WO2021212477A1 - 校正点云数据的方法和相关装置 - Google Patents

校正点云数据的方法和相关装置 Download PDF

Info

Publication number
WO2021212477A1
WO2021212477A1 PCT/CN2020/086728 CN2020086728W WO2021212477A1 WO 2021212477 A1 WO2021212477 A1 WO 2021212477A1 CN 2020086728 W CN2020086728 W CN 2020086728W WO 2021212477 A1 WO2021212477 A1 WO 2021212477A1
Authority
WO
WIPO (PCT)
Prior art keywords
point cloud
cloud data
features
feature
adjustment parameter
Prior art date
Application number
PCT/CN2020/086728
Other languages
English (en)
French (fr)
Inventor
石峰
刘建琴
乔得志
Original Assignee
华为技术有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 华为技术有限公司 filed Critical 华为技术有限公司
Priority to CN202080004997.1A priority Critical patent/CN112739983B/zh
Priority to PCT/CN2020/086728 priority patent/WO2021212477A1/zh
Publication of WO2021212477A1 publication Critical patent/WO2021212477A1/zh

Links

Images

Classifications

    • GPHYSICS
    • G01MEASURING; TESTING
    • G01CMEASURING DISTANCES, LEVELS OR BEARINGS; SURVEYING; NAVIGATION; GYROSCOPIC INSTRUMENTS; PHOTOGRAMMETRY OR VIDEOGRAMMETRY
    • G01C21/00Navigation; Navigational instruments not provided for in groups G01C1/00 - G01C19/00
    • G01C21/26Navigation; Navigational instruments not provided for in groups G01C1/00 - G01C19/00 specially adapted for navigation in a road network
    • G01C21/28Navigation; Navigational instruments not provided for in groups G01C1/00 - G01C19/00 specially adapted for navigation in a road network with correlation of data from several navigational instruments
    • G01C21/30Map- or contour-matching
    • G01C21/32Structuring or formatting of map data

