Point cloud data quality inspection method and device and electronic equipment

Abstract

The invention relates to the technical field of maps, and provides a point cloud data quality inspection method and device and electronic equipment. The method comprises the following steps: acquiring single-frame point cloud data and inertial navigation trajectory data of a moving object of the point cloud data, carrying out ground extraction on the single-frame point cloud data to obtain ground point cloud data and ground point cloud data, and carrying out point cloud layer detection on the ground point cloud data to obtain first candidate problem point cloud data; carrying out registration on the ground point cloud data to obtain second candidate problem point cloud data, and respectively carrying out loopback layering detection and track smoothness detection on the inertial navigation track data to obtain third candidate problem point cloud data and fourth candidate problem point cloud data; and finally, performing classification prediction on the first to fourth candidate problem data by using the trained decision tree classifier to obtain a quality inspection result, so that quality inspection of the point cloud data can be comprehensively and accurately realized, the quality inspection quality is improved, and accurate and reliable point cloud data is provided for subsequent operations such as construction of a high-precision map.

Description

Technical Field

[0001] The present application relates to the field of map technology, and in particular to a point cloud data quality inspection method, device and electronic equipment. Background Art

[0002] At present, the implementation of L4 level autonomous driving is inseparable from high-precision maps, which provide long-range prior information for autonomous driving vehicles. With the gradual advancement of the commercialization of unmanned vehicles, the collection range of high-precision maps is also gradually increasing. However, due to the influence of the environment, the inertial navigation system has no way to provide accurate position and attitude information under viaducts and lush street trees. Although there is post-processing optimization of the position and attitude, it cannot guarantee absolute accuracy. Therefore, point cloud quality inspection becomes a very necessary step and the last checkpoint to ensure the quality of high-precision point cloud delivery.

[0003] In related technologies, when quality inspection is performed on point cloud data, usually only a simple point cloud integrity check is performed, or only one of stratification and misalignment is detected, or only the standard deviation of the posture information provided by the INS (Inertial Navagation System) is screened. However, sections with a smaller standard deviation will still have stratification problems, and it is impossible to accurately detect all problematic point cloud data. Summary of the invention

[0004] In view of this, the embodiments of the present application provide a point cloud data quality inspection method, device and electronic device to solve the problem of poor point cloud data quality inspection effect in the prior art.

[0005] A first aspect of an embodiment of the present application provides a point cloud data quality inspection method, comprising:

[0006] Acquire single-frame point cloud data and inertial navigation trajectory data of the moving object that collects the point cloud data;

[0007] Perform ground extraction on single-frame point cloud data to obtain ground point cloud data and ground point cloud data;

[0008] Performing point cloud layer detection on the ground point cloud data to obtain the first candidate problem point cloud data, and registering the ground point cloud data with the point cloud map to obtain the second candidate problem point cloud data;

[0009] Perform loop closure layer detection and trajectory smoothness detection on the inertial navigation trajectory data to obtain the third candidate problem point cloud data and the fourth candidate problem point cloud data;

[0010] The first to fourth candidate problem point cloud data are input into the trained decision tree classifier to obtain the point cloud data quality inspection result.

[0011] A second aspect of an embodiment of the present application provides a point cloud data quality inspection device, comprising:

[0012] An acquisition module is configured to acquire single-frame point cloud data and inertial navigation trajectory data of a moving object that acquires the point cloud data;

[0013] An extraction module is configured to perform ground extraction on the single-frame point cloud data to obtain ground point cloud data and ground point cloud data;

[0014] The detection module is configured to perform point cloud hierarchical detection on the ground point cloud data to obtain first candidate problem point cloud data, and align the ground point cloud data with the point cloud map to obtain second candidate problem point cloud data;

[0015] The detection module is further configured to perform loop closure layer detection and trajectory smoothness detection on the inertial navigation trajectory data to obtain third candidate problem point cloud data and fourth candidate problem point cloud data;

[0016] The quality inspection module is configured to input the first to fourth candidate problem point cloud data into the trained decision tree classifier to obtain the point cloud data quality inspection result.

[0017] According to a third aspect of an embodiment of the present application, an electronic device is provided, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the above method when executing the computer program.

[0018] Compared with the prior art, the embodiments of the present application have the following beneficial effects: the embodiments of the present application obtain single-frame point cloud data and inertial navigation trajectory data of the moving object that collects the point cloud data, perform ground extraction on the single-frame point cloud data to obtain ground point cloud data and ground point cloud data, perform point cloud hierarchy detection on the ground point cloud data to obtain first candidate problem point cloud data, perform registration on the ground point cloud data to obtain second candidate problem point cloud data, perform loop closure hierarchy detection and trajectory smoothness detection on the inertial navigation trajectory data, respectively, to obtain third candidate problem point cloud data and fourth candidate problem point cloud data, and finally use the trained decision tree classifier to classify and predict the first to fourth candidate problem data to obtain quality inspection results, which can comprehensively and accurately realize quality inspection of point cloud data, improve the quality of quality inspection, and provide accurate and reliable point cloud data for subsequent operations such as building high-precision maps. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0020] Figure 1 It is a flow chart of a point cloud data quality inspection method provided in an embodiment of the present application.

