also provide a network device, which comprises at least one processor, a memory, and a computer program stored in the memory and executable on the at least one processor, and the processor implements the steps in any of the above method embodiments when executing the computer program.

[0124] The embodiments of the present application also provide a computer readable storage medium, which stores a computer program, and the computer program is executable on a processor to implement the steps in any of the above method embodiments.

[0125] The embodiments of the present application provide a computer program product, which, when running on a mobile terminal, enables the mobile terminal to implement the steps in any of the above method embodiments.

[0126] The integrated unit, if implemented in the form of a software function unit and sold or used as an independent product, can be stored in a computer readable storage medium. Based on such understanding, the present application can implement all or part of the processes in the above-mentioned embodiment methods through a computer program to instruct relevant hardware to complete, and the computer program can be stored in a computer readable storage medium. When the computer program is executed by a processor, the steps of each method embodiment described above can be implemented. The computer program includes computer program code, which can be in the form of source code, object code, executable files or some intermediate forms. The computer readable medium at least includes any entity or device capable of carrying the computer program code to the photographing device / terminal equipment, recording medium, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signal, telecommunication signal and software distribution medium. For example, U disk, mobile hard disk, magnetic disk or optical disk, etc. In some jurisdictions, according to legislation and patent practice, the computer readable medium can not be an electrical carrier signal and a telecommunication signal.

[0127] In the above embodiments, the description of each embodiment has its own focus, and the parts not described or recorded in detail in a certain embodiment can be referred to the relevant description of other embodiments.

[0128] Those skilled in the art can appreciate that the units and algorithm steps of the examples described in combination with the embodiments disclosed herein can be realized by electronic hardware or a combination of computer software and electronic hardware. Whether the functions are realized in hardware or software depends on the specific application and design constraints of the technical solution. The skilled person can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0129] In the embodiments provided by the present application, it should be understood that the disclosed apparatus / network device and method can be implemented in other ways. For example, the above-described apparatus / network device embodiments are merely schematic, for example, the division of the modules or units is only a logical function division, and actual implementation can have another division manner, for example, a plurality of units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the coupling or direct coupling or communication connection between the displayed or discussed each other can be indirect coupling or communication connection through some interfaces, devices or units, and can be electrical, mechanical or other forms.

[0130] The units described as separate components may or may not be physically separate, and the components displayed as units may or may not be physical units, that is, may be located in one place, or may also be distributed to multiple network units. Part or all of the units can be selected to achieve the purpose of the embodiment scheme according to actual needs.

[0131] The above embodiments are only used to illustrate the technical solutions of the present application, but not limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that the technical solutions recorded in the foregoing embodiments can still be modified, or some technical features can be replaced by equivalents; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application, and should be included in the protection scope of the present application.

Claims

1. A point cloud data filtering method, characterized in that, include: Receive several sets of point cloud data; wherein each set of point cloud data includes point cloud data collected by one lidar from a first viewpoint and point cloud data collected by another lidar from a second viewpoint. Obtain the overlapping range of point clouds, and extract a set of target point cloud data from each set of point cloud data based on the overlapping range of point clouds; Extract the point cloud features from the target point cloud data; If, based on the point cloud features of the target point cloud data, it is confirmed that there is ghosting in a corresponding set of point cloud data, the corresponding set of point cloud data is filtered out. Before filtering out the corresponding set of point cloud data after confirming the presence of ghosting based on the point cloud features of the target point cloud data, the process includes: By fusing the point cloud features of the target point cloud data, a point cloud fusion feature is obtained; Obtain a point cloud fusion feature threshold. If the point cloud fusion feature exceeds the point cloud fusion feature threshold, it is confirmed that there is ghosting in the corresponding set of point cloud data. The point cloud fusion feature threshold is determined based on the standard deviation of a normal distribution, where the normal distribution is the distribution presented by several point cloud fusion features. The point cloud features include a first mean feature, a first variance feature, a second mean feature, a second variance feature, and a centroid distance feature. The point cloud features obtained by fusing the target point cloud data to obtain point cloud fusion features include: Obtain the mean difference feature between the first mean feature and the second mean feature, and the variance difference feature between the first variance feature and the second variance feature; The point cloud fusion feature is obtained based on the mean difference feature, the preset mean fusion parameter, the variance difference feature, the preset variance fusion parameter, the centroid distance feature, and the preset centroid distance fusion parameter.