Definitions

  • This application relates to the technical field of autonomous driving and intelligent networked vehicles, and more specifically, to a method and related devices for correcting point cloud data.
  • the point cloud is a collection of point data on the surface of an object obtained by measuring instruments (such as cameras, lidars, etc.).
  • the point cloud data may include three-dimensional coordinate information of the object.
  • Point cloud data can be used for target detection and recognition.
  • point cloud data can be used to identify cars, road traffic markings, road traffic signs and other information in the scene. Therefore, point cloud data can be used in areas such as autonomous driving and intelligent robot navigation.
  • the automatic driving system needs to accurately predict the road conditions ahead. For the range that the physical sensor cannot perceive, it is also necessary to provide corresponding information. Therefore, the automatic driving system needs to obtain a high-precision map in advance as the prior knowledge provided by the automatic driving.
  • the accuracy of the map is determined by the Global Navigation Satellite System (GNSS), Real-Time Kinematic (RTK) or Inertial Navigation System (INS) Sure.
  • GNSS Global Navigation Satellite System
  • RTK Real-Time Kinematic
  • INS Inertial Navigation System
  • the present application provides a method and a related device for correcting point cloud data, which can correct the point cloud data, thereby improving the accuracy of a high progress map.
  • an embodiment of the present application provides a method for correcting point cloud data, including: determining the N first reference features of a reference map in a first area and a target map in a first point cloud data set of the first area N first features of, the first point cloud data set is a collection of point cloud data of the first region, the N first reference features and the N first features have a one-to-one correspondence, and N is a positive integer;
  • the poses of the N first reference features in the reference map and the poses of the N first features in the first point cloud data set determine a first adjustment parameter set; according to the first adjustment parameter set, adjust Location information of each point cloud data in the first point cloud data set.
  • the above technical solution can correct the point cloud data without increasing the cost of collecting the point cloud data. Further, the above technical solution does not need to construct a loop in the process of correcting the point cloud data, so that the collection efficiency of the point cloud data can be improved. In addition, the accuracy of the point cloud data corrected by the above technical solution can be controlled.
  • the reference map is a digital orthophoto map, or the reference map is high-confidence point cloud data; or, the reference map is a construction design picture.
  • the above technical solution uses a high-precision map as a reference map for correcting point cloud data. In this way, the accuracy of the corrected point cloud data can be guaranteed.
  • the method further includes: adjusting the position information of each point cloud data in the second point cloud data set according to the first adjustment parameter set, and the second point cloud data set
  • the point cloud data set is a collection of point cloud data of the target map in the second area.
  • the method before adjusting the position information of each point cloud data in the second point cloud data set according to the first adjustment parameter set, the method further includes: It is determined that there is no corresponding feature between the reference map and the second point cloud data set.
  • the above technical solution can correct the point cloud data that can not be corrected by the reference map by using the point cloud data that can be corrected by the reference map. Therefore, the above technical solution can also correct the point cloud data without increasing the cost of the point cloud collection technology. Further, the above technical solution does not need to construct a loop in the process of correcting the point cloud data, so that the collection efficiency of the point cloud data can be improved.
  • the first area and the second area belong to a first road
  • the method further includes: determining a second adjustment parameter according to the first adjustment parameter set Set; according to the second adjustment parameter set, adjust the position information of each point cloud data in the third point cloud data set, the third point cloud data set is the set of point cloud data of the target map in the third area, the first The third area and the second area belong to a second road, and the second area is located in an intersection area of the first road and the second road.
  • the first adjustment parameter set and the second adjustment parameter set may be the same or different.
  • the above technical solution can correct the point cloud data that can not be corrected by the reference map by using the point cloud data that can be corrected by the reference map. Therefore, the above technical solution can also correct the point cloud data without increasing the cost of the point cloud collection technology. Further, the above technical solution does not need to construct a loop in the process of correcting the point cloud data, so that the collection efficiency of the point cloud data can be improved.
  • the poses of the N first reference features in the reference map and the N first features in the first point cloud data set includes: determining a second feature, the second feature being the first feature closest to the point cloud data collection device among the N first features; according to the second reference feature in the Determine the first adjustment parameter set by referring to the pose in the map and the pose of the second feature in the first point cloud data set, and the second reference feature is the N first reference features corresponding to the second feature The first reference feature.
  • the poses of the N first reference features in the reference map and the N first features in the first point cloud data set To determine the first adjustment parameter set, including: according to the poses of the N first reference features in the reference map and the poses of the N first features in the first point cloud data set, determining The K first features in the reference map whose errors between the corresponding first reference features are greater than the error threshold, where K is a positive integer less than or equal to N; according to the K first features in the first point cloud data set The pose, and the poses of the K first reference features corresponding to the K first features one-to-one in the reference map, determine the first adjustment parameter set.
  • an embodiment of the present application provides an apparatus for correcting point cloud data.
  • the apparatus for correcting point cloud data includes a unit for implementing the first aspect or any possible implementation manner of the first aspect.
  • an embodiment of the present application provides an apparatus for correcting point cloud data, and the apparatus for correcting point cloud data includes a processor.
  • the processor is configured to be coupled with the memory, read and execute computer program instructions in the memory, so as to implement the method in any one of the possible implementation manners in the method design of the first aspect described above.
  • an embodiment of the present application provides a computer-readable medium, the computer-readable medium includes computer instructions, and when the computer instructions are executed by a processor, the device for correcting point cloud data executes the above-mentioned first aspect Any one of the possible implementation methods in the method design.
  • embodiments of the present application provide a computer program product, which when the computer program product runs on a processor, causes the device for correcting point cloud data to execute any possible implementation of the method design of the first aspect above The method in the way.
  • FIG. 1 is a schematic diagram of an application scenario in which an embodiment of the present application is applied to the vehicle side and an action position of the embodiment of the present application in a map system architecture.
  • Fig. 2 is a schematic diagram of the deviation of the position information of the point cloud data.
  • Fig. 3 is a schematic diagram of a reference map used in an embodiment of the present application.
  • Fig. 4 is a schematic diagram of another reference map used in an embodiment of the present application.
  • Fig. 5 is a schematic diagram of dividing the ground-level road in Fig. 4 into three parts.
  • Fig. 6 is a flowchart of a method for correcting point cloud data provided by an embodiment of the present application.
  • FIG. 7 is a schematic diagram of N first reference features and N first features in area 1 shown in FIG. 4.
  • FIG. 8 is another schematic diagram of the N first reference features and the N first features in the area 1 shown in FIG. 4.
  • FIG. 9 is a flowchart of another method for correcting point cloud data provided by an embodiment of the present application.
  • FIG. 10 is a flowchart of another method for correcting point cloud data provided by an embodiment of the present application.
  • FIG. 11 is a flowchart of another method for correcting point cloud data provided by an embodiment of the present application.
  • FIG. 12 is a flowchart of another method for correcting point cloud data provided by an embodiment of the present application.
  • Fig. 13 is a schematic structural block diagram of an apparatus for correcting point cloud data according to an embodiment of the present application.
  • Fig. 14 is a structural block diagram of an apparatus for correcting point cloud data provided according to an embodiment of the present application.
  • the subscript sometimes as W 1 may form a clerical error at non-target as W1, while not emphasize the difference, to express their meaning is the same.
  • references described in this specification to "one embodiment” or “some embodiments”, etc. mean that one or more embodiments of the present application include a specific feature, structure, or characteristic described in combination with the embodiment. Therefore, the sentences “in one embodiment”, “in some embodiments”, “in some other embodiments”, “in some other embodiments”, etc. appearing in different places in this specification are not necessarily All refer to the same embodiment, but mean “one or more but not all embodiments” unless it is specifically emphasized otherwise.
  • the terms “including”, “including”, “having” and their variations all mean “including but not limited to”, unless otherwise specifically emphasized.
  • At least one refers to one or more, and “multiple” refers to two or more.
  • “And/or” describes the association relationship of the associated objects, indicating that there can be three relationships, for example, A and/or B, which can mean: A alone exists, A and B exist at the same time, and B exists alone, where A, B can be singular or plural.
  • the character “/” generally indicates that the associated objects before and after are in an “or” relationship.
  • the following at least one item (a)” or similar expressions refers to any combination of these items, including any combination of a single item (a) or a plurality of items (a).
  • at least one item (a) of a, b, or c can mean: a, b, c, ab, ac, bc, or abc, where a, b, and c can be single or multiple .
  • the high-precision map referred to in the embodiments of the present application refers to a map that can at least provide lane-level navigation. Under normal circumstances, a map with an error of less than 30 cm can provide lane-level navigation for the autonomous driving system. For example, a map with an error of less than 25 cm, 15 cm, or 10 cm.
  • FIG. 1 is a schematic diagram of an application scenario in which an embodiment of the present application is applied to the vehicle side.
  • a measurement system 120 and a computing device 130 may be installed in the vehicle 110.
  • the measurement system 120 may include a sensor for detecting and scanning a point cloud in a target scene.
  • the aforementioned sensors may include (light detection and ranging, LiDAR), three-dimensional scanners, depth cameras, etc., which are not limited in this application.
  • the collection of point cloud data collected by the sensor can be referred to as the original point cloud data set.
  • the measurement system 120 may also include a GNSS/RTK module.
  • the GNSS/RTK module is used to obtain location data of the vehicle 110.
  • the measurement system 120 may also include an inertial measurement unit (IMU).
  • IMU inertial measurement unit
  • the computing device 130 is connected to the measurement system 120, and is used to obtain the original point cloud data set, the position data of the vehicle 110, and the attitude information of the vehicle 110 from the measurement system 120, and to fuse the original point cloud data set, position data, and attitude information, Obtain the fused point cloud data set.
  • the point cloud data set referred to in the embodiments of the present application refers to a point cloud data set after fusion. In the process of fusing the original point cloud data set, you can delete outliers and/or correct the point cloud data according to existing methods (such as simultaneous localization and mapping (simultaneous localization and mapping, SLAM) loop repair). .
  • (B) in FIG. 1 is a schematic diagram of the position of action in the map system architecture of the embodiment of the present application.
  • the GNSS/RTK module cannot obtain accurate location information of the vehicle 110.
  • the attitude information collected by the IMU module may have cumulative errors. Therefore, the point cloud data will have some deviations from the actual information.
  • the technical solution of the present application can be used to correct the point cloud data, so as to reduce the error of the point cloud data and improve the accuracy of the map.
  • FIG. 2 is a schematic diagram of the deviation of the position information of the point cloud data.
  • the difference includes position difference and posture difference.
  • the position difference is reflected in the position deviation of the center point of the lane line.
  • the difference in attitude is reflected in the angular deviation of the lane line pointing direction.
  • Fig. 3 is a schematic diagram of a reference map used in an embodiment of the present application.
  • the reference map shown in FIG. 3 is a schematic diagram of a high-precision map used to correct point cloud data.
  • the reference map as shown in FIG. 3 may be a two-dimensional map with an aerial top view angle presented by using a digital orthophoto map (DOM).
  • DOM digital orthophoto map
  • the reference maps used in the embodiments of the present application are not limited to digital orthophotos, and may also include high-reliability point cloud data, construction design drawings, and other forms of maps with high-precision and high-reliability location information.
  • DOM is based on the aerial (or aerospace) photos to correct each aerial photographic photo data to the digital ground model based on the pixel, eliminate the aerial photograph inclination error and the projection difference caused by the terrain undulation, and then go through mosaic and cutting.
  • DOM has the characteristics of high precision. For example, if the absolute spatial error of the 1:500 DOM is less than 0.3 meters, and the object resolution is greater than 0.05 meters, the internal error of the photo after de-distortion is more equal to 0. For objects within the DOM range, the relative error is 0.05 meters, and the absolute spatial error is 0.35 meters.
  • Figure 3 is a two-dimensional map generated from the DOM.
  • the two-dimensional map shown in Figure 3 converts real objects in the photo into two-dimensional graphics. In this way, it is convenient to zoom in and out of the map and to mark information on the map.
  • the accuracy of the two-dimensional map is the same as the DOM for generating the two-dimensional map.
  • the DOM in addition to the two-dimensional map generated from the DOM as shown in FIG. 3, which can be used as a reference map, the DOM can also be used as a reference map.
  • any high-precision map can be used as a reference map. For example, construction design drawings, maps based on manual measurements, or high-confidence point cloud data.
  • the following also introduces the technical solution of the present application by taking the two-dimensional map generated according to the DOM as shown in FIG. 3 as an example.
  • the DOM since the DOM is generated from aerial (or aerospace) photos, the DOM reflects a bird's-eye view of the ground. Therefore, some objects on the ground cannot be reflected in the DOM due to occlusion.
  • the two-dimensional map generated from the DOM does not have these occluded objects.
  • Fig. 4 is a schematic diagram of another reference map used in an embodiment of the present application.
  • Fig. 4 shows objects (such as traffic markings, traffic lights, and police booths) blocked by viaducts and traffic markings blocked by trees.
  • Fig. 5 is a schematic diagram of dividing the ground-level road in Fig. 4 into three parts.
  • the first part is the exposed part of the east-west road from the aerial view.
  • the second part is the area where the east-west road is obscured by the viaduct when viewed from the air.
  • the third part is the area of the north-south road that is blocked under the viaduct, except for the area that overlaps with the east-west road.
  • Fig. 6 is a schematic flowchart of a method for correcting point cloud data provided according to an embodiment of the present application.
  • Point cloud data set 1 Determine the N first reference features of the reference map in area 1 and the N first features of the point cloud data set (hereinafter referred to as point cloud data set 1) of the target map in area 1, where N is greater than or equal to 1. Is a positive integer.
  • the target map is a map obtained from the point cloud data in the point cloud data set.
  • Point cloud data set 1 is a collection of point cloud data included in area 1.
  • the point cloud data in area 1 is the range of the point cloud data collected by the sensor in one frame.
  • each of the regions 2 to 4 shown in FIG. 4 represents the range corresponding to the point cloud data collected by the sensor in one frame.