[0021] Figure 2 It is a flowchart of a method for performing point cloud hierarchy detection on ground point cloud data to obtain first candidate problem point cloud data, as provided in an embodiment of the present application.

[0022] Figure 3 It is a schematic diagram of a method for obtaining sub-point cloud data provided in an embodiment of the present application.

[0023] Figure 4 It is a flow chart of a method provided in an embodiment of the present application for aligning ground point cloud data with a point cloud map to obtain second candidate problem point cloud data.

[0024] Figure 5 It is a flowchart of a method for performing loop closure layered detection on inertial navigation trajectory data to obtain third candidate problem point cloud data provided in an embodiment of the present application.

[0025] Figure 6 It is a schematic diagram of a method for performing loop closure layered detection on inertial navigation trajectory data provided in an embodiment of the present application.

[0026] Figure 7 It is a flow chart of a method for performing trajectory smoothness detection on inertial navigation trajectory data to obtain fourth candidate problem point cloud data provided in an embodiment of the present application.

[0027] Figure 8 It is a schematic diagram of the structure of the decision tree classifier provided in the embodiment of the present application.

[0028] Fig. 9 It is a flow chart of another point cloud data quality inspection method provided in an embodiment of the present application.

[0029] Fig.10 It is a schematic diagram of the result after quality inspection of point cloud data using the point cloud data quality inspection provided in an embodiment of the present application.

[0030] Fig.11 It is a schematic diagram of a point cloud data quality inspection device provided in an embodiment of the present application.

[0031] Fig.12 It is a schematic diagram of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0032] In the following description, specific details such as specific system structures, technologies, etc. are provided for the purpose of illustration rather than limitation, so as to provide a thorough understanding of the embodiments of the present application. However, it should be clear to those skilled in the art that the present application may also be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to prevent unnecessary details from obstructing the description of the present application.

[0033] A point cloud data quality inspection method and device according to an embodiment of the present application will be described in detail below with reference to the accompanying drawings.

[0034] As mentioned above, point cloud quality inspection is a very necessary step and the last checkpoint to ensure the quality of high-precision point cloud delivery. The most important quality inspections in point cloud quality inspection include stratification and misalignment detection. Stratification refers to the phenomenon that the original point cloud ground is highly inconsistent due to multiple acquisitions, and misalignment is the phenomenon that the point cloud road surface is horizontally offset due to posture jumps during the acquisition process.

[0035] In related technologies, when quality inspection is performed on point cloud data, usually only a simple point cloud integrity check is performed, or only one of stratification and misalignment is detected, or only the standard deviation of the posture information provided by INS is screened. However, sections with a smaller standard deviation will still have stratification problems, and it is impossible to accurately detect all problematic point cloud data.

[0036] In view of this, an embodiment of the present application provides a point cloud data quality inspection method, which obtains single-frame point cloud data and inertial navigation trajectory data of a moving object that collects the point cloud data, performs ground extraction on the single-frame point cloud data to obtain ground point cloud data and ground point cloud data, performs point cloud hierarchy detection on the ground point cloud data to obtain first candidate problem point cloud data, performs registration on the ground point cloud data to obtain second candidate problem point cloud data, performs loop closure hierarchy detection and trajectory smoothness detection on the inertial navigation trajectory data, respectively, to obtain third candidate problem point cloud data and fourth candidate problem point cloud data, and finally uses the trained decision tree classifier to classify and predict the first to fourth candidate problem data to obtain quality inspection results, which can comprehensively and accurately realize point cloud data quality inspection, improve the quality of quality inspection, and provide accurate and reliable point cloud data for subsequent operations such as building high-precision maps.

[0037] Figure 1 Schematic diagram of a point cloud data quality inspection method provided in an embodiment of the present application. Figure 1 As shown, the method comprises the following steps:

[0038] In step S101, single-frame point cloud data and inertial navigation trajectory data of a moving object for collecting point cloud data are acquired.

[0039] In step S102, ground extraction is performed on the single-frame point cloud data to obtain ground point cloud data and above-ground point cloud data.

[0040] In step S103, point cloud layer detection is performed on the ground point cloud data to obtain first candidate problem point cloud data, and the ground point cloud data is aligned with the point cloud map to obtain second candidate problem point cloud data.

[0041] In step S104, loop closure layer detection and trajectory smoothness detection are performed on the inertial navigation trajectory data to obtain third candidate problem point cloud data and fourth candidate problem point cloud data.

