Automatic localization geometry detection

• US 10,066,946 B2
• Filed: 08/26/2016
• Issued: 09/04/2018
• Est. Priority Date: 08/26/2016
• Status: Active Grant

• Current Assignee: HERE Global B.V. (HERE Holding Corporation)
• Original Assignee: HERE Global B.V. (HERE Holding Corporation)
• Inventors: Doria, David, Chen, Xin
• Primary Examiner(s): Moyer, Dale
• Application Number:
  US15/248,714
• Publication Number:
  US 20180058861A1
• Time in Patent Office: 739 Days
• CPC Class Codes: G01C 21/3837 Data obtained from a single source
• CPC Class Codes: G01C 21/3867 Geometry of map features, e.g. shape points, polygons or for simplified maps
• CPC Class Codes: G01S 17/89 for mapping or imaging
• CPC Class Codes: G01S 7/4808 Evaluating distance, position or velocity data
• CPC Class Codes: G06T 17/05 Geographic models
• CPC Class Codes: G06T 2215/16 Using real world measurements to influence rendering
• CPC Class Codes: G06T 7/00 Image analysis

Abstract

Embodiments include apparatus and methods for generating a localization geometry or occupancy grid for a geographic location. Point cloud that describes a vicinity of a pathway is collected by a distance sensor and describing a vicinity of the pathway. The point cloud data is reduced or filtered to a predetermined volume with respect to the roadway. The remaining point cloud data is projected onto a two-dimensional plane including at least one pixel formation. A volumetric grid is defined according to the at least one pixel formation, and a voxel occupancy for each of a voxels forming the volumetric grid is determined. The arrangement of the voxel occupancies or a sequence of data describing the voxel occupancies is a localization geometry that describes the geographic location of the pathway.

Claims

1. A method for automatic generation of a localization geometry, the method comprising:

receiving point cloud data collected by a distance sensor and describing a vicinity of a pathway;

reducing, by a processor, the point cloud data to a predetermined volume with respect to the pathway;

projecting, by the processor, the point cloud data to a two-dimensional plane including at least one pixel formation;

defining, by the processor, a volumetric grid according to the at least one pixel formation;

determining, by the processor, a voxel occupancy for each of a plurality of voxels forming the volumetric grid; and

generating, by the processor, the localization geometry according to the voxel occupancy.

2. The method of claim 1, wherein reducing the point cloud data to the predetermined volume includes:

determining a first horizontal border of the pathway;

determining a second horizontal border of the pathway; and

filtering data of the point cloud data to remove data between the first horizontal border and the second horizontal border.

3. The method of claim 1, wherein reducing the point cloud data to the predetermined volume includes:

determining a first vertical border in a plane parallel to the pathway;

determining a second vertical border in a plane parallel to the pathway; and

filtering data of the point cloud data to remove data between the first vertical border and the second vertical border.

4. The method of claim 3, wherein the first vertical border corresponds to a height of a collection vehicle for the point cloud data.

5. The method of claim 1, wherein reducing the point cloud data to the predetermined volume includes:

determining an outer border for the pathway; and

filtering data of the point cloud data to remove data farther from the pathway than the outer border.

6. The method of claim 1, further comprising:

defining a plurality of chunks of the pathway, each chunk having a predetermined length, wherein at least one object described in the point cloud data extends across multiple chunks.

7. The method of claim 1, wherein the two-dimensional plane includes a two-dimensional grid having cells of a predetermined area, the method further comprising:

filtering data in cells having less than a predetermined number of data points.

8. The method of claim 1, further comprising:

identifying the at least one pixel formation from the two-dimensional plane; and

generating the volumetric grid from the at least one pixel formation.

9. The method of claim 1, further comprising:

defining a primary direction in the volumetric grid;

defining a secondary direction in the volumetric grid;

defining a tertiary direction in the volumetric grid; and

storing occupancy data for the volumetric data according to a sequence of the primary direction, the secondary direction, and the tertiary direction.

10. The method of claim 1, further comprising:

transforming the volumetric grid to the localization geometry according to a transformation matrix or a transformation quaternion.

11. The method of claim 10, wherein the volumetric grid is in local coordinates and the localization geometry is in geodetic coordinates.

12. The method of claim 1, wherein subsequent point cloud data is compared to the localization geometry in order to determine a geographic position of a mobile device.

13. An apparatus for automatic generation of a localization geometry of a roadway, the apparatus comprising:

a communication interface configured to receive point cloud data collected by a distance sensor and describing a vicinity of a roadway; and

a controller configured to reduce the point cloud data to a predetermined volume with respect to the roadway, project the point cloud data to a two-dimensional plane including at least one pixel formation, and determine a voxel occupancy for each of a plurality of voxels corresponding to the at least one pixel formation,

wherein the voxel occupancy for a grid defines the localization geometry of the roadway.

14. The apparatus of claim 13, wherein the controller reduces the point cloud data according to removal of data between borders associated with the roadway.

15. The apparatus of claim 13, wherein the controller reduces the point cloud data according to removal of data between a first vertical border in a plane parallel to the roadway and a second vertical border in a plane parallel to the roadway.

16. The apparatus of claim 13, wherein the controller reduces the point cloud data according to removal of data farther from the roadway than an outer border for the roadway.

17. The apparatus of claim 13, wherein the controller defines a plurality of chunks of the roadway, each chunk having a predetermined length, wherein at least one object described in the point cloud data extends across multiple chunks.

18. The apparatus of claim 13, further comprising:

a memory comprising occupancy data for the grid arranged in an order that spans multiple dimensions of the grid.

19. The apparatus of claim 13, wherein the controller transforms the grid to the localization geometry according to a transformation matrix or a transformation quaternion.

20. A system for automatic generation of a localization geometry of a roadway, the system comprising:

a distance sensor configured to receive point cloud data collected by a distance sensor and describing a vicinity of a roadway; and

a controller configured to reduce the point cloud data to a predetermined volume with respect to the roadway and project the point cloud data to a two-dimensional plane including at least one pixel formation,

wherein the localization geometry includes a volumetric grid is populated according to a voxel occupancy value for at least one pixel formation.