  • the N first reference features and the N first features have a one-to-one correspondence.
  • the first reference feature may be, for example, a specific object in a reference map, including but not limited to lane lines, road boundary lines, building boundaries, or traffic facilities; the first feature may be, for example, a specific object in the target map, Including but not limited to lane lines, road boundary lines, building boundaries or transportation facilities.
  • the method for determining the reference feature in the reference map and the feature in the point cloud data set can adopt the existing technology.
  • algorithms such as scale-invariant feature transform (SIFT) and accelerated robust features (speeded up robust features, SURF) can be used for feature extraction.
  • SIFT scale-invariant feature transform
  • SURF speeded up robust features
  • one feature or multiple features can be extracted by referring to an object in the map (for example, a section of traffic marking, a traffic light, etc.).
  • a point cloud data set corresponding to an object can extract one feature or multiple features.
  • each feature corresponds to an object.
  • feature matching is performed on the extracted features to obtain the corresponding relationship between the first reference feature and the first feature.
  • all features in the reference map and the point cloud data set 1 that have a corresponding relationship may be determined first, and then the N first reference features and the N first features are determined from all the features that have a corresponding relationship.
  • the point cloud data set 1 and the reference map share M groups of corresponding features, and each group of corresponding features includes a first candidate reference feature and a first candidate feature.
  • each group of corresponding features includes a first candidate reference feature and a first candidate feature.
  • M first candidate reference features in area 1 on the reference map M first candidate features in point cloud data set 1, the M first candidate reference features and the M first candidate features One-to-one correspondence.
  • the error threshold may be equal to the absolute spatial error of the DOM, the relative error of the objects within the DOM, or the absolute error of the objects within the DOM. Spatial error.
  • the error threshold may be equal to ⁇ , where ⁇ represents the absolute spatial error of the DOM, the relative error of the objects within the DOM, Or the absolute spatial error of objects within the DOM range, ⁇ is a coefficient.
  • can be an empirical value. For example, ⁇ can be a number greater than zero.
  • the error threshold may be an empirical value. For example, it can be 0.5 meters, 0.3 meters, or 0.4 meters.
  • the error between each of the M first candidate reference features and the corresponding first candidate feature may be determined, and the N groups of features with the largest error may be determined as the N first candidates Features and the N first features.
  • N can be a preset value.
  • M is equal to 10.
  • M there are a total of M first candidate reference features in the reference map, which are the first candidate reference feature 1 to the first candidate reference feature 10 respectively.
  • M first candidate features in the point cloud data set 1 which are the first candidate feature 1 to the first candidate feature 10, respectively.
  • the first candidate reference feature 1 corresponds to the first candidate feature 1
  • the first candidate reference feature 2 corresponds to the first candidate reference feature 2
  • the first candidate reference feature corresponds to the first candidate reference feature 3 and so on.
  • ⁇ 1 represents the error between the first candidate feature 1 and the first candidate reference feature 1
  • ⁇ 2 represents the error between the first candidate feature 2 and the first candidate reference feature 2 and so on.
  • N is equal to 5. Then the N first features are the first candidate feature 6 to the first candidate feature 10, and correspondingly, the N first reference features are the first candidate reference feature 6 to the first candidate reference feature 10.
  • the N groups of corresponding features closest to the point cloud data collection device can be determined.
  • N can be a preset value.
  • the distance from each first feature of the N first features to the point cloud data collection device is less than the distance from the features other than the N first features in the point cloud data set 1 to the point cloud data collection device the distance.
  • M is equal to 10.
  • M there are a total of M first candidate reference features in the reference map, which are the first candidate reference feature 1 to the first candidate reference feature 10 respectively.
  • M first candidate features in the point cloud data set 1 which are the first candidate feature 1 to the first candidate feature 10, respectively.
  • the first candidate reference feature 1 corresponds to the first candidate feature 1
  • the first candidate reference feature 2 corresponds to the first candidate reference feature 2
  • the first candidate reference feature corresponds to the first candidate reference feature 3, and so on.
  • D1 to D10 respectively represent the distances from the first candidate feature 1 to the first candidate feature 10 to the point cloud data collection device.
  • D1 to D10 have the following relationship: D1 ⁇ D2 ⁇ D3 ⁇ D4 ⁇ D5 ⁇ D6 ⁇ D7 ⁇ D8 ⁇ D9 ⁇ D10, and assume that N is equal to 5. Then the N first features are the first candidate feature 1 to the first candidate feature 5, and correspondingly, the N first reference features are the first candidate reference feature 1 to the first candidate reference feature 5.
  • the error between each of the M first candidate reference features and the corresponding first candidate feature may be determined first, and multiple sets of features whose errors are greater than the error threshold may be selected. Then, from the multiple sets of features, the N sets of features closest to the point cloud data collection device are selected as the N first features and the N first reference features.
  • ⁇ 1 to ⁇ 10 (the meaning of ⁇ 1 to ⁇ 10 is the same as above) and the error threshold ⁇ th have the following relationship: ⁇ 1 ⁇ 2 ⁇ th ⁇ 3 ⁇ 4 ⁇ 5 ⁇ 6 ⁇ 7 ⁇ 8 ⁇ 9 ⁇ 10; correspondingly, D1 to D10 (the meaning of D1 to D10 is the same as above) has the following relationship: D1 ⁇ D2 ⁇ D3 ⁇ D4 ⁇ D5 ⁇ D6 ⁇ D7 ⁇ D8 ⁇ D9 ⁇ D10, and assume that N is equal to 5. Then the N first features are the first candidate feature 3 to the first candidate feature 7, and correspondingly, the N first reference features are the first candidate reference feature 3 to the first candidate reference feature 7.
  • the N first reference features and the N first features are all features in the reference map and the point cloud data set 1 that have a corresponding relationship.
  • the i-th first feature can be adjusted according to the i-th candidate adjustment parameter set.
  • the adjusted pose of the i-th first feature in the point cloud data set 1 and the corresponding i-th first reference feature are in the reference map
  • the poses of are the same, or the pose of the adjusted i-th first feature in the point cloud data set 1 and the corresponding i-th first reference feature in the reference map are smaller than the error threshold.
  • FIG. 7 is a schematic diagram of N first reference features and N first features in area 1 shown in FIG. 4.
  • area 1 includes 9 first reference features and 9 first features in total. Each first reference feature and the corresponding first feature can form a feature group. As shown in Figure 7, there are 9 feature groups in area 1.
  • the first feature in the i-th feature group of the 9 feature groups can be adjusted to the pose of the corresponding first reference feature according to the candidate adjustment parameter set AD i.
  • AD i can include adjusting in the X-direction parameter X i and the Y-direction adjustment parameter Y i.
  • the first adjustment feature i X i in the X direction, the Y direction adjustment Y i can be adjusted to a pose of a first reference feature is located i.
  • FIG. 8 is a schematic diagram of N first reference features and N first features of area 1 shown in FIG. 4.
  • FIG. 8 is the result of adjusting the first feature 9 according to the parameter AD 9 on the basis of FIG. 7. As shown in FIG. 8, the adjusted first feature 9 and the first reference feature 9 overlap.
  • determining the first adjustment parameter set according to the determined N candidate adjustment parameter sets may be determining the average value of the N candidate adjustment parameter sets.
  • the average value of the N candidate adjustment parameter sets may be the arithmetic average, geometric average, or weighted average of the corresponding parameters in the N candidate adjustment parameter sets.
  • adjusting the first set of parameters may include parameters with in Is the average value of the adjustment parameters of the N candidate adjustment parameter sets in the X direction, It is the average value of the adjustment parameters of the N candidate adjustment parameter sets in the Y direction.
  • the weight of the N candidate adjustment parameter sets may be the error between the corresponding first feature and the first reference feature It is proportional. In other words, if the error between the first feature and the corresponding first reference feature is greater, the weight of the candidate adjustment parameter set corresponding to the first feature is greater.
  • the first adjustment parameter set may be the maximum value of the N candidate adjustment parameter sets.
  • the sum of the parameters in the X direction and the parameters in the Y direction of each adjustment parameter set in the N candidate adjustment parameter sets may be determined.
  • the candidate adjustment parameter set with the largest sum of the parameters in the X direction and the parameters in the Y direction may be used as the first adjustment parameter set.
  • the average value of the parameters in the X direction and the parameters in the Y direction of each adjustment parameter set in the N candidate adjustment parameter sets may be determined, and the candidate adjustment parameter set with the largest average value may be used as the first adjustment. Parameter set.
  • the first adjustment parameter set may be the median of N candidate adjustment parameter sets.
  • the sum of the parameters in the X direction and the parameters in the Y direction of each adjustment parameter set in the N candidate adjustment parameter sets may be determined. Arrange the sum of the parameters in the X direction and the parameters in the Y direction in descending order, and the candidate adjustment parameter set corresponding to the median can be used as the first adjustment parameter set. In other embodiments, the average value of the parameters in the X direction and the parameters in the Y direction of each adjustment parameter set in the N candidate adjustment parameter sets can be determined, and the candidate adjustment parameter set corresponding to the median of the average value can be used as The first adjustment parameter set.
  • a weighted sum operation may be performed on the N candidate adjustment parameter sets.
  • the result of the weighted sum operation can be used as the first adjustment parameter set.
  • a first feature (may be referred to as a second feature) that is closest to the point cloud data collection device among the N first features can be determined.
  • the first reference feature corresponding to the second feature among the N first reference features may be referred to as the second reference feature.
  • the first adjustment parameter set may be determined according to the position of the second reference feature in the reference map and the position of the second feature in the point cloud data set 1.
  • the error threshold is not used when determining the N first features and the N first reference features, then it can be determined that the error value of the N first features and the N first reference features is greater than The characteristics of the error threshold. For example, suppose that the error values of the K first features and the corresponding K first reference features in the N first features are greater than the error threshold. Then the first adjustment parameter set can be determined according to the positions of the K first reference features in the reference map and the poses of the K first features in the point cloud data set 1, where K is a positive integer less than or equal to N .
  • the manner of determining the first adjustment parameter set based on the K first reference features and the K first characteristics is similar to the manner of determining the first adjustment parameter set based on the N first reference features and the N first features.
  • K candidate adjustment parameter sets may be determined, and the first adjustment parameter set may be determined according to the K candidate adjustment parameter sets.
  • the first adjustment parameter set may be determined according to the first feature closest to the point cloud data collection device among the K first features and the corresponding first reference feature.
  • 603 Adjust the position information of the point cloud data in the point cloud data set 1 according to the first adjustment parameter set.
  • the N first features can be adjusted according to the first adjustment parameter set.
  • the position information of the point cloud data other than the N first features in the point cloud data set 1 may be adjusted according to the first adjustment parameter set.
  • the three unmatched point cloud data shown in FIG. 7 indicate that there is no feature corresponding to the unmatched point cloud data in the reference map.
  • the pose of the feature in point cloud dataset 1 can be corrected, so that the pose of the feature in point cloud dataset 1 in point cloud dataset 1 and the corresponding reference feature are in reference
  • the poses in the map are the same or the error is within the allowable range (for example, less than the error threshold).
  • the method shown in Figure 6 can also be used to adjust the points without corresponding reference features, so that the pose of the point cloud data in the point cloud data set is the same as the actual pose of the corresponding object or the error is allowed. In the range.
  • using the method shown in Figure 6 can improve the accuracy of point cloud data and reduce errors. For example, the errors before and after correction using the method shown in FIG. 6 are shown in Table 1.
  • Relative horizontal error The difference between the positional relationship and the true positional relationship between the objects collected within the scope of the map collection operation is called the relative error.
  • the horizontal relative error is the relative error in the horizontal direction of general movement called the horizontal relative error.
  • Absolute horizontal error The objects collected in the map collection are in the absolute coordinate system established with the center of the earth as the center. The difference between the absolute position measured by collecting and the actual position in the absolute coordinate system.
  • the horizontal absolute error is the relative error in the horizontal direction of general motion called the horizontal absolute error.
  • the method shown in Figure 6 can significantly reduce the maximum horizontal relative error and the maximum horizontal absolute error.
  • the application of the method shown in Fig. 6 does not need to increase the cost of collecting point cloud data.
  • a level sensor needs to be installed.
  • mapping surveying and mapping also requires obliquely mounted sensors to make the ground have sufficient point cloud density. In other words, if you need to build a loop to repair the cumulative error, then at least one horizontal sensor and one oblique sensor need to be installed on the point cloud data collection device, which increases the cost of the point cloud data collection device.
  • the point cloud in the method shown in FIG. 6 can be collected using only one sensor, for example, only one diagonally mounted sensor can be used.
  • FIG. 6 can be implemented using only a point cloud collected by one sensor, in order to improve accuracy, multiple sensors can also be set on the point cloud data collection device to collect the point cloud.
  • the loopback can only eliminate errors through closed point matching.
  • the cumulative error in the middle of the trajectory will be reduced and will not be eliminated.
  • the errors in the point cloud data of each frame can be corrected.
  • the relative cumulative error in the middle of the loop is related to the scene and the sensor and is not easy to evaluate.
  • Fig. 9 is a schematic flowchart of another method for correcting point cloud data provided according to an embodiment of the present application.
  • the reference map in area 2 is blocked by the viaduct. Therefore, the reference feature in area 2 cannot be extracted from the reference map. Therefore, the features in the point cloud data set 2 do not have the corresponding features in the reference map in the region 2.
  • the point cloud data can be corrected using the adjustment parameter set determined according to the reference map.
  • the above technical solution can also modify the point cloud data.
  • the above technical solution does not need to install multiple sensors, does not need to construct a loop, and the corrected point cloud data is also known.
  • FIG. 10 is a schematic flowchart of another method for correcting point cloud data provided according to an embodiment of the present application.
  • area 1 is located on an east-west road (may be referred to as road 1).
  • Area 3 is located on the road under the viaduct (may be called road 2).
  • Area 2 is located at the intersection of road 1 and road 2. Similar to area 2, the reference map is blocked by a viaduct in area 3. Therefore, the reference feature in area 3 cannot be extracted from the reference map. Therefore, the features in the point cloud data set 3 do not correspond to the features in the reference map in the region 3.
  • the second set of adjustment parameters may be the same as the first set of adjustment parameters.
  • the second adjustment parameter set may be determined based on multiple adjustment parameter sets including the first adjustment parameter set.
  • area 2 is only a partial area located in the intersection area.
  • the intersection area also includes other areas, such as area 6 and area 8.
  • the method shown in FIG. 6 can also be used to determine other adjustment parameter sets.