[0042] In step S105, the first to fourth candidate problem point cloud data are input into the trained decision tree classifier to obtain the point cloud data quality inspection result.

[0043] In some embodiments of the present application, the method may be executed by a server, or by a terminal with certain processing capabilities.

[0044] In certain embodiments of the present application, a single frame of point cloud data can be acquired, and INS trajectory data of a mobile object that acquires the point cloud data can be acquired. The mobile object that acquires the point cloud data can be, for example, a vehicle equipped with a point cloud acquisition device such as a laser radar, or other movable devices. The mobile object can continuously acquire and output point cloud data, and then perform quality inspection on the acquired point cloud data frame by frame.

[0045] In some implementations, ground extraction may be first performed on single-frame point cloud data to obtain ground point cloud data and ground point cloud data. Ground extraction may be implemented using a threshold-based method, a shape-based method, a statistical method, or other methods, which are not limited here.

[0046] In some embodiments, the extracted ground point cloud data may be subjected to point cloud layer detection to obtain the first candidate problem point cloud data. At the same time, a point cloud map generated using the collected point cloud data may be obtained, and the ground point cloud data extracted from the single frame point cloud data may be registered with the point cloud map to obtain the second candidate problem point cloud data.

[0047] Furthermore, loop closure layer detection and trajectory smoothness detection can be performed on the inertial navigation trajectory data to obtain the third candidate problem point cloud data and the fourth candidate problem point cloud data. In one example, the problem position in the inertial navigation trajectory data can be determined by loop closure layer detection and trajectory smoothness detection, and then the point cloud data at the position corresponding to the problem position in the single frame point cloud data is determined as the candidate problem point cloud data.

[0048] In another embodiment, the first to fourth candidate problem point cloud data may be input into a trained decision tree classifier to obtain a point cloud data quality inspection result. The decision tree classifier may be a decision tree classifier trained based on existing point cloud data quality inspection results. The decision tree classifier receives the candidate problem point cloud data as input, classifies each candidate problem point cloud data, and finally determines that the candidate problem point cloud data classified as failing the quality inspection is the problem point cloud data, and obtains the quality inspection result of the point cloud data.

[0049] According to the technical solution provided in the embodiment of the present application, by acquiring single-frame point cloud data and inertial navigation trajectory data of the moving object that collects the point cloud data, ground extraction is performed on the single-frame point cloud data to obtain ground point cloud data and ground point cloud data, point cloud hierarchy detection is performed on the ground point cloud data to obtain first candidate problem point cloud data, the ground point cloud data is aligned to obtain second candidate problem point cloud data, loop closure hierarchy detection and trajectory smoothness detection are performed on the inertial navigation trajectory data, respectively, to obtain third candidate problem point cloud data and fourth candidate problem point cloud data, and finally, the trained decision tree classifier is used to classify and predict the first to fourth candidate problem data to obtain quality inspection results, which can comprehensively and accurately realize quality inspection of point cloud data, improve the quality of quality inspection, and provide accurate and reliable point cloud data for subsequent operations such as building high-precision maps.

[0050] Figure 2 1 is a flow chart of a method for performing point cloud layer detection on ground point cloud data to obtain first candidate problem point cloud data provided by an embodiment of the present application. Figure 2 As shown, the method comprises the following steps:

[0051] In step S201, the ground point cloud data is spatially segmented based on the inertial navigation trajectory to obtain a first sub-point cloud data set.

[0052] The first sub-point cloud data set includes a plurality of sub-point cloud data, each sub-point cloud data corresponds to a subspace, and each subspace has the same volume and does not overlap.

[0053] In step S202, each sub-point cloud data in the first sub-point cloud data set is clustered to obtain N categories.

[0054] In step S203, plane fitting is performed for each category to obtain N fitting planes.

[0055] In step S204, in the first sub-point cloud data set, sub-point cloud data in the fitting plane whose projection of the plane normal vector relative to the reference plane in the vertical direction is smaller than a first preset threshold is removed to obtain a second sub-point cloud data set.

[0056] In step S205, the sub-point cloud data in the category whose height difference between the cluster center and the cluster centers of at least M sub-point cloud data is greater than a second preset threshold in the second sub-point cloud data set is determined as the first candidate problem point cloud data.

[0057] Among them, the reference plane is the ground, and M is a positive integer.

[0058] In some embodiments of the present application, performing point cloud hierarchical detection on ground point cloud data to obtain the first candidate problem point cloud data may be to first perform spatial segmentation on the ground point cloud data based on the INS trajectory to obtain a first sub-point cloud data set. The first sub-point cloud data set includes a plurality of sub-point cloud data, each of which corresponds to a subspace, and each subspace has the same volume and does not overlap.