2. The point cloud data filtering method as described in claim 1, characterized in that, The acquisition of the point cloud overlap range includes: Acquire the data acquisition range from the first-person perspective, the data acquisition range from the second-person perspective, and the calibration parameters of the data acquisition device; The point cloud overlap range is obtained based on the data acquisition range under the first viewpoint, the data acquisition range under the second viewpoint, and the calibration parameters of the data acquisition device.

3. The point cloud data filtering method as described in claim 1, characterized in that, The set of target point cloud data includes first target point cloud data and second target point cloud data. Extracting the point cloud features of the target point cloud data includes: Extract the first mean feature and the first variance feature from the first target point cloud data; Extract the second mean feature and the second variance feature from the second target point cloud data; Based on the first centroid of the first target point cloud data and the second centroid of the second target point cloud data, centroid distance features are extracted.

4. The point cloud data filtering method according to any one of claims 1 to 3, characterized in that, A set of point cloud data is stored in a PCD format file. The step of extracting a target point cloud data set from each set of point cloud data based on the overlapping range of the point clouds includes: Based on the point cloud overlap range, a set of target point cloud data is extracted from each of the PCD format files; The filtering out of the corresponding set of point cloud data includes: Filter out the corresponding PCD format file.

5. A point cloud data filtering device, characterized in that, include: A receiving unit is used to receive several sets of point cloud data; wherein each set of point cloud data includes point cloud data corresponding to a first viewpoint and point cloud data corresponding to a second viewpoint. The interception unit is used to obtain the overlapping range of the point cloud and, based on the overlapping range of the point cloud, to intercept a set of target point cloud data from each set of point cloud data. Extraction unit, used to extract point cloud features from the target point cloud data; The filtering unit is used to filter out the corresponding set of point cloud data if, based on the point cloud features of the target point cloud data, it is confirmed that there is ghosting in the corresponding set of point cloud data. A fusion unit is used to fuse the point cloud features of the target point cloud data to obtain point cloud fusion features; The judgment unit is used to obtain the point cloud fusion feature threshold. If the point cloud fusion feature exceeds the point cloud fusion feature threshold, it is confirmed that there is ghosting in the corresponding set of point cloud data. The point cloud fusion feature threshold is determined based on the standard deviation of a normal distribution, where the normal distribution is the distribution presented by several point cloud fusion features. The point cloud features include a first mean feature, a first variance feature, a second mean feature, a second variance feature, and a centroid distance feature. The fusion unit is specifically used for: Obtain the mean difference feature between the first mean feature and the second mean feature, and the variance difference feature between the first variance feature and the second variance feature; The point cloud fusion feature is obtained based on the mean difference feature, the preset mean fusion parameter, the variance difference feature, the preset variance fusion parameter, the centroid distance feature, and the preset centroid distance fusion parameter.

6. The point cloud data filtering device as described in claim 5, characterized in that, The interception unit includes: The first acquisition unit is used to acquire the data acquisition range under the first perspective, the data acquisition range under the second perspective, and the calibration parameters of the data acquisition device. The second acquisition unit acquires the point cloud overlap range based on the data acquisition range under the first viewpoint, the data acquisition range under the second viewpoint, and the calibration parameters of the data acquisition device.

7. A terminal device, comprising: A processor, a memory, and a computer program stored in the memory and executable on the processor, characterized in that the processor, when executing the computer program, implements the steps of the method as claimed in any one of claims 1 to 4.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the method as described in any one of claims 1 to 4.

Images