  • the method shown in FIG. 6 may be used to determine two adjustment parameter sets (may be referred to as adjustment parameter set 5 and adjustment parameter set 7 respectively).
  • the determination of the second adjustment parameter set according to the first adjustment parameter set may be: according to the first adjustment parameter set, the adjustment parameter set 5 and the adjustment parameter set 7 determine the second adjustment parameter set.
  • the manner of determining the second adjustment parameter set based on multiple adjustment parameter sets is similar to the manner of determining the first adjustment parameter set based on multiple candidate adjustment parameter sets. For example, you can determine the average, median, and maximum of multiple adjustment parameter sets, calculate a weighted sum, and so on. For the sake of brevity, I won't repeat them here.
  • the method shown in FIG. 10 can correct uncorrected point cloud data based on the corrected point cloud data.
  • the above technical solution can also modify the point cloud data.
  • the above technical solution does not need to install multiple sensors, does not need to construct a loop, and the corrected point cloud data is also known.
  • FIG. 11 is a schematic flowchart of another method for correcting point cloud data provided according to an embodiment of the present application.
  • the reference map is blocked by the viaduct in area 2. Therefore, the reference feature in area 2 cannot be extracted from the reference map. Therefore, the features in the point cloud data set 2 do not have the corresponding features in the reference map in the region 2.
  • the reference map can include some objects that are not occluded. Therefore, the reference feature in the area 4 (which can be called the fourth reference feature) in the reference map and the feature in the point cloud data set 4 (which can be called the fourth feature) may have the following situations:
  • one or more fourth features do not have a corresponding fourth reference feature.
  • the objects in area 4 are blocked by trees. Therefore, there is no fourth reference feature determined based on these objects in the reference map.
  • point cloud data collection equipment can collect objects corresponding to these objects blocked by trees. Therefore, the point cloud data 4 includes fourth features corresponding to these objects. However, these fourth features do not have corresponding fourth reference features.
  • Case 2 One or more fourth reference features do not have corresponding fourth features.
  • the fourth reference feature can be obtained from the reference map.
  • the point cloud data collection device failed to collect the point cloud of the object corresponding to the fourth reference feature. Therefore, the point cloud data 4 does not include the fourth features corresponding to these objects.
  • the reference map has T fourth reference features in area 4. Although T fourth features can be found one-to-one corresponding to these T fourth reference features, each fourth reference feature corresponds to the fourth reference feature.
  • the number of matching points between features is less than or equal to the matching point threshold.
  • the matching point threshold can be a preset number.
  • the matching point threshold may be a positive integer greater than or equal to 5 and less than or equal to 15.
  • the matching point threshold can be 5, 6, or 7. If the number of matching points between the fourth reference feature and the corresponding fourth feature is less than the matching point threshold, the error between the two features cannot be determined, or it is determined that the error and the actual error range are larger. Therefore, it can also be considered that the T fourth reference features have no corresponding fourth features.
  • each fourth reference feature in the region 4 in the reference map and each fourth feature in the point cloud data set 4 meet one of the above-mentioned cases 1 to 3.
  • part of the fourth reference feature in the area 4 of the reference map and part of the fourth feature in the point cloud data set 4 conform to one of the foregoing cases 1 to 3.
  • Another part of the fourth reference feature in the area 4 of the reference map corresponds to another part of the fourth feature in the point cloud data 4 in a one-to-one correspondence.
  • the error between these fourth reference features and the corresponding fourth feature is less than or equal to the error threshold.
  • This fourth reference feature and the fourth feature can also be considered as not corresponding.
  • the point cloud data can be corrected using the adjustment parameter set determined according to the reference map when the reference map cannot be used.
  • the above technical solution can also modify the point cloud data.
  • the above technical solution does not need to install multiple sensors, does not need to construct a loop, and the corrected point cloud data is also known.
  • the correction process of the point cloud data on the viaduct is the same as the correction process of area 1. For the sake of brevity, it will not be repeated here.
  • Fig. 12 is a schematic flowchart of a method for correcting point cloud data according to an embodiment of the present application.
  • the reference map is a digital orthophoto map, or the reference map is high-confidence point cloud data; or, the reference map is a construction design drawing.
  • the method further includes: adjusting the position information of each point cloud data in a second point cloud data set according to the first adjustment parameter set, where the second point cloud data set indicates that the target map is in the second area Collection of point cloud data.
  • the method before adjusting the position information of each point cloud data in the second point cloud data set according to the first adjustment parameter set, the method further includes: determining the reference map and the second point cloud data set There is no corresponding feature between them.
  • the first area and the second area belong to a first road
  • the method further includes: determining a second adjustment parameter set according to the first adjustment parameter set; and adjusting the second adjustment parameter set according to the second adjustment parameter set.
  • the location information of each point cloud data in a three point cloud data set, the third point cloud data set is a collection of point cloud data of the target map in a third area, the third area and the second area belong to the second road, And the second area is located at the intersection of the first road and the second road.
  • the determining the first adjustment parameter set according to the poses of the N first reference features in the reference map and the poses of the N first features in the first point cloud data set includes : Determine the second feature, which is the first feature closest to the point cloud data collection device among the N first features; according to the pose of the second reference feature in the reference map and the location of the second feature The pose in the first point cloud data set determines a first adjustment parameter set, and the second reference feature is a first reference feature corresponding to the second feature among the N first reference features.
  • the determining the first adjustment parameter set according to the poses of the N first reference features in the reference map and the poses of the N first features in the first point cloud data set includes : According to the poses of the N first reference features in the reference map and the poses of the N first features in the first point cloud data set, determine the difference between the corresponding first reference features in the reference map K first features with an error greater than the error threshold, where K is a positive integer less than or equal to N; according to the poses of the K first features in the first point cloud data set, and the same as the K first features A corresponding pose of the K first reference features in the reference map is used to determine the first adjustment parameter set.
  • Fig. 13 is a schematic structural block diagram of an apparatus for correcting point cloud data according to an embodiment of the present application.
  • the apparatus 1300 for correcting point cloud data as shown in FIG. 13 may be used to execute the method for correcting point cloud data described in the foregoing embodiments.
  • the apparatus 1300 includes an acquiring unit 1301 and a processing unit 1302.
  • the obtaining unit 1301 is used to obtain a reference map and a target map.
  • the processing unit 1302 is configured to determine the N first reference features of the reference map in the first area and the N first features of the target map in the first point cloud data set of the first area, the first point cloud
  • the data set is a collection of point cloud data of the first region, the N first reference features and the N first features have a one-to-one correspondence, and N is a positive integer.
  • the processing unit 1302 is further configured to determine a first adjustment parameter set according to the poses of the N first reference features in the reference map and the poses of the N first features in the first point cloud data set.
  • the processing unit 1302 is further configured to adjust the position information of each point cloud data in the first point cloud data set according to the first adjustment parameter set.
  • the reference map is a digital orthophoto map, or the reference map is high-confidence point cloud data; or, the reference map is a construction design drawing.
  • the processing unit 1302 is further configured to adjust the position information of each point cloud data in the second point cloud data set according to the first adjustment parameter set.
  • the second point cloud data set indicates that the target map is in the first The collection of point cloud data in the second area.
  • the processing unit 1302 is further configured to determine the reference map and the second point cloud before adjusting the position information of each point cloud data in the second point cloud data set according to the first adjustment parameter set There is no corresponding feature between the data sets.
  • the first area and the second area belong to the first road
  • the processing unit 1302 is further configured to determine a second adjustment parameter set according to the first adjustment parameter set; according to the second adjustment parameter set, Adjust the position information of each point cloud data in the third point cloud data set.
  • the third point cloud data set is a collection of point cloud data of the target map in the third area.
  • the third area and the second area belong to the second area. Road, and the second area is located at the intersection of the first road and the second road.
  • the processing unit 1302 is specifically configured to determine a second feature, and the second feature is the first feature that is closest to the point cloud data collection device among the N first features; Determine the first adjustment parameter set by referring to the pose in the map and the pose of the second feature in the first point cloud data set, and the second reference feature is the N first reference features corresponding to the second feature The first reference feature.
  • the processing unit 1302 is specifically configured to determine the position of the i-th first reference feature in the reference map according to the position of the i-th first reference feature among the N first reference features and the i-th first feature among the N first features.
  • the processing unit 1302 is specifically configured to determine and according to the pose of the N first reference features in the reference map and the pose of the N first features in the first point cloud data set The K first features in the reference map whose errors between the corresponding first reference features are greater than the error threshold, where K is a positive integer less than or equal to N; according to the K first features in the first point cloud data set The pose, and the poses of the K first reference features corresponding to the K first features one-to-one in the reference map, determine the first adjustment parameter set.
  • the device 1300 for correcting point cloud data may be a server or a vehicle, and may also be a component in the server or a vehicle.
  • the component includes a chip, such as a system on chip (SoC), a central processing unit ( Central processor unit, CPU) implementation, application-specific integrated circuit (ASIC), or programmable logic device (programmable logic device, PLD), the above PLD can be a complex programmable logical device (CPLD) , Field-programmable gate array (FPGA), generic array logic (GAL) or any combination thereof.
  • SoC system on chip
  • CPU central processing unit
  • ASIC application-specific integrated circuit
  • PLD programmable logic device
  • the above PLD can be a complex programmable logical device (CPLD) , Field-programmable gate array (FPGA), generic array logic (GAL) or any combination thereof.
  • CPLD complex programmable logical device
  • FPGA Field-programmable gate array
  • GAL generic array logic
  • the software or firmware includes but is not limited to computer program instructions or codes, and can be executed by a hardware processor.
  • the hardware includes, but is not limited to, various integrated circuits, such as a central processing unit (CPU), a digital signal processor (DSP), a field programmable gate array (FPGA), or an application specific integrated circuit (ASIC).
  • CPU central processing unit
  • DSP digital signal processor
  • FPGA field programmable gate array
  • ASIC application specific integrated circuit
  • the apparatus 1300 for correcting point cloud data may correspond to the method described in the embodiment of the present application, and the above-mentioned and other operations and/or functions of each unit in the apparatus 1300 for correcting point cloud data are respectively implemented to implement For the sake of brevity, the corresponding process of the above method will not be repeated here.
  • Fig. 14 is a structural block diagram of an apparatus for correcting point cloud data provided according to an embodiment of the present application.
  • the device 1400 for correcting point cloud data shown in FIG. 14 can be used to execute the method for correcting point cloud data described in the foregoing embodiments.
  • the device 1400 includes: a processor 1401, a memory unit 1402, and a storage medium 1403.
  • the processor 1401, the memory unit 1402, and the storage medium 1403 can communicate through a bus 1404.
  • the processor 1401 is the control center of the computing device 1400, and provides sequencing and processing facilities for executing instructions, executing interrupt actions, providing timing functions, and other functions.
  • the processor 1401 includes one or more central processing units (CPUs). CPU 0 and CPU 1 as shown in Figure 14.
  • the computing device 1400 includes multiple processors.
  • the processor 1401 may be a single-core (single CPU) processor or a multi-core (multi-CPU) processor. Unless otherwise specified, the processor or memory used to perform tasks may be implemented as temporarily configured A general-purpose component that performs a task at a given time or a specific component that is manufactured to perform a task.
  • the term "processor" refers to one or more devices or circuits.
  • the processor 1401 may also be other general-purpose processors, such as SOC, ASIC, FPGA or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, and so on.
  • the general-purpose processor may be a microprocessor or any conventional processor.
  • the program code executed by the CPU of the processor 1401 may be stored in the memory unit 1402 or the storage medium 1403.
  • the program code (for example, the kernel, the program to be debugged) is stored in the storage medium 1403 and copied to the memory unit 1402 for the processor 1401 to execute.
  • the processor 1401 can execute at least one operating system, and the operating system may be LINUX TM , UNIX TM, or the like.
  • the processor 1401 controls the execution of other programs or processes, controls communication with peripheral devices, and controls the use of data processing device resources, thereby controlling the operation of the computing device 1400, so as to implement the operation steps of the above method.
  • bus 1404 may also include a power bus, a control bus, and a status signal bus. However, for clear description, various buses are marked as bus 1404 in the figure.
  • the device 1400 for correcting point cloud data further includes a communication interface (not shown in the figure), and the communication interface is used to implement communication between the device 1400 for correcting point cloud data and external devices or equipment, for example, correcting point cloud data.
  • the data acquisition device 1400 can communicate with a point cloud data acquisition device for the point cloud data collected by the point cloud data acquisition device.
  • the present application also provides a computer program product, which when the computer program product runs on a processor, causes the device for correcting point cloud data to execute the above method.
  • the present application also provides a computer-readable medium, the computer-readable medium includes computer instructions, when the computer instructions are executed by a processor, the device for correcting point cloud data executes the above method.
  • the above-mentioned embodiments may be implemented in whole or in part by software, hardware, firmware or any other combination.
  • the above-mentioned embodiments may be implemented in the form of a computer program product in whole or in part.
  • the computer program product includes one or more computer instructions.
  • the computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices.
  • the computer instructions may be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium.
  • the computer instructions may be transmitted from a website, computer, server, or data center.
  • the computer-readable storage medium may be any available medium that can be accessed by a computer or a data storage device such as a server or a data center that includes one or more sets of available media.
  • the usable medium may be a magnetic medium (for example, a floppy disk, a hard disk, a magnetic tape), an optical medium (for example, a DVD), or a semiconductor medium.
  • the semiconductor medium may be a solid state drive (SSD).