[0059] Figure 3 Schematic diagram of a method for obtaining sub-point cloud data provided by an embodiment of the present application. Figure 3 As shown, based on the INS trajectory, a cuboid with preset length and width values ​​can be taken at every preset distance in the ground point cloud data, and the point cloud data in the cuboid is a sub-point cloud data. The preset distance is, for example, 2 meters, and the preset length and width values ​​can be, for example, both 1 meter.

[0060] After obtaining a first sub-point cloud data set including a plurality of sub-point cloud data, each sub-point cloud data in the first sub-point cloud data set can be clustered to obtain N categories. Further, plane fitting is performed on each of the N categories to obtain fitting planes of each category. For each fitting plane, its plane normal vector relative to the reference plane is determined. The projection of each plane normal vector in the vertical direction, i.e., the z-axis direction of the world coordinate system, is obtained, and the sub-point cloud data corresponding to the fitting plane whose projection value is less than a first preset threshold, i.e., the non-vertical fitting plane, is removed from the first sub-point cloud data set to obtain a second sub-point cloud data set.

[0061] Further, the sub-point cloud data in the second sub-point cloud data set whose height difference between the cluster center and the cluster centers of at least M sub-point cloud data is greater than the second preset threshold is determined as the first candidate problem point cloud data. In other words, the height value of the cluster center of each category in the second sub-point cloud data set, that is, the z value in the world coordinate system, can be obtained, and then the difference between the height values ​​of each category can be calculated. If the difference between the height values ​​of a category and other multiple categories is greater than the second preset threshold, the sub-point cloud data in the category can be determined as the first candidate problem point cloud data.

[0062] The values ​​of the first preset threshold and the second preset threshold can be set according to actual needs and are not limited here. In one example, the first preset threshold can be, for example, 0.97, and the second preset threshold can be, for example, 20 centimeters.

[0063] Figure 4 1 is a flow chart of a method for aligning ground point cloud data with a point cloud map to obtain second candidate problem point cloud data provided by an embodiment of the present application. Figure 4 As shown, the method comprises the following steps:

[0064] In step S401, a point cloud map is obtained, and the ground point cloud data is aligned with the point cloud map.

[0065] In step S402, the ground point cloud data having a registration score greater than a third preset threshold is determined as the second candidate problem point cloud data.

[0066] In some embodiments of the present application, when registering the ground point cloud data with the point cloud map to obtain the second candidate problem point cloud data, the point cloud map can be first obtained, and then the ground point cloud data and the point cloud map are registered to obtain a registration score. Further, the ground point cloud data obtained by the registration that is greater than the third preset threshold is determined to be the second candidate problem point cloud data.

[0067] The value of the third preset threshold can be set according to actual needs and is not limited here. In one example, the third preset threshold can be, for example, 20 centimeters.

[0068] Figure 5 1 is a flow chart of a method for performing loop closure layered detection on inertial navigation trajectory data to obtain third candidate problem point cloud data provided by an embodiment of the present application. Figure 5 As shown, the method comprises the following steps:

[0069] In step S501, adjacent inertial navigation trajectory data of the inertial navigation trajectory data is obtained.

[0070] The distance between adjacent inertial navigation trajectory data is less than a preset distance threshold, and the time interval is greater than a preset time interval.

[0071] In step S502, the minimum difference between the vertical coordinate values ​​of the inertial navigation trajectory data and the adjacent inertial navigation trajectory data is determined.

[0072] In step S503, a first target position point of the inertial navigation trajectory data whose minimum difference is greater than a fourth preset threshold is determined.

[0073] In step S504, the point cloud data corresponding to the first target position point in the single frame point cloud data is determined as the third candidate problem point cloud data.

[0074] In some embodiments of the present application, when performing loop closure layered detection on INS trajectory data, adjacent INS trajectory data of the INS trajectory data may be first obtained. The distance between the adjacent INS trajectory data and the INS trajectory data is less than a preset distance threshold, and the time interval is greater than a preset time interval. For example, adjacent INS trajectory data within a preset distance threshold range of the driving trajectory of the mobile object collecting point cloud data and whose collection time is greater than a preset time interval may be obtained, that is, adjacent INS trajectory data within a certain range collected by the mobile object at different times may be collected.

[0075] The values ​​of the preset distance threshold and the preset time interval can be set according to actual needs and are not limited here. In one example, the preset distance threshold can be, for example, 40 meters, and the preset time interval can be, for example, 10 seconds.

[0076] Further, the difference between the numerical coordinates of the INS trajectory data and the adjacent INS trajectory data can be determined, and the first target position point of the INS trajectory data whose difference is greater than the fourth preset threshold is determined. Finally, the point cloud data corresponding to the first target position point in the single frame point cloud data is determined as the third candidate problem point cloud data.

[0077] In some implementations, the first candidate problem point cloud data may be updated according to the loop closure layered detection result to remove false positive data in the first candidate problem point cloud data. That is, if a first candidate problem point cloud data is displayed as normal data in the loop closure layered detection result, the point cloud data may be removed from the first candidate problem data, and then the updated first candidate problem data and the second to fourth candidate problem point cloud data may be input into the trained decision tree classifier to obtain the point cloud data quality inspection result.

[0078] The value of the fourth preset threshold value can be set according to actual needs and is not limited here. In one example, the fourth preset threshold value can be, for example, 20 centimeters.

[0079] Figure 6 Schematic diagram of a method for performing loop closure layered detection on inertial navigation trajectory data provided by an embodiment of the present application. Figure 6 As shown, the adjacent INS trajectory data 1 and 2 of the INS trajectory data 3 can be obtained, and their minimum z value differences can be calculated respectively. If the minimum z value difference deltaz=2.68378 decimeters between the INS trajectory data 3 and the adjacent INS trajectory data 2 is greater than a fourth preset threshold value, such as 20 centimeters, the first target position point in the INS trajectory data can be determined, and then the point cloud data corresponding to the first target position point in the single frame point cloud data is determined as the third candidate problem point cloud data.

[0080] Furthermore, since INS trajectory data can distinguish elevated roads, if Figure 6If the INS trajectory data in is the INS trajectory data of the elevated road, it can be determined that it is not layered. If the data of some positions in the INS trajectory data are identified as the first candidate problem point cloud data during point cloud layer detection, the point cloud data corresponding to these positions can be considered as false positive data, and then removed from the first candidate problem point cloud data.

[0081] Figure 7 1 is a flow chart of a method for performing trajectory smoothness detection on inertial navigation trajectory data to obtain fourth candidate problem point cloud data provided by an embodiment of the present application. Figure 7 As shown, the method comprises the following steps:

[0082] In step S701 , the curvature of each data point in the inertial navigation trajectory data is determined.

[0083] In step S702, a second target position point of a data point whose curvature is greater than a fifth preset threshold is determined.

[0084] In step S703, the point cloud data corresponding to the second target position point in the single frame point cloud data is determined as the fourth candidate problem point cloud data.

[0085] In some embodiments of the present application, when performing trajectory smoothness detection on INS trajectory data, the curvature of each data point in the INS trajectory data can be first determined. In one example, the curvature of each data point can be obtained by calculating the curvature of three points, or the curvature can be determined by other methods, which are not limited here. Further, a second target position point of the data point whose curvature is greater than the fifth preset threshold is determined. This position represents an uneven point of the INS. Since uneven points are prone to stratification or misalignment, the point cloud data corresponding to the second target position point in the single-frame point cloud data can be determined as the fourth candidate problem point cloud data.

[0086] The value of the fifth preset threshold value can be set according to actual needs and is not limited here. In one example, the fifth preset threshold value can be 1, for example.

[0087] In some embodiments of the present application, the signal standard deviation of the INS trajectory data can also be considered as a screening condition for the problem point cloud data. That is to say, the signal standard deviation of each point data in the inertial navigation trajectory data can also be determined to obtain the third target position point of the data point whose signal standard deviation is greater than the sixth preset threshold. Next, the point cloud data corresponding to the third target position point in the single-frame point cloud data is determined as the fifth candidate problem point cloud data. Finally, the first to fifth candidate problem point cloud data are input into the trained decision tree classifier to obtain the point cloud data quality inspection result.

[0088] In some embodiments of the present application, the decision tree classifier can determine the point cloud data quality inspection result in the following manner: perform point cloud hierarchical detection scoring and registration scoring on the input candidate problem data. Perform loop hierarchical detection scoring, trajectory smoothness detection scoring, and signal standard deviation scoring on the candidate problem data that have completed the point cloud hierarchical scoring and the candidate problem data that have completed the registration scoring, respectively, to obtain the scoring results. Determine the direct result based on the scoring results.

[0089] Figure 8 Schematic diagram of the structure of the decision tree classifier provided in the embodiment of the present application. Figure 8 As shown, the decision tree classifier includes a point cloud input module, in which candidate problem point cloud data can be input. The candidate problem point cloud data is input into the ground point cloud registration detection module and the ground point cloud hierarchical detection module for detection. The candidate point cloud detected by the ground point cloud registration detection module is then input into the loop closure hierarchical detection module, the trajectory smoothness detection module and the INS signal standard deviation detection module for detection. The candidate point cloud detected by the ground point cloud hierarchical detection module is also input into the loop closure hierarchical detection module, the estimated smoothness detection module and the INS signal standard deviation detection module for detection. Finally, considering all factors comprehensively, the point cloud quality inspection results are obtained, including point cloud quality inspection passed and point cloud quality inspection failed.