Landscapes

  • Engineering & Computer Science (AREA)
  • Radar, Positioning & Navigation (AREA)
  • Remote Sensing (AREA)
  • Automation & Control Theory (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Navigation (AREA)
  • Instructional Devices (AREA)

Abstract

本申请提供了一种校正点云数据的方法和装置。本申请利用参考地图对点云数据集合中的点云数据进行校正,可以在不增加采集点云数据的成本的情况下,对点云数据进行校正。本申请在校正点云数据的过程中无需构建回环,从而提高了点云数据的采集效率,并且本申请中校正的点云数据的精度是可以控制的。

Description

校正点云数据的方法和相关装置 技术领域
本申请涉及自动驾驶和智能网联车技术领域,更具体地,涉及校正点云数据的方法和相关装置。
背景技术
点云是通过测量仪器(例如摄像头、激光雷达等)得到的物体表面的点的数据的集合。点云数据可以包括物体的三维坐标信息。点云数据可以用于进行目标检测与识别。例如,点云数据可以用于识别场景中的汽车、道路交通标线、道路交通标志等信息。因此,点云数据可以用于自动驾驶、智能机器人导航等领域。
自动驾驶***要对前方的道路状况进行精确地预判。对于物理传感器感知不到的范围,也需要提供相应的信息。因此,自动驾驶***需要预先获取高精度地图作为自动驾驶提供的先验知识。
目前,高精度地图的制作过程中,地图的精度是由全球卫星定位***(Global Navigation Satellite System,GNSS)、实时差分技术(Real-Time Kinematic,RTK)或惯性导航***(Inertial Navigation System,INS)确定。但是在一些场景下,由于遮挡、多径效应等影响,现有的***的精度会受到影响,进而影响所制作的地图的精度。
发明内容
本申请提供一种校正点云数据的方法和相关装置,能够对点云数据进行校正,从而提高高进度地图的精度。
第一方面,本申请实施例提供一种校正点云数据的方法,包括:确定参考地图在第一区域中的N个第一参考特征和目标地图在该第一区域的第一点云数据集中的N个第一特征,该第一点云数据集为该第一区域的点云数据的集合,该N个第一参考特征和该N个第一特征一一对应,N为正整数;根据该N个第一参考特征在该参考地图中的位姿和该N个第一特征在该第一点云数据集中的位姿,确定第一调整参数集;根据该第一调整参数集,调整该第一点云数据集中的每个点云数据的位置信息。上述技术方案可以在不增加采集点云数据的成本的情况下,对点云数据进行校正。进一步,上述技术方案在校正点云数据的过程中无需构建回环,从而可以提供点云数据的采集效率。此外,利用上述技术方案校正的点云数据的精度是可以控制的。
结合第一方面,在第一方面的一种可能的实现方式中,该参考地图为数字正射影像图,或者,该参考地图为高可信度点云数据;或者,该参考地图为施工设计图。上述技术方案利用高精度地图作为用于校正点云数据的参考地图。这样,可以保证校正后的点云数据的精度。
结合第一方面,在第一方面的一种可能的实现方式中,该方法还包括:根据该第一调 整参数集,调整第二点云数据集中每个点云数据的位置信息,该第二点云数据集为该目标地图在第二区域的点云数据的集合。
结合第一方面,在第一方面的一种可能的实现方式中,在该根据该第一调整参数集,调整第二点云数据集中每个点云数据的位置信息之前,该方法还包括:确定该参考地图与该第二点云数据集之间没有相对应的特征。上述技术方案可以通过能够利用参考地图校正的点云数据,对没有无法利用参考地图的点云数据进行校正。因此上述技术方案也可以在不增加采集点云技术的成本的情况下,对点云数据进行校正。进一步,上述技术方案在校正点云数据的过程中无需构建回环,从而可以提供点云数据的采集效率。
结合第一方面,在第一方面的一种可能的实现方式中,该第一区域与该第二区域属于第一道路,该方法还包括:根据该第一调整参数集,确定第二调整参数集;根据该第二调整参数集,调整第三点云数据集中每个点云数据的位置信息,该第三点云数据集为该目标地图在第三区域的点云数据的集合,该第三区域与该第二区域属于第二道路,并且该第二区域位于该第一道路与该第二道路的交叉区域。所述第一调整参数集与所述第二调整参数集可能相同,也可能不同。上述技术方案可以通过能够利用参考地图校正的点云数据,对没有无法利用参考地图的点云数据进行校正。因此上述技术方案也可以在不增加采集点云技术的成本的情况下,对点云数据进行校正。进一步,上述技术方案在校正点云数据的过程中无需构建回环,从而可以提供点云数据的采集效率。
结合第一方面,在第一方面的一种可能的实现方式中,该根据该N个第一参考特征在该参考地图中的位姿和该N个第一特征在该第一点云数据集中的位姿,确定第一调整参数集,包括:确定第二特征,该第二特征为该N个第一特征中距离点云数据采集设备最近的一个第一特征;根据第二参考特征在该参考地图中的位姿和该第二特征在该第一点云数据集中的位姿,确定第一调整参数集,该第二参考特征为该N个第一参考特征中与该第二特征对应的第一参考特征。
结合第一方面,在第一方面的一种可能的实现方式中,该根据该N个第一参考特征在该参考地图中的位姿和该N个第一特征在该第一点云数据集中的位姿,确定第一调整参数集,包括:根据该N个第一参考特征中第i个第一参考特征在该参考地图中的位姿和该N个第一特征中第i个第一特征在该第一点云数据集中的位姿,确定第i个候选调整参数集,其中该第i个第一参考特征与该第i个第一特征相对应,i=1,…,N;根据N个该候选调整参数集,确定该第一调整参数集。
结合第一方面,在第一方面的一种可能的实现方式中,该根据该N个第一参考特征在该参考地图中的位姿和该N个第一特征在该第一点云数据集中的位姿,确定第一调整参数集,包括:根据该N个第一参考特征在该参考地图中的位姿和该N个第一特征在该第一点云数据集中的位姿,确定与参考地图中的对应第一参考特征之间的误差大于误差阈值的K个第一特征,该K为小于或等于N的正整数;根据该K个第一特征在该第一点云数据集中的位姿,以及与该K个第一特征一一对应的K个第一参考特征在该参考地图中的位姿,确定该第一调整参数集。
第二方面,本申请实施例提供一种校正点云数据的装置,该校正点云数据的装置包括用于实现第一方面或第一方面的任一种可能的实现方式的单元。
第三方面,本申请实施例提供一种校正点云数据的装置,该校正点云数据的装置包括 处理器。所述处理器用于与存储器耦合,读取并执行所述存储器中的计算机程序指令,以实现上述第一方面的方法设计中任意一种可能的实现方式中的方法。
第四方面,本申请实施例提供一种计算机可读介质,所述计算机可读介质包括计算机指令,当所述计算机指令在被处理器运行时,使得校正点云数据的装置执行上述第一方面的方法设计中任意一种可能的实现方式中的方法。
第五方面,本申请实施例提供一种计算机程序产品,当所述计算机程序产品在处理器上运行时,使得校正点云数据的装置执行上述第一方面的方法设计中任意一种可能的实现方式中的方法。
附图说明
图1是本申请实施例应用于车辆侧的应用场景以及本申请实施例在地图***架构中的作用位置的示意图。
图2是点云数据位置信息偏差的示意图。
图3是本申请实施例使用的一个参考地图的示意图。
图4是本申请实施例使用的另一个参考地图的示意图。
图5是将图4中的地面层道路划分为三个部分的示意图。
图6是本申请实施例提供的校正点云数据的方法的流程图。
图7是图4所示的区域1中的N个第一参考特征和N个第一特征的一个示意图。
图8是图4所示的区域1中的N个第一参考特征和N个第一特征的另一个示意图。
图9是本申请实施例提供的另一校正点云数据的方法的流程图。
图10是本申请实施例提供的另一校正点云数据的方法的流程图。
图11是本申请实施例提供的另一校正点云数据的方法的流程图。
图12是本申请实施例提供的另一种校正点云数据的方法的流程图。
图13是根据本申请实施例提供的一种校正点云数据的装置的示意性结构框图。
图14是根据本申请实施例提供的校正点云数据的装置的结构框图。
具体实施方式
下面将结合附图,对本申请中的技术方案进行描述。
本申请将围绕可包括多个设备、组件、模块等的***来呈现各个方面、实施例或特征。应当理解和明白的是,各个***可以包括另外的设备、组件、模块等,并且/或者可以并不包括结合附图讨论的所有设备、组件、模块等。此外,还可以使用这些方案的组合。
另外,在本申请实施例中,“示例的”、“例如”等词用于表示作例子、例证或说明。本申请中被描述为“示例”的任何实施例或设计方案不应被解释为比其它实施例或设计方案更优选或更具优势。确切而言,使用示例的一词旨在以具体方式呈现概念。
本申请实施例中,“相应的(corresponding,relevant)”和“对应的(corresponding)”有时可以混用,应当指出的是,在不强调其区别时,其所要表达的含义是一致的。
本申请实施例中,有时候下标如W 1可能会笔误为非下标的形式如W1,在不强调其区别时,其所要表达的含义是一致的。
本申请实施例描述的网络架构以及业务场景是为了更加清楚的说明本申请实施例的 技术方案,并不构成对于本申请实施例提供的技术方案的限定,本领域普通技术人员可知,随着网络架构的演变和新业务场景的出现,本申请实施例提供的技术方案对于类似的技术问题,同样适用。
在本说明书中描述的参考“一个实施例”或“一些实施例”等意味着在本申请的一个或多个实施例中包括结合该实施例描述的特定特征、结构或特点。由此,在本说明书中的不同之处出现的语句“在一个实施例中”、“在一些实施例中”、“在其他一些实施例中”、“在另外一些实施例中”等不是必然都参考相同的实施例,而是意味着“一个或多个但不是所有的实施例”,除非是以其他方式另外特别强调。术语“包括”、“包含”、“具有”及它们的变形都意味着“包括但不限于”,除非是以其他方式另外特别强调。
本申请中,“至少一个”是指一个或者多个,“多个”是指两个或两个以上。“和/或”,描述关联对象的关联关系,表示可以存在三种关系,例如,A和/或B,可以表示:单独存在A,同时存在A和B,单独存在B的情况,其中A,B可以是单数或者复数。字符“/”一般表示前后关联对象是一种“或”的关系。“以下至少一项(个)”或其类似表达,是指的这些项中的任意组合,包括单项(个)或复数项(个)的任意组合。例如,a,b,或c中的至少一项(个),可以表示:a,b,c,a-b,a-c,b-c,或a-b-c,其中a,b,c可以是单个,也可以是多个。
本申请实施例中所称的高精度地图是指至少能够提供车道级别导航的地图。通常情况下,误差小于30厘米的地图就可以为自动驾驶***提供车道级别的导航。例如,误差小于25厘米,15厘米或者10厘米的地图。
本申请实施例既可以用于车辆侧,也可以用于网络侧。图1中的(a)是本申请实施例应用于车辆侧的应用场景示意图。如图1所示,在车辆110中可以安装测量***120和计算设备130。
测量***120可以包括用于检测和扫描目标场景中的点云的传感器。作为示例,上述传感器可以包括(light detection and ranging,LiDAR)、三维扫描仪、深度相机等,本申请不作限定。传感器采集到的点云数据的集合可以称为原始点云数据集。
测量***120还可以包括GNSS/RTK模块。GNSS/RTK模块用于获取车辆110的位置数据。
测量***120还可以包括惯性测量单元(inertial measurement unit,IMU)。IMU用于获取车辆110的姿态信息。
计算设备130与测量***120相连,用于从测量***120获取原始点云数据集,车辆110的位置数据以及车辆110的姿态信息,并对原始点云数据集、位置数据以及姿态信息进行融合,得到融合后的点云数据集。本申请实施例中所称的点云数据集是指融合后的点云数据集。在对原始点云数据集进行融合的过程中可以进行删除离群点和/或按照现有方式(例如进行同步定位与建图(simultaneous localization and mapping,SLAM)回环修复)校正点云数据等操作。
图1中的(b)是本申请实施例在地图***架构中的作用位置的示意图。
在一些情况下,GNSS/RTK模块无法获取车辆110的准确的位置信息。在另一些情况下,IMU模块采集的姿态信息可能会出现累计误差。因此,点云数据与实际的信息会产生一些偏差。如图1中的(b)所示,本申请的技术方案可以用于对点云数据进行校正,以减少点云数据的误差,提高地图精度。例如,图2是点云数据位置信息偏差的示意图。
如图2所示,点云数据与实际的道路交通标志线会出现一定的差异,所述差异包括位置差异和姿态差异两方面,所述位置差异体现为车道线中心点的位置偏差,所述姿态差异体现为车道线指向方向的角度偏差。
下面结合图3至图6,对本申请的技术方案进行介绍。
图3是本申请实施例使用的一个参考地图的示意图。如图3所示的参考地图是用于对点云数据进行校正的一种高精度地图的示意图。如图3所示的参考地图可以是利用数字正射影像(digital orthophoto map,DOM)呈现的空中俯视角度的二维地图。本申请实施例所使用的参考地图不限于数字正射影像,还可以包括高可信度点云数据、施工设计图等其他形式的具有较高精度和较高可信度的位置信息的地图。
DOM是在航空(或航天)照片的基础上以像元为基础把每张航空摄影照片数据纠正到数字地面模型上,消除航摄照片倾斜误差和地形起伏引起的投影差,再经过镶嵌、切割,按一定图幅范围裁剪生成的数字正射影像集。所以在航空(或航天)照片具备影像特征的图像基础上,同时具备有地图几何精度,使图像中显像的对象具备准确的地理坐标。
DOM具有高精度的特点。例如,1:500的DOM的绝对空间误差小于0.3米,对象分辨率大于0.05米,去畸变后照片内误差越等于0。对于DOM范围内的对象的相对误差为0.05米,绝对空间误差为0.35米。
如上所述,DOM是基于航空(或航天)照片生成的。因此,DOM反映的是地面的真实物体。图3是根据DOM生成的二维地图。如图3所示的二维地图将照片中的真实物体转换为二维图形。这样,可以便于地图的缩放,而且便于在地图上标注信息。但由于该二维地图是根据DOM生成的,因此该二维地图的精度与生成该二维地图的DOM是相同的。
当然,除了如图3所示的根据DOM生成的二维地图可以作为参考地图外,DOM也可以作为参考地图。此外,只要是高精度地图都可以作为参考地图。例如,施工设计图、基于人工测量得到的地图或者高可信度点云数据等。
下面还以图3所示的根据DOM生成的二维地图为例对本申请的技术方案进行介绍。如上所述,由于DOM是根据航空(或航天)照片生成的,因此DOM反映的是地面的鸟瞰照片。因此,某些地面上的对象由于被遮挡的原因不能在DOM中反映出。相应的,根据DOM生成的二维地图也没有这些被遮挡的对象。
如图3所示的参考地图中可以看到南北向的高架桥以及高架桥上的交通标线,东西向道路,东西向道路上未被树木遮挡的交通标线以及东西向道路边的树木。
南北向高架桥下的交通标线被高架桥遮挡住,东西向道路上的部分交通标线被数目挡住。图4是本申请实施例使用的另一参考地图的示意图。图4示出了被高架桥遮挡的对象(例如交通标线、交通信号灯以及警亭)以及被树木遮挡的交通标线。
如图4所示的地图中的地面层道路可以划分为三个部分,分别为第一部分,第二部分和第三部分。图5是将图4中的地面层道路划分为三个部分的示意图。