[0090] Fig. 9 FIG. 1 is a flow chart of another point cloud data quality inspection method provided in an embodiment of the present application. Fig. 9 As shown in the figure, first input the single-frame point cloud and INS trajectory data. The single-frame point cloud can be ground extracted to obtain ground point cloud data and ground point cloud data. Perform point cloud layer detection on the ground point cloud data, and input the detection results into the decision tree classifier. Perform registration processing on the ground point cloud data and the point cloud map, and input the processing results into the decision tree classifier. For the INS trajectory data, loop layer detection, trajectory smoothness detection and INS signal standard deviation detection are performed respectively, and the detection results are all input into the decision tree classifier. The decision tree classifier receives the results of each single test and outputs the coordinates of the problem point cloud after comprehensive evaluation.

[0091] Fig.10 1 is a schematic diagram of the result of quality inspection of point cloud data using the point cloud data quality inspection provided by the embodiment of the present application. Fig.10 As shown, the lines represent the INS trajectory and the circles represent the coordinates of the problem point cloud obtained by quality inspection.

[0092] By adopting the technical solution of the embodiment of the present application, all information is comprehensively considered, the point cloud is layered and dislocation checked, and the position coordinates of the point cloud that fails the quality inspection are output, thereby improving the quality inspection efficiency of the point cloud data.

[0093] All the above optional technical solutions can be arbitrarily combined to form optional embodiments of the present application, which will not be described one by one here.

[0094] The following is an embodiment of the device of the present application, which can be used to execute the embodiment of the method of the present application. For details not disclosed in the embodiment of the device of the present application, please refer to the embodiment of the method of the present application.

[0095] Fig.11 Schematic diagram of a point cloud data quality inspection device provided in an embodiment of the present application. Fig.11 As shown, the device comprises:

[0096] The acquisition module 1101 is configured to acquire single-frame point cloud data and inertial navigation trajectory data of a moving object that acquires the point cloud data.

[0097] The extraction module 1102 is configured to perform ground extraction on the single-frame point cloud data to obtain ground point cloud data and ground point cloud data.

[0098] The detection module 1103 is configured to perform point cloud layer detection on the ground point cloud data to obtain first candidate problem point cloud data, and align the ground point cloud data with the point cloud map to obtain second candidate problem point cloud data.

[0099] The detection module 1103 is further configured to perform loop closure layer detection and trajectory smoothness detection on the inertial navigation trajectory data to obtain third candidate problem point cloud data and fourth candidate problem point cloud data.

[0100] The quality inspection module 1104 is configured to input the first to fourth candidate problem point cloud data into the trained decision tree classifier to obtain a point cloud data quality inspection result.

[0101] According to the technical solution provided in the embodiment of the present application, by acquiring single-frame point cloud data and inertial navigation trajectory data of the moving object that collects the point cloud data, ground extraction is performed on the single-frame point cloud data to obtain ground point cloud data and ground point cloud data, point cloud hierarchy detection is performed on the ground point cloud data to obtain first candidate problem point cloud data, the ground point cloud data is aligned to obtain second candidate problem point cloud data, loop closure hierarchy detection and trajectory smoothness detection are performed on the inertial navigation trajectory data, respectively, to obtain third candidate problem point cloud data and fourth candidate problem point cloud data, and finally, the trained decision tree classifier is used to classify and predict the first to fourth candidate problem data to obtain quality inspection results, which can comprehensively and accurately realize quality inspection of point cloud data, improve the quality of quality inspection, and provide accurate and reliable point cloud data for subsequent operations such as building high-precision maps.

[0102] In some embodiments, point cloud hierarchy detection is performed on ground point cloud data to obtain first candidate problem point cloud data, including: spatially segmenting the ground point cloud data based on the inertial navigation trajectory to obtain a first sub-point cloud data set, the first sub-point cloud data set including multiple sub-point cloud data, each sub-point cloud data corresponds to a subspace, and each subspace has the same volume and does not overlap; clustering the sub-point cloud data in the first sub-point cloud data set to obtain N categories; performing plane fitting on each category to obtain N fitting planes; in the first sub-point cloud data set, removing the sub-point cloud data in the fitting plane whose projection of the plane normal vector relative to the reference plane in the vertical direction is less than a first preset threshold to obtain a second sub-point cloud data set; determining the sub-point cloud data in the category in which the height difference between the cluster center and the cluster centers of at least M sub-point cloud data is greater than a second preset threshold in the second sub-point cloud data set as the first candidate problem point cloud data, where M is a positive integer; wherein the reference plane is the ground.

[0103] In some embodiments, the ground point cloud data is aligned with the point cloud map to obtain second candidate problem point cloud data, including: obtaining the point cloud map, and aligning the ground point cloud data with the point cloud map; determining that the ground point cloud data having a alignment score greater than a third preset threshold is the second candidate problem point cloud data.

[0104] In some embodiments, loop closure hierarchical detection is performed on the inertial navigation trajectory data to obtain third candidate problem point cloud data, including: obtaining adjacent inertial navigation trajectory data of the inertial navigation trajectory data, the distance between the adjacent inertial navigation trajectory data and the inertial navigation trajectory data is less than a preset distance threshold, and the time interval is greater than a preset time interval; determining the minimum difference between the vertical coordinate values ​​of the inertial navigation trajectory data and the adjacent inertial navigation trajectory data; determining the first target position point of the inertial navigation trajectory data whose minimum difference is greater than a fourth preset threshold; and determining that the point cloud data corresponding to the first target position point in the single frame point cloud data is the third candidate problem point cloud data.

[0105] In some embodiments, a trajectory smoothness test is performed on the inertial navigation trajectory data to obtain fourth candidate problem point cloud data, including: determining the curvature of each data point in the inertial navigation trajectory data; determining a second target position point of the data point whose curvature is greater than a fifth preset threshold; and determining that the point cloud data corresponding to the second target position point in a single frame of point cloud data is the fourth candidate problem point cloud data.

[0106] In some implementations, it also includes: determining the signal standard deviation of each point data in the inertial navigation trajectory data; obtaining a third target position point whose signal standard deviation is greater than a sixth preset threshold value; determining that the point cloud data corresponding to the third target position point in the single frame point cloud data is the fifth candidate problem point cloud data; and inputting the first to fifth candidate problem point cloud data into the trained decision tree classifier to obtain a point cloud data quality inspection result.

[0107] In some implementations, it also includes: updating the first candidate problem point cloud data according to the loop closure layered detection result; inputting the second to fourth candidate problem point cloud data and the updated first candidate problem data into the trained decision tree classifier to obtain the point cloud data quality inspection result.

[0108] In some embodiments, the decision tree classifier determines the point cloud data quality inspection results in the following manner: performing point cloud hierarchical detection scoring and registration scoring on the input candidate problem data; performing loop hierarchical detection scoring, trajectory smoothness detection scoring, and signal standard deviation scoring on the candidate problem data that have completed the point cloud hierarchical scoring and the candidate problem data that have completed the registration scoring to obtain scoring results; and determining direct results based on the scoring results.

[0109] It should be understood that the size of the serial numbers of the steps in the above embodiments does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0110] Fig.12 Schematic diagram of an electronic device provided in an embodiment of the present application. Fig.12 As shown, the electronic device 12 of this embodiment includes: a processor 1201, a memory 1202, and a computer program 1203 stored in the memory 1202 and executable on the processor 1201. When the processor 1201 executes the computer program 1203, the steps in the above-mentioned method embodiments are implemented. Alternatively, when the processor 1201 executes the computer program 1203, the functions of the modules / units in the above-mentioned device embodiments are implemented.

[0111] The electronic device 12 may be a desktop computer, a notebook, a PDA, a cloud server, or other electronic device. The electronic device 12 may include, but is not limited to, a processor 1201 and a memory 1202. Those skilled in the art will appreciate that Fig.12 The electronic device 12 is merely an example and does not limit the electronic device 12 , and may include more or fewer components than those shown in the figure, or different components.

[0112] Processor 1201 can be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc.

[0113] The memory 1202 may be an internal storage unit of the electronic device 12, for example, a hard disk or memory of the electronic device 12. The memory 1202 may also be an external storage device of the electronic device 12, for example, a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card (FlashCard), etc. equipped on the electronic device 12. The memory 1202 may also include both an internal storage unit of the electronic device 12 and an external storage device. The memory 1202 is used to store computer programs and other programs and data required by the electronic device.

[0114] Those skilled in the art can clearly understand that for the convenience and simplicity of description, only the division of the above-mentioned functional units and modules is used as an example. In actual applications, the above-mentioned functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiments can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of software functional units.

[0115] If the integrated module / unit 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. Based on this understanding, the present application implements all or part of the processes in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium, and the computer program can implement the steps of the above-mentioned various method embodiments when executed by the processor. The computer program may include computer program code, which may be in source code form, object code form, executable file or some intermediate form. Computer-readable media may include: any entity or device capable of carrying computer program code, recording medium, U disk, mobile hard disk, disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electric carrier signal, telecommunication signal and software distribution medium, etc.

[0116] The above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application, and should all be included in the protection scope of the present application.

Claims

1. A point cloud data quality inspection method, characterized in that: include: Acquire single-frame point cloud data and inertial navigation trajectory data of a moving object that acquires the point cloud data; Performing ground extraction on the single-frame point cloud data to obtain ground point cloud data and ground point cloud data; Performing point cloud layer detection on the ground point cloud data to obtain first candidate problem point cloud data, and registering the ground point cloud data with a point cloud map to obtain second candidate problem point cloud data; Performing loop closure layer detection and trajectory smoothness detection on the inertial navigation trajectory data respectively to obtain third candidate problem point cloud data and fourth candidate problem point cloud data; The first to fourth candidate problem point cloud data are input into the trained decision tree classifier to obtain the point cloud data quality inspection result.