第一部分是东西向道路在空中俯视角度露出的部分。第二部分是东西向道路在空中俯视角度被高架桥遮挡区域。第三部分是高架桥下面被遮挡的南北方向道路中除与东西向道路重叠的区域以外的区域。
下面将结合图6和图7对如何校正点云数据进行详细描述。
图6是根据本申请实施例提供的校正点云数据的方法的示意性流程图。
601,确定参考地图在区域1中的N个第一参考特征和目标地图在区域1中的点云数据集(以下简称点云数据集1)的N个第一特征,N为大于或等于1的正整数。目标地图是点云数据集中的点云数据得到的地图。点云数据集1是区域1包括的点云数据的集合。
区域1中的点云数据是传感器在一帧内采集到的点云数据所在的范围。类似的,如图4所示的区域2至区域4中的每个区域都代表传感器在一帧中采集到的点云数据对应的范围。
该N第一参考特征和该N个第一特征一一对应。所述第一参考特征例如可以为参考地图中的特定对象,包括但不限于车道线、道路边界线、建筑物的边界或交通设施;所述第一特征例如可以为目标地图中的特定对象,包括但不限于车道线、道路边界线、建筑物的边界或交通设施。
确定参考地图中的参考特征以及点云数据集中的特征的确定方式可以采用现有技术。例如,可以利用尺度不变特征变换(scale-invariant feature transform,SIFT)、加速稳健特征(speeded up robust features,SURF)等算法进行特征提取。
本领域技术人员可以理解,参考地图中的一个对象(例如一段交通标线、一个交通信号灯等)可以提取出一个特征也可以提取出多个特征。类似的,点云数据集中对应于一个对象(例如一段交通标线、一个交通信号灯等)可以提取出一个特征也可以提取出多个特征。但是为了便于描述,以下实施例中假设每个特征就是对应于一个对象。
在提取特征之后,对提取到的特征进行特征匹配,得到第一参考特征和第一特征之间的对应关系。
在一些实施例中,可以先确定参考地图和点云数据集1中全部具有对应关系的特征,然后从所有具有对应关系的特征中确定该N个第一参考特征和该N个第一特征。
为了便于描述,以下假设点云数据集1和参考地图中共有M组对应的特征,每组对应的特征包括一个第一候选参考特征和一个第一候选特征。换句话说,在参考地图在区域1中有M个第一候选参考特征,点云数据集1中有M个第一候选特征,该M个第一候选参考特征和该M个第一候选特征一一对应。
在一些实施例中,可以确定该M个第一候选参考特征中的每个第一候选参考特征和对应的第一候选特征的误差。如果第一候选参考特征和对应的第一候选特征的误差大于误差阈值,则该第一候选参考特征可以作为第一参考特征,对应的第一候选特征可以作为第一特征。换句话说,该N个第一参考特征中的每个第一参考特征与对应的第一特征的误差大于该误差阈值。
在一些实施例中,如果该参考地图是DOM或者根据DOM生成的二维地图,那么该误差阈值可以等于DOM的绝对空间误差,DOM范围内的对象的相对误差,或者DOM范围内的对象的绝对空间误差。
在另一些实施例中,如果该参考地图是DOM或者根据DOM生成的二维地图,那么该误差阈值可以等于α×Δ,其中Δ表示DOM的绝对空间误差,DOM范围内的对象的相对误差,或者DOM范围内的对象的绝对空间误差,α是一个系数。α可以是经验值。例如α可以是大于0的数。
在另一些实施例中,该误差阈值可以是一个经验值。例如,可以是0.5米,0.3米或者0.4米等。
在另一些实施例中,可以确定该M个第一候选参考特征中的每个第一候选参考特征和对应的第一候选特征的误差,确定误差最大的N组特征为该N个第一候选特征和该N个第一特征。N可以是一个预设值。
假设M等于10。换句话说,该参考地图中共有M个第一候选参考特征,分别为第一候选参考特征1至第一候选参考特征10。相应的,该点云数据集1中共有M个第一候选特征,分别为第一候选特征1至第一候选特征10。第一候选参考特征1与第一候选特征1对应,第一候选参考特征2与第一候选参考特征2对应,第一候选参考特征与第一候选参考特征3对应,以此类推。假设Δ1表示第一候选特征1与第一候选参考特征1的误差,Δ2表示第一候选特征2与第一候选参考特征2的误差,以此类推。假设Δ1至Δ10有如下关系:Δ1<Δ2<Δ3<Δ4<Δ5<Δ6<Δ7<Δ8<Δ9<Δ10,且假设N等于5。那么该N第一特征是第一候选特征6至第一候选特征10,相应的,该N个第一参考特征是第一候选参考特征6至第一候选参考特征10。
在另一些实施例中,可以确定距离点云数据采集设备(例如如图1所示的车辆110)最近的N组对应的特征。在此情况下,N可以是一个预设值。换句话说,该N个第一特征中的每个第一特征到该点云数据采集设备的距离小于点云数据集1中除该N个第一特征以外的特征到该点云数据采集设备的距离。
假设M等于10。换句话说,该参考地图中共有M个第一候选参考特征,分别为第一候选参考特征1至第一候选参考特征10。相应的,该点云数据集1中共有M个第一候选特征,分别为第一候选特征1至第一候选特征10。第一候选参考特征1与第一候选特征1对应,第一候选参考特征2与第一候选参考特征2对应,第一候选参考特征与第一候选参考特征3对应,以此类推。假设D1至D10分别表示第一候选特征1到第一候选特征10到点云数据采集设备的距离。假设D1至D10有如下关系:D1<D2<D3<D4<D5<D6<D7<D8<D9<D10,且假设N等于5。那么该N第一特征是第一候选特征1至第一候选特征5,相应的,该N个第一参考特征是第一候选参考特征1至第一候选参考特征5。
在另一些实施例中,可以先确定可以确定该M个第一候选参考特征中的每个第一候选参考特征和对应的第一候选特征的误差,选择误差大于误差阈值的多组特征。然后再从该多组特征中,选择出距离点云数据采集设备最近的N组特征作为该N个第一特征和该N个第一参考特征。
例如,Δ1至Δ10(Δ1至Δ10的含义同上)和误差阈值Δ th有如下关系:Δ1<Δ2<Δ th<Δ3<Δ4<Δ5<Δ6<Δ7<Δ8<Δ9<Δ10;相应的,D1至D10(D1至D10的含义同上)有如下关系:D1<D2<D3<D4<D5<D6<D7<D8<D9<D10,且假设N等于5。那么该N个第一特征是第一候选特征3至第一候选特征7,相应的,该N个第一参考特征是第一候选参考特征3至第一候选参考特征7。
在另一些实施例中,该N个第一参考特征和该N个第一特征是参考地图和点云数据集1中全部具有对应关系的特征。
602,根据该N个第一参考特征在参考地图中的位姿和该N个第一特征在点云数据集1中的位姿,确定第一调整参数集。
在一些实施例中,可以根据根据该N个第一参考特征中第i个第一参考特征在该参考 地图中的位姿和该N个第一特征中第i个第一特征在该点云数据集1中的位姿,确定第i个候选调整参数集,其中该第i个第一参考特征与所述第i个第一特征相对应,i=1,…,N。根据确定的N个候选调整参数集,确定该第一调整参数集。
第i个第一特征可以根据第i个候选调整参数集调整,调整后的第i个第一特征所在点云数据集1中的位姿与对应的第i个第一参考特征在参考地图中的位姿相同,或者,调整后的第i个第一特征所在点云数据集1中的位姿与对应的第i个第一参考特征在参考地图中的位姿小于误差阈值。
图7是图4所示的区域1中的N个第一参考特征和N个第一特征的示意图。
如图7所示,区域1中共包括9个第一参考特征和9个第一特征。每个第一参考特征和对应的第一特征可以组成一个特征组。如图7所示,区域1中共有9个特征组。例如,9个特征组中的第i个特征组中的第一特征可以根据候选调整参数集AD i调整到对应的第一参考特征的位姿。在一些实施例中,AD i可以包括在X方向上调整的参数X i和在Y方向调整的参数Y i。换句话说,第一特征i向X方向调整X i,向Y方向调整Y i就可以调整到第一参考特征i所在的位姿。
图8是图4所示的区域1的N个第一参考特征和N个第一特征的示意图。图8是在图7的基础上将第一特征9按照参数AD 9调整后得到的结果。如图8所示,调整后的第一特征9和第一参考特征9重合。
在一些实施例中,根据确定的N个候选调整参数集,确定该第一调整参数集可以是确定该N个候选调整参数集的平均值。该N个候选调整参数集的平均值可以是该N个候选调整参数集中的对应的参数的算数平均值、几何平均值或者加权平均值等。
还以AD i包括在X方向上调整的参数X i和在Y方向调整的参数Y i为例,该第一调整参数集可以包括参数
Figure PCTCN2020086728-appb-000001
Figure PCTCN2020086728-appb-000002
其中
Figure PCTCN2020086728-appb-000003
是N个候选调整参数集在X方向的调整参数的平均值,
Figure PCTCN2020086728-appb-000004
是N个候选调整参数集在Y方向的调整参数的平均值。
例如,如果该N个候选调整参数集的平均值是该N个候选调整参数集的加权平均值,那么该N个候选调整参数集的权重可以与对应的第一特征与第一参考特征的误差呈正比。换句话说,如果第一特征与对应的第一参考特征的误差越大,该第一特征对应的候选调整参数集的权重就越大。
在另一些实施例中,该第一调整参数集可以是N个候选调整参数集的中的最大值。
还以AD i包括在X方向上调整的参数X i和在Y方向调整的参数Y i为例。在一些实施例中,可以确定N个候选调整参数集中的每个调整参数集在X方向的参数和在Y方向的参数的和。X方向的参数和Y方向的参数的和最大的候选调整参数集可以作为该第一调整参数集。在另一些实施例中,可以确定N个候选调整参数集中的每个调整参数集在X方向的参数和在Y方向的参数的平均值,平均值最大的候选调整参数集可以作为该第一调整参数集。
在另一些实施例中,该第一调整参数集可以是N个候选调整参数集的中位数。
还以AD i包括在X方向上调整的参数X i和在Y方向调整的参数Y i为例。在一些实施例中,可以确定N个候选调整参数集中的每个调整参数集在X方向的参数和在Y方向的参数的和。将X方向的参数和Y方向的参数的和由大到小排列,中位数对应的候选调整参数集可以作为该第一调整参数集。在另一些实施例中,可以确定N个候选调整参数集中 的每个调整参数集在X方向的参数和在Y方向的参数的平均值,平均值的中位数对应的候选调整参数集可以作为该第一调整参数集。
在另一些实施例中,可以对该N个候选调整参数集进行加权求和运算。加权求和运算的结果可以作为该第一调整参数集。
在另一些实施例中,可以确定该N个第一特征中距离点云数据采集设备最近的一个第一特征(可以称为第二特征)。相应的,该N个第一参考特征中与该第二特征对应的第一参考特征可以称为第二参考特征。可以根据该第二参考特征在参考地图中的位置和第二特征在点云数据集1中的位置确定该第一调整参数集。
在另一些实施例中,如果在确定该N个第一特征和N个第一参考特征时没有用到误差阈值,那么可以确定该N个第一特征和N个第一参考特征中误差值大于误差阈值的特征。例如,假设该N个第一特征中的K个第一特征和对应的K个第一参考特征的误差值大于误差阈值。那么可以根据该K个第一参考特征在参考地图中的位置以及该K个第一特征在点云数据集1中的位姿确定该第一调整参数集,K为小于或等于N的正整数。
根据K个第一参考特征和K个第一特征确定第一调整参数集的方式与根据N个第一参考特征和N个第一特征确定第一调整参数集的方式类似。例如,可以确定K个候选调整参数集,根据K个候选调整参数集确定第一调整参数集。又如,可以根据K个第一特征中距离点云数据采集设备最近的第一特征和对应的第一参考特征确定该第一调整参数集。
603,根据该第一调整参数集,调整该点云数据集1中的点云数据的位置信息。
还以图7为例,在确定了第一调整参数集后,可以根据该第一调整参数集调整该N个第一特征。此外,还可以根据该第一调整参数集调整点云数据集1中除该N个第一特征以外的点云数据的位置信息。例如,如图7所示的三个未匹配点云数据的位置信息。如图7所示的三个未匹配点云数据表示在参考地图中没有与未匹配点云数据对应的特征。
通过图6所示的方法,可以对点云数据集1中的特征的位姿进行校正,使得点云数据集1中的特征在点云数据集1中的位姿与对应的参考特征在参考地图中的位姿相同或者误差在允许的范围内(例如小于误差阈值)。除此之外,利用图6所示的方法还可以调整没有对应参考特征的点,从而使得点云数据集中的点云数据的位姿与对应的对象的实际位姿是相同的或者误差在允许的范围内。综上所述,利用图6所示的方法可以提高点云数据的精度,缩小误差。例如,在利用图6所示方法进行修正之前和之后的误差如表1所示。
表1
Figure PCTCN2020086728-appb-000005
相对水平误差:地图采集作业范围内所采集得到的对象之间位置关系与真实的位置关系的差距,称为相对误差。而水平相对误差是以一般运动的水平方向的相对误差称为水平 相对误差。
绝对水平误差:地图采集所采集对象,在以地心为中心建立的绝对坐标系下。通过采集所测算的绝对位置,和真实的在绝对坐标系下的位置的差距。而水平绝对误差是以一般运动的水平方向的相对误差称为水平绝对误差。
如表1所示,利用图6所示的方法可以显著缩小最大水平相对误差和最大水平绝对误差。
此外,应用图6所示的方法不需要增加采集点云数据的成本。例如,在构建回环修复累计误差时,为了提高回环效果,需要增大用于采集点云的传感器与帧之间的共视特征。因此,需要安装水平传感器。此外,建图测绘还需要斜装传感器使得地面有足够的点云密度。换句话说,如果需要利用构建回环修复累计误差,那么至少需要在点云数据采集设备上安装一个水平传感器和一个斜装传感器,这样增加了点云数据采集设备的成本。而图6所示的方法中的点云可以只使用一个传感器采集,例如可以只使用一个斜装传感器。
本领域人员可以理解,虽然图6所示的方法可以只使用一个传感器采集的点云来实现,但是为了提高精度也可以在点云数据采集设备上设置多个传感器来采集点云。
进一步,通过回环方式修复累计误差需要构建回环,因此点云数据采集设备的采集路径会大量重复,降低采集效率。而图6所示的方法不需要构建回环,因此,可以提高采集效率。
再进一步,回环只能通过闭合点匹配消除误差。轨迹中途的累计误差会减小不会消除。而图6所示的方法中,每一帧点云数据中的误差都可以被校正。此外,回环中间的相对累计误差与场景和传感器相关,不易于评估。而利用图6所示的方法,由于点云数据中的对象是校正至对应的参考地图中对应的对象,而参考地图的精度是已知的。因此,校正后的点云数据的精度也是已知的。
以上是结合图6至图8介绍了如何校正图4中的区域1中的点云数据。下面结合图9介绍如何校正图4中的区域2中的点云数据。
图9是根据本申请实施例提供的另一校正点云数据的方法的示意性流程图。
901,确定参考地图与目标地图在区域2的点云数据的集合(可以称为点云数据集2)没有相对应的特征。
如上所述,参考地图在区域2被高架桥遮挡。因此,无法从参考地图中提取区域2中的参考特征。因此,点云数据集2中的特征在区域2中没有对应的参考地图中的特征。
902,根据第一调整参数集,调整点云数据集2中每个点云数据的位置信息。
利用图9所示的方法,可以在无法利用参考地图的情况下,使用根据参考地图确定的调整参数集对点云数据进行校正。上述技术方案也可以对点云数据进行修正。同样的,上述技术方案也可以不需要安装多个传感器,不需要构建回环,并且校正后的点云数据也是已知的。
下面结合图10介绍如何校正图4中的区域3中的点云数据。
图10是根据本申请实施例提供的另一校正点云数据的方法的示意性流程图。
1001,确定参考地图与目标地图在区域3的点云数据的集合(可以称为点云数据集3)没有相对应的特征。
如图4所示,区域1位于东西向的道路(可以称为道路1)。区域3位于高架桥下的 道路(可以称为道路2)。区域2位于道路1和道路2的交叉区域。与区域2类似,参考地图在区域3被高架桥遮挡。因此,无法从参考地图中提取区域3中的参考特征。因此,点云数据集3中的特征在区域3中没有对应的参考地图中的特征。
1002,根据第一调整参数集确定第二调整参数集。
在一些实施例中,第二调整参数集可以与第一调整参数集相同。
在另一些实施例中,可以根据包括第一调整参数集在内的多个调整参数集确定第二调整参数集。如图4所示,区域2只是位于交叉区域中的部分区域。除了区域2以外,交叉区域中还包括其他区域,例如区域6和区域8。利用图6所示的方法还可以确定其他调整参数集。例如,对于区域5和区域7可以分别利用图6所示的方法确定两个调整参数集(可以分别称为调整参数集5和调整参数集7)。根据第一调整参数集确定第二调整参数集可以是:根据第一调整参数集,调整参数集5和调整参数集7确定第二调整参数集。
根据多个调整参数集确定第二调整参数集的方式与根据多个候选调整参数集确定第一调整参数集的方式类似。例如,可以确定多个调整参数集的平均值,中位数,最大值,计算加权求和等。为了简洁,在此就不再赘述。
1003,根据第二调整参数集,调整点云数据集3中每个点云数据的位置信息。
图10所示的方法可以根据已经校正过的点云数据对未经过校正的点云数据进行校正。上述技术方案也可以对点云数据进行修正。同样的,上述技术方案也可以不需要安装多个传感器,不需要构建回环,并且校正后的点云数据也是已知的。
下面结合图11介绍如何校正图4中的区域4中的点云数据。
图11是根据本申请实施例提供的另一校正点云数据的方法的示意性流程图。
1101,确定参考地图与目标地图在区域4的点云数据的集合(可以称为点云数据集4)没有相对应的特征。
在图9所示的方法中参考地图在区域2被高架桥遮挡。因此,无法从参考地图中提取区域2中的参考特征。因此,点云数据集2中的特征在区域2中没有对应的参考地图中的特征。
但是在区域4中,地面上的对象的一部分被树遮挡,另一部分未被遮挡。所以参考地图中可以包括未被遮挡的部分对象。因此,参考地图中的在区域4中的参考特征(可以称为第四参考特征)与点云数据集4中的特征(可以称为第四特征)可能会出现以下几种情况:
情况1,一个或多个第四特征没有对应的第四参考特征。例如,区域4的中的对象被树木遮挡住。因此,参考地图中并没有根据这些对象确定的第四参考特征。但是点云数据采集设备能够采集到对应于这些被树木挡住对象。因此,点云数据4中包括对应于这些对象的第四特征。但是这些第四特征并没有对应的第四参考特征。
情况2,一个或多个第四参考特征没有对应的第四特征。例如,可以从参考地图第四参考特征。但是点云数据采集设备没能够采集对应于第四参考特征的对象的点云。因此,点云数据4中没有包括对应于这些对象的第四特征。
情况3,参考地图在区域4中有T个第四参考特征,虽然能够找到和这T个第四参考特征一一对应的T个第四特征,但是每个第四参考特征与对应的第四特征之间的匹配点数目小于或等于匹配点阈值。匹配点阈值可以是一个预设的数字。例如匹配点阈值可以是大 于或等于5且小于或等于15的正整数。例如,匹配点阈值可以是5,6或者7。如果第四参考特征与对应的第四特征之间的匹配点数目小于该匹配点阈值,则无法确定这两个特征之间的误差,或者,确定出误差与实际误差范围较大。因此,可以也认为这T个第四参考特征没有对应的第四特征。
在一些实施例中,参考地图中的在区域4中的每个第四参考特征与点云数据集4中的每个第四特征都符合上述情况1至情况3中的一种。
在另一些实施例中,参考地图在区域4中的部分第四参考特征与点云数据集4中的部分第四特征符合上述情况1至情况3中的一种。参考地图在区域4中的另一部分第四参考特征与点云数据4中的另一部分第四特征一一对应。但是这些第四参考特征和对应的第四特征之间的误差小于或等于误差阈值。这种第四参考特征和第四特征也可以认为是没有对应。
1102,根据第一调整参数集,调整点云数据集4中每个点云数据的位置信息。
利用图11所示的方法,可以在无法利用参考地图的情况下,使用根据参考地图确定的调整参数集对点云数据进行校正。上述技术方案也可以对点云数据进行修正。同样的,上述技术方案也可以不需要安装多个传感器,不需要构建回环,并且校正后的点云数据也是已知的。
高架桥上的对象由于没有被任何物体遮挡,因此高架桥上的点云数据的校正过程与区域1的校正过程相同,为了简洁,在此就不再赘述。
图12是根据本申请实施例提供的一种校正点云数据的方法的示意性流程图。
1201,确定参考地图在第一区域中的N个第一参考特征和目标地图在该第一区域的第一点云数据集中的N个第一特征,该第一点云数据集为该第一区域的点云数据的集合,该N个第一参考特征和该N个第一特征一一对应,N为正整数。
1202,根据该N个第一参考特征在该参考地图中的位姿和该N个第一特征在该第一点云数据集中的位姿,确定第一调整参数集。
1203,根据该第一调整参数集,调整该第一点云数据集中的每个点云数据的位置信息。
在一些实施例中,该参考地图为数字正射影像图,或者,该参考地图为高可信度点云数据;或者,该参考地图为施工设计图。
在一些实施例中,该方法还包括:根据该第一调整参数集,调整第二点云数据集中每个点云数据的位置信息,该第二点云数据集为该目标地图在第二区域的点云数据的集合。
在一些实施例中,在该根据该第一调整参数集,调整第二点云数据集中每个点云数据的位置信息之前,该方法还包括:确定该参考地图与该第二点云数据集之间没有相对应的特征。
在一些实施例中,该第一区域与该第二区域属于第一道路,该方法还包括:根据该第一调整参数集,确定第二调整参数集;根据该第二调整参数集,调整第三点云数据集中每个点云数据的位置信息,该第三点云数据集为该目标地图在第三区域的点云数据的集合,该第三区域与该第二区域属于第二道路,并且该第二区域位于该第一道路与该第二道路的交叉区域。
在一些实施例中,该根据该N个第一参考特征在该参考地图中的位姿和该N个第一特征在该第一点云数据集中的位姿,确定第一调整参数集,包括:确定第二特征,该第二 特征为该N个第一特征中距离点云数据采集设备最近的一个第一特征;根据第二参考特征在该参考地图中的位姿和该第二特征在该第一点云数据集中的位姿,确定第一调整参数集,该第二参考特征为该N个第一参考特征中与该第二特征对应的第一参考特征。
在一些实施例中,该根据该N个第一参考特征在该参考地图中的位姿和该N个第一特征在该第一点云数据集中的位姿,确定第一调整参数集,包括:根据该N个第一参考特征中第i个第一参考特征在该参考地图中的位姿和该N个第一特征中第i个第一特征在该第一点云数据集中的位姿,确定第i个候选调整参数集,其中该第i个第一参考特征与该第i个第一特征相对应,i=1,…,N;根据N个该候选调整参数集,确定该第一调整参数集。
在一些实施例中,该根据该N个第一参考特征在该参考地图中的位姿和该N个第一特征在该第一点云数据集中的位姿,确定第一调整参数集,包括:根据该N个第一参考特征在该参考地图中的位姿和该N个第一特征在该第一点云数据集中的位姿,确定与参考地图中的对应第一参考特征之间的误差大于误差阈值的K个第一特征,该K为小于或等于N的正整数;根据该K个第一特征在该第一点云数据集中的位姿,以及与该K个第一特征一一对应的K个第一参考特征在该参考地图中的位姿,确定该第一调整参数集。
图12所示方法的具体步骤和有益效果可以参考图6、图9至图11中的描述,为了简洁,在此就不再赘述。
图13是根据本申请实施例提供的一种校正点云数据的装置的示意性结构框图。如图13所示的校正点云数据的装置1300可以用于执行上述各实施例所述的校正点云数据的方法,装置1300包括获取单元1301和处理单元1302。
获取单元1301,用于获取参考地图和目标地图。
处理单元1302,用于确定该参考地图在第一区域中的N个第一参考特征和该目标地图在该第一区域的第一点云数据集中的N个第一特征,该第一点云数据集为该第一区域的点云数据的集合,该N个第一参考特征和该N个第一特征一一对应,N为正整数。
处理单元1302,还用于根据该N个第一参考特征在该参考地图中的位姿和该N个第一特征在该第一点云数据集中的位姿,确定第一调整参数集。
处理单元1302,还用于根据该第一调整参数集,调整该第一点云数据集中的每个点云数据的位置信息。
在一些实施例中,该参考地图为数字正射影像图,或者,该参考地图为高可信度点云数据;或者,该参考地图为施工设计图。
在一些实施例中,处理单元1302,还用于根据该第一调整参数集,调整第二点云数据集中每个点云数据的位置信息,该第二点云数据集为该目标地图在第二区域的点云数据的集合。
在一些实施例中,处理单元1302,还用于在该根据该第一调整参数集,调整第二点云数据集中每个点云数据的位置信息之前,确定该参考地图与该第二点云数据集之间没有相对应的特征。
在一些实施例中,该第一区域与该第二区域属于第一道路,处理单元1302,还用于根据该第一调整参数集,确定第二调整参数集;根据该第二调整参数集,调整第三点云数据集中每个点云数据的位置信息,该第三点云数据集为该目标地图在第三区域的点云数据 的集合,该第三区域与该第二区域属于第二道路,并且该第二区域位于该第一道路与该第二道路的交叉区域。
在一些实施例中,处理单元1302,具体用于确定第二特征,该第二特征为该N个第一特征中距离点云数据采集设备最近的一个第一特征;根据第二参考特征在该参考地图中的位姿和该第二特征在该第一点云数据集中的位姿,确定第一调整参数集,该第二参考特征为该N个第一参考特征中与该第二特征对应的第一参考特征。
在一些实施例中,处理单元1302,具体用于根据该N个第一参考特征中第i个第一参考特征在该参考地图中的位姿和该N个第一特征中第i个第一特征在该第一点云数据集中的位姿,确定第i个候选调整参数集,其中该第i个第一参考特征与该第i个第一特征相对应,i=1,…,N;根据N个该候选调整参数集,确定该第一调整参数集。
在一些实施例中,处理单元1302,具体用于根据该N个第一参考特征在该参考地图中的位姿和该N个第一特征在该第一点云数据集中的位姿,确定与参考地图中的对应第一参考特征之间的误差大于误差阈值的K个第一特征,该K为小于或等于N的正整数;根据该K个第一特征在该第一点云数据集中的位姿,以及与该K个第一特征一一对应的K个第一参考特征在该参考地图中的位姿,确定该第一调整参数集。
在一些实施例中,校正点云数据的装置1300可以是服务器或者车辆,还可以是服务器中或者车辆中部件,所述部件包括芯片,例如***芯片(system on chip,SoC),中央处理器(central processor unit,CPU)实现,专用集成电路(application-specific integrated circuit,ASIC),或可编程逻辑器件(programmable logic device,PLD),上述PLD可以是复杂程序逻辑器件(complex programmable logical device,CPLD),现场可编程门阵列(field-programmable gate array,FPGA),通用阵列逻辑(generic array logic,GAL)或其任意组合。
图13中的各个单元的只一个或多个可以软件、硬件、固件或其结合实现。所述软件或固件包括但不限于计算机程序指令或代码,并可以被硬件处理器所执行。所述硬件包括但不限于各类集成电路,如中央处理单元(CPU)、数字信号处理器(DSP)、现场可编程门阵列(FPGA)或专用集成电路(ASIC)。
根据本申请实施例的校正点云数据的装置1300可对应于执行本申请实施例中描述的方法,并且校正点云数据的装置1300中的各个单元的上述和其它操作和/或功能分别为了实现上述方法的相应流程,为了简洁,在此不再赘述。
图14是根据本申请实施例提供的校正点云数据的装置的结构框图。图14所示的校正点云数据的装置1400可以用于执行上述各实施例所述的校正点云数据的方法,装置1400包括:处理器1401,内存单元1402,存储介质1403。
处理器1401,内存单元1402和存储介质1403可以通过总线1404相通信。
处理器1401是计算设备1400的控制中心,提供执行指令、执行中断动作、提供定时功能和其他功能的排序和处理设施。可选的,处理器1401包括一个或多个中央处理器(CPU)。如图14所示的CPU 0和CPU 1。可选的,计算设备1400包括多个处理器。处理器1401可以是单核(单CPU)处理器,也可以是多核(多CPU)处理器,除非另有说明,否则用于执行任务的处理器或存储器等部件可以实现为临时配置的用于在给定时间执行任务的通用组件或制造用于执行任务的特定组件,如本文所使用的术语“处理器”是指一个 或多个设备或电路。处理器1401还可以是其他通用处理器、例如SOC,ASIC,FPGA或者其他可编程逻辑器件、分立门或者晶体管逻辑器件、分立硬件组件等。通用处理器可以是微处理器或者是任何常规的处理器等。
处理器1401的CPU执行的程序代码可以存储在内存单元1402或存储介质1403中。可选的,程序代码(例如,内核、待调试程序)存储在存储介质1403中,被复制到存储器单元1402中供处理器1401执行。处理器1401可执行至少一个操作***,该操作***可以是LINUX TM、UNIX TM等。处理器1401通过控制其他程序或进程的执行,控制与周边设备的通信,控制数据处理设备资源的使用,从而控制计算设备1400的运行,以此实现上述方法的操作步骤。
总线1404除包括数据总线之外,还可以包括电源总线、控制总线和状态信号总线等。但是为了清楚说明起见,在图中将各种总线都标为总线1404。
可选地,上述校正点云数据的装置1400还包括通信接口(图中未示出),该通信接口用于实现校正点云数据的装置1400与外部器件或设备的通信,例如,校正点云数据的装置1400可以与点云数据采集装置进行通信,用于该点云数据采集装置采集的点云数据。
应理解,根据本申请实施例的校正点云数据的装置1400可对应于本申请实施例中的校正点云数据的装置1300,并且校正点云数据的装置1400中的各个模块的上述和其它操作和/或功能分别为了实现上述方法的相应流程,为了简洁,在此不再赘述。
本申请还提供一种计算机程序产品,当所述计算机程序产品在处理器上运行时,使得校正点云数据的装置执行执行上述方法。
根据本申请实施例提供的方法,本申请还提供一种计算机可读介质,该计算机可读介质包括计算机指令,当所述计算机指令在被处理器运行时,使得校正点云数据的装置执行上述方法。
上述实施例,可以全部或部分地通过软件、硬件、固件或其他任意组合来实现。当使用软件实现时,上述实施例可以全部或部分地以计算机程序产品的形式实现。所述计算机程序产品包括一个或多个计算机指令。在计算机上加载或执行所述计算机程序指令时,全部或部分地产生按照本申请实施例所述的流程或功能。所述计算机可以为通用计算机、专用计算机、计算机网络、或者其他可编程装置。所述计算机指令可以存储在计算机可读存储介质中,或者从一个计算机可读存储介质向另一个计算机可读存储介质传输,例如,所述计算机指令可以从一个网站站点、计算机、服务器或数据中心通过有线(例如同轴电缆、光纤、数字用户线(DSL))或无线(例如红外、无线、微波等)方式向另一个网站站点、计算机、服务器或数据中心进行传输。所述计算机可读存储介质可以是计算机能够存取的任何可用介质或者是包含一个或多个可用介质集合的服务器、数据中心等数据存储设备。所述可用介质可以是磁性介质(例如,软盘、硬盘、磁带)、光介质(例如,DVD)、或者半导体介质。半导体介质可以是固态硬盘(solid state drive,SSD)。
以上所述,仅为本申请的具体实施方式,但本申请的保护范围并不局限于此,任何熟悉本技术领域的技术人员在本申请揭露的技术范围内,可轻易想到变化或替换,都应涵盖在本申请的保护范围之内。因此,本申请的保护范围应以所述权利要求的保护范围为准。