2. The method according to claim 1, characterized in that The step of performing point cloud hierarchical detection on the ground point cloud data to obtain first candidate problem point cloud data includes: Based on the inertial navigation trajectory, the ground point cloud data is spatially segmented to obtain a first sub-point cloud data set, wherein the first sub-point cloud data set includes a plurality of sub-point cloud data, each sub-point cloud data corresponds to a subspace, and each subspace has the same volume and does not overlap; Clustering each sub-point cloud data in the first sub-point cloud data set to obtain N categories; Perform plane fitting for each category to obtain N fitting planes; In the first sub-point cloud data set, sub-point cloud data in the fitting plane whose projection of the plane normal vector relative to the reference plane in the vertical direction is less than a first preset threshold is removed to obtain a second sub-point cloud data set; Determine, in the second sub-point cloud data set, sub-point cloud data in a category whose height difference between a cluster center and cluster centers of at least M sub-point cloud data is greater than a second preset threshold as first candidate problem point cloud data, where M is a positive integer; The reference plane is the ground.

3. The method according to claim 1, characterized in that The step of registering the ground point cloud data with the point cloud map to obtain second candidate problem point cloud data includes: Acquire a point cloud map, and align the ground point cloud data with the point cloud map; The ground point cloud data having a registration score greater than a third preset threshold is determined as the second candidate problem point cloud data.

4. The method according to claim 1, characterized in that Perform loop closure and hierarchical detection on the inertial navigation trajectory data to obtain the third candidate problem point cloud data, including: Acquire adjacent inertial navigation trajectory data of the inertial navigation trajectory data, where the distance between the adjacent inertial navigation trajectory data and the inertial navigation trajectory data is less than a preset distance threshold, and the time interval is greater than a preset time interval; Determine a minimum difference between the vertical coordinate values ​​of the inertial navigation trajectory data and adjacent inertial navigation trajectory data; Determine a first target position point of the inertial navigation trajectory data whose minimum difference is greater than a fourth preset threshold; Determine that the point cloud data corresponding to the first target position point in the single frame point cloud data is the third candidate problem point cloud data.

5. The method according to claim 1, characterized in that The fourth candidate problem point cloud data is obtained by performing trajectory smoothness detection on the inertial navigation trajectory data, including: Determining the curvature of each data point in the inertial navigation trajectory data; Determine a second target position point of a data point whose curvature is greater than a fifth preset threshold; Determine that the point cloud data corresponding to the second target position point in the single frame point cloud data is the fourth candidate problem point cloud data.

6. The method according to claim 1, characterized in that The method further comprises: Determining the signal standard deviation of each point data in the inertial navigation trajectory data; Acquire a third target position point of a data point whose signal standard deviation is greater than a sixth preset threshold; Determine that the point cloud data corresponding to the third target position point in the single frame point cloud data is the fifth candidate problem point cloud data; The first to fifth candidate problem point cloud data are input into the trained decision tree classifier to obtain the point cloud data quality inspection results.

7. The method according to claim 1, characterized in that The method further comprises: Update the first candidate problem point cloud data according to the loop closure layered detection result; The second to fourth candidate problem point cloud data and the updated first candidate problem data are input into the trained decision tree classifier to obtain the point cloud data quality inspection result.

8. The method according to claim 6, characterized in that The decision tree classifier determines the point cloud data quality inspection result in the following way: Perform point cloud hierarchical detection and registration scoring on the input candidate problem data; The candidate problem data that have completed the point cloud hierarchical scoring and the candidate problem data that have completed the registration scoring are respectively subjected to loop closure hierarchical detection scoring, trajectory smoothness detection scoring and signal standard deviation scoring to obtain the scoring results; An immediate result is determined based on the scoring result.

9. A point cloud data quality inspection device, characterized in that: include: An acquisition module is configured to acquire single-frame point cloud data and inertial navigation trajectory data of a moving object that acquires the point cloud data; An extraction module is configured to perform ground extraction on the single frame point cloud data to obtain ground point cloud data and ground point cloud data; A detection module is configured to perform point cloud layer detection on the ground point cloud data to obtain first candidate problem point cloud data, and to register the ground point cloud data with a point cloud map to obtain second candidate problem point cloud data; The detection module is further configured to perform loop closure layer detection and trajectory smoothness detection on the inertial navigation trajectory data to obtain third candidate problem point cloud data and fourth candidate problem point cloud data; The quality inspection module is configured to input the first to fourth candidate problem point cloud data into the trained decision tree classifier to obtain the point cloud data quality inspection result.

10. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 8 are implemented.

Images