Claims (19)

  1. 一种校正点云数据的方法,其特征在于,包括:
    确定参考地图在第一区域中的N个第一参考特征和目标地图在所述第一区域的第一点云数据集中的N个第一特征,所述第一点云数据集为所述第一区域的点云数据的集合,所述N个第一参考特征和所述N个第一特征一一对应,N为正整数;
    根据所述N个第一参考特征在所述参考地图中的位姿和所述N个第一特征在所述第一点云数据集中的位姿,确定第一调整参数集;
    根据所述第一调整参数集,调整所述第一点云数据集中的每个点云数据的位置信息。
  2. 如权利要求1所述的方法,其特征在于,所述参考地图为数字正射影像图,或者,所述参考地图为高可信度点云数据;或者,所述参考地图为施工设计图。
  3. 如权利要求1或2所述的方法,其特征在于,所述方法还包括:
    根据所述第一调整参数集,调整第二点云数据集中每个点云数据的位置信息,所述第二点云数据集为所述目标地图在第二区域的点云数据的集合。
  4. 如权利要求3所述的方法,其特征在于,在所述根据所述第一调整参数集,调整第二点云数据集中每个点云数据的位置信息之前,所述方法还包括:
    确定所述参考地图与所述第二点云数据集之间没有相对应的特征。
  5. 如权利要求3所述的方法,其特征在于,所述第一区域与所述第二区域属于第一道路,所述方法还包括:
    根据所述第一调整参数集,确定第二调整参数集;
    根据所述第二调整参数集,调整第三点云数据集中每个点云数据的位置信息,所述第三点云数据集为所述目标地图在第三区域的点云数据的集合,所述第三区域与所述第二区域属于第二道路,并且所述第二区域位于所述第一道路与所述第二道路的交叉区域。
  6. 如权利要求1至5中任一项所述的方法,其特征在于,所述根据所述N个第一参考特征在所述参考地图中的位姿和所述N个第一特征在所述第一点云数据集中的位姿,确定第一调整参数集,包括:
    确定第二特征,所述第二特征为所述N个第一特征中距离点云数据采集设备最近的一个第一特征;
    根据第二参考特征在所述参考地图中的位姿和所述第二特征在所述第一点云数据集中的位姿,确定第一调整参数集,所述第二参考特征为所述N个第一参考特征中与所述第二特征对应的第一参考特征。
  7. 如权利要求1至5中任一项所述的方法,其特征在于,所述根据所述N个第一参考特征在所述参考地图中的位姿和所述N个第一特征在所述第一点云数据集中的位姿,确定第一调整参数集,包括:
    根据所述N个第一参考特征中第i个第一参考特征在所述参考地图中的位姿和所述N个第一特征中第i个第一特征在所述第一点云数据集中的位姿,确定第i个候选调整参数集,其中所述第i个第一参考特征与所述第i个第一特征相对应,i=1,…,N;
    根据N个所述候选调整参数集,确定所述第一调整参数集。
  8. 如权利要求1至5中任一项所述的方法,其特征在于,所述根据所述N个第一参 考特征在所述参考地图中的位姿和所述N个第一特征在所述第一点云数据集中的位姿,确定第一调整参数集,包括:
    根据所述N个第一参考特征在所述参考地图中的位姿和所述N个第一特征在所述第一点云数据集中的位姿,确定与参考地图中的对应第一参考特征之间的误差大于误差阈值的K个第一特征,所述K为小于或等于N的正整数;
    根据所述K个第一特征在所述第一点云数据集中的位姿,以及与所述K个第一特征一一对应的K个第一参考特征在所述参考地图中的位姿,确定所述第一调整参数集。
  9. 一种校正点云数据的装置,其特征在于,包括:
    获取单元,用于获取参考地图和目标地图;
    处理单元,用于确定所述参考地图在第一区域中的N个第一参考特征和所述目标地图在所述第一区域的第一点云数据集中的N个第一特征,所述第一点云数据集为所述第一区域的点云数据的集合,所述N个第一参考特征和所述N个第一特征一一对应,N为正整数;
    所述处理单元,还用于根据所述N个第一参考特征在所述参考地图中的位姿和所述N个第一特征在所述第一点云数据集中的位姿,确定第一调整参数集;
    所述处理单元,还用于根据所述第一调整参数集,调整所述第一点云数据集中的每个点云数据的位置信息。
  10. 如权利要求9所述的装置,其特征在于,所述参考地图为数字正射影像图,或者,所述参考地图为高可信度点云数据;或者,所述参考地图为施工设计图。
  11. 如权利要求9或10所述的装置,其特征在于,所述处理单元,还用于根据所述第一调整参数集,调整第二点云数据集中每个点云数据的位置信息,所述第二点云数据集为所述目标地图在第二区域的点云数据的集合。
  12. 如权利要求11所述的装置,其特征在于,所述处理单元,还用于在所述根据所述第一调整参数集,调整第二点云数据集中每个点云数据的位置信息之前,确定所述参考地图与所述第二点云数据集之间没有相对应的特征。
  13. 如权利要求11所述的装置,其特征在于,所述第一区域与所述第二区域属于第一道路,所述处理单元,还用于根据所述第一调整参数集,确定第二调整参数集;根据所述第二调整参数集,调整第三点云数据集中每个点云数据的位置信息,所述第三点云数据集为所述目标地图在第三区域的点云数据的集合,所述第三区域与所述第二区域属于第二道路,并且所述第二区域位于所述第一道路与所述第二道路的交叉区域。
  14. 如权利要求9至13中任一项所述的装置,其特征在于,所述处理单元,具体用于确定第二特征,所述第二特征为所述N个第一特征中距离点云数据采集设备最近的一个第一特征;
    根据第二参考特征在所述参考地图中的位姿和所述第二特征在所述第一点云数据集中的位姿,确定第一调整参数集,所述第二参考特征为所述N个第一参考特征中与所述第二特征对应的第一参考特征。
  15. 如权利要求9至13中任一项所述的装置,其特征在于,所述处理单元,具体用于根据所述N个第一参考特征中第i个第一参考特征在所述参考地图中的位姿和所述N个第一特征中第i个第一特征在所述第一点云数据集中的位姿,确定第i个候选调整参数集, 其中所述第i个第一参考特征与所述第i个第一特征相对应,i=1,…,N;
    根据N个所述候选调整参数集,确定所述第一调整参数集。
  16. 如权利要求9至13中任一项所述的装置,其特征在于,所述处理单元,具体用于根据所述N个第一参考特征在所述参考地图中的位姿和所述N个第一特征在所述第一点云数据集中的位姿,确定与参考地图中的对应第一参考特征之间的误差大于误差阈值的K个第一特征,所述K为小于或等于N的正整数;
    根据所述K个第一特征在所述第一点云数据集中的位姿,以及与所述K个第一特征一一对应的K个第一参考特征在所述参考地图中的位姿,确定所述第一调整参数集。
  17. 一种校正点云数据的装置,其特征在于,其特征在于,包括:处理器,所述处理器用于与存储器耦合,读取并执行所述存储器中的计算机程序指令,以执行如权利要求1-8中任一项所述的方法。
  18. 一种计算机存储介质,其特征在于,包括计算机指令,当所述计算机指令在被处理器运行时,使得校正点云数据的装置执行如权利要求1-8任一项所述的方法。
  19. 一种计算机程序产品,其特征在于,当所述计算机程序产品在处理器上运行时,使得校正点云数据的装置执行如权利要求1-8任一项所述的方法。
PCT/CN2020/086728 2020-04-24 2020-04-24 校正点云数据的方法和相关装置 WO2021212477A1 (zh)

Priority Applications (2)

Application Number Priority Date Filing Date Title
CN202080004997.1A CN112739983B (zh) 2020-04-24 2020-04-24 校正点云数据的方法和相关装置
PCT/CN2020/086728 WO2021212477A1 (zh) 2020-04-24 2020-04-24 校正点云数据的方法和相关装置

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/CN2020/086728 WO2021212477A1 (zh) 2020-04-24 2020-04-24 校正点云数据的方法和相关装置

Publications (1)

Publication Number Publication Date
WO2021212477A1 true WO2021212477A1 (zh) 2021-10-28

Family

ID=75609540

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2020/086728 WO2021212477A1 (zh) 2020-04-24 2020-04-24 校正点云数据的方法和相关装置

Country Status (2)

Country Link
CN (1) CN112739983B (zh)
WO (1) WO2021212477A1 (zh)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113984071A (zh) * 2021-09-29 2022-01-28 云鲸智能(深圳)有限公司 地图匹配方法、装置、机器人和计算机可读存储介质
CN114812447A (zh) * 2022-06-23 2022-07-29 质子汽车科技有限公司 地形面积计量方法及电动车辆

Families Citing this family (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113495281B (zh) * 2021-06-21 2023-08-22 杭州飞步科技有限公司 可移动平台的实时定位方法及装置
CN113503883B (zh) * 2021-06-22 2022-07-19 北京三快在线科技有限公司 采集用于构建地图的数据的方法、存储介质及电子设备

Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20130051658A1 (en) * 2011-08-22 2013-02-28 Samsung Electronics Co., Ltd. Method of separating object in three dimension point cloud
CN109613543A (zh) * 2018-12-06 2019-04-12 深圳前海达闼云端智能科技有限公司 激光点云数据的修正方法、装置、存储介质和电子设备
CN110136182A (zh) * 2019-05-28 2019-08-16 北京百度网讯科技有限公司 激光点云与2d影像的配准方法、装置、设备和介质
CN110148099A (zh) * 2019-05-29 2019-08-20 北京百度网讯科技有限公司 投影关系的修正方法及装置、电子设备、计算机可读介质
CN110197615A (zh) * 2018-02-26 2019-09-03 北京京东尚科信息技术有限公司 用于生成地图的方法及装置
CN110223379A (zh) * 2019-06-10 2019-09-10 于兴虎 基于激光雷达的三维点云重建方法
CN110849374A (zh) * 2019-12-03 2020-02-28 中南大学 地下环境定位方法、装置、设备及存储介质
CN111028161A (zh) * 2019-11-22 2020-04-17 维沃移动通信有限公司 图像校正方法及电子设备

Family Cites Families (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US10270642B2 (en) * 2012-12-05 2019-04-23 Origin Wireless, Inc. Method, apparatus, and system for object tracking and navigation
CN105528588B (zh) * 2015-12-31 2019-04-23 百度在线网络技术(北京)有限公司 一种车道线识别方法及装置
CN108334802B (zh) * 2017-01-20 2022-10-28 腾讯科技(深圳)有限公司 道路特征物的定位方法及装置
CN108036793B (zh) * 2017-12-11 2021-07-23 北京奇虎科技有限公司 基于点云的定位方法、装置及电子设备
CN108734654A (zh) * 2018-05-28 2018-11-02 深圳市易成自动驾驶技术有限公司 绘图与定位方法、***及计算机可读存储介质
CN108765487B (zh) * 2018-06-04 2022-07-22 百度在线网络技术(北京)有限公司 重建三维场景的方法、装置、设备和计算机可读存储介质
CN110634183A (zh) * 2018-06-21 2019-12-31 北京京东尚科信息技术有限公司 地图构建方法、装置和无人设备

Patent Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20130051658A1 (en) * 2011-08-22 2013-02-28 Samsung Electronics Co., Ltd. Method of separating object in three dimension point cloud
CN110197615A (zh) * 2018-02-26 2019-09-03 北京京东尚科信息技术有限公司 用于生成地图的方法及装置
CN109613543A (zh) * 2018-12-06 2019-04-12 深圳前海达闼云端智能科技有限公司 激光点云数据的修正方法、装置、存储介质和电子设备
CN110136182A (zh) * 2019-05-28 2019-08-16 北京百度网讯科技有限公司 激光点云与2d影像的配准方法、装置、设备和介质
CN110148099A (zh) * 2019-05-29 2019-08-20 北京百度网讯科技有限公司 投影关系的修正方法及装置、电子设备、计算机可读介质
CN110223379A (zh) * 2019-06-10 2019-09-10 于兴虎 基于激光雷达的三维点云重建方法
CN111028161A (zh) * 2019-11-22 2020-04-17 维沃移动通信有限公司 图像校正方法及电子设备
CN110849374A (zh) * 2019-12-03 2020-02-28 中南大学 地下环境定位方法、装置、设备及存储介质

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113984071A (zh) * 2021-09-29 2022-01-28 云鲸智能(深圳)有限公司 地图匹配方法、装置、机器人和计算机可读存储介质
CN114812447A (zh) * 2022-06-23 2022-07-29 质子汽车科技有限公司 地形面积计量方法及电动车辆

Also Published As

Publication number Publication date
CN112739983B (zh) 2022-07-12
CN112739983A (zh) 2021-04-30

Similar Documents

Publication Publication Date Title
WO2021212477A1 (zh) 校正点云数据的方法和相关装置
US10598489B2 (en) Visual odometry and pairwise alignment for high definition map creation
US20220028163A1 (en) Computer Vision Systems and Methods for Detecting and Modeling Features of Structures in Images
US10964054B2 (en) Method and device for positioning
US8571265B2 (en) Measurement apparatus, measurement method, and feature identification apparatus
US11625851B2 (en) Geographic object detection apparatus and geographic object detection method
CN111492403A (zh) 用于生成高清晰度地图的激光雷达到相机校准
CN108810473B (zh) 一种在移动平台上实现gps映射摄像机画面坐标的方法及***
JP2013186816A (ja) 動画処理装置、動画処理方法および動画処理用のプログラム
CN112414417B (zh) 自动驾驶地图生成方法、装置、电子设备及可读存储介质
CN103411587A (zh) 定位定姿方法及***
George et al. Towards drone-sourced live video analytics for the construction industry
KR20210034253A (ko) 위치 추정 장치 및 방법
CN114969221A (zh) 一种更新地图的方法及相关设备
KR102050995B1 (ko) 공간좌표의 신뢰성 평가 장치 및 방법
CN108801225A (zh) 一种无人机倾斜影像定位方法、***、介质及设备
CN116958218A (zh) 一种基于标定板角点对齐的点云与图像配准方法及设备
CN116203976A (zh) 变电站室内巡检方法、装置、无人机和存储介质
EP3664038A1 (en) Geospatial surveying tool
KR20200142315A (ko) 도로 네트워크를 갱신하는 방법 및 장치
JP7271413B2 (ja) 物体検出装置および物体検出方法
JP7236377B2 (ja) 物体検出装置および物体検出方法
JP7117408B1 (ja) 位置算出装置、プログラム及び位置算出方法
JP7439969B2 (ja) 地図データ収集装置及び地図データ収集用コンピュータプログラム
JP2024044400A (ja) 情報処理装置、プログラム、システム、及び情報処理方法

Legal Events

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

Ref document number: 20932493

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 20932493

Country of ref document: EP

Kind code of ref document: A1