Determining extrinsic calibration parameters for a sensor

• US 9,297,899 B2
• Filed: 09/27/2012
• Issued: 03/29/2016
• Est. Priority Date: 09/30/2011
• Status: Active Grant

• Current Assignee: Oxford University Innovation Limited (University of Oxford)
• Original Assignee: Chancellor Masters & Scholars of The University of Cambridge (University of Cambridge)
• Inventors: Newman, Paul Michael, Maddern, William Paul, Harrison, Alastair Robin, Sheehan, Mark Christopher
• Primary Examiner(s): RATCLIFFE, LUKE D
• Application Number:
  US14/348,482
• Publication Number:
  US 20140240690A1
• Time in Patent Office: 1,279 Days
• Field of Search: 356 301- 315, 356 401- 41, 356 501- 515, 356 6- 22, 356/28, 356/28.5
• US Class Current: 1/1
• CPC Class Codes: G01S 17/02 Systems using the reflection of electromagnetic waves other than radio waves G01S17/66 takes precedence
• CPC Class Codes: G01S 17/87 Combinations of systems using electromagnetic waves other than radio waves
• CPC Class Codes: G01S 17/875 for determining attitude
• 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: G01S 7/4972 Alignment of sensor

Abstract

A method of determining extrinsic calibration parameters for at least one sensor (102, 104, 106) mounted on transportable apparatus (100). The method includes receiving (202) data representing pose history of the transportable apparatus and receiving (202) sensor data from at least one sensor mounted on transportable apparatus. The method generates (204) at least one point cloud data using the sensor data received from the at least one sensor, each point in a said point cloud having a point covariance derived from the pose history data. The method then maximizes (206) a value of a quality function for the at least one point cloud, and uses (208) the maximized quality function to determine extrinsic calibration parameters for the at least one sensor.

Claims

1. A method of determining extrinsic calibration parameters for at least one sensor mounted on a transportable apparatus, the method including:

receiving, by a computer processor, data representing pose history of the transportable apparatus with respect to a global frame;

receiving, by the computer processor, sensor data from at least one sensor mounted on the transportable apparatus;

generating, by the computer processor, at least one point cloud datum using the sensor data received from the at least one sensor, each point in a said point cloud having a corresponding point covariance derived from the pose history data;

maximizing, by the computer processor, a value of a quality function for the at least one point cloud; and

determining, by the computer processor, extrinsic calibration parameters for the at least one sensor using the maximized quality function.

2. A method according to claim 1, wherein the value of the quality function depends on pairwise distances between said points in a said point cloud (X^) and associated said point covariance (Σ

) of each said point derived from the pose history data.

3. A method according to claim 1, wherein the quality function (E) is expressed as E(Θ

|Z, Y), where Z represents the data received from one said sensor, the sensor data representing a set of measurements over a period of time, and where Y represents a set of poses of the transportable apparatus from the pose history over the period of time, and where Θ

provides a most likely estimate for the extrinsic calibration parameters for the sensor.

4. A method according to claim 1, wherein the quality function comprises an entropy-based point cloud quality metric.

5. A method according to claim 4, wherein the quality function is based on Renyi Quadratic Entropy.

6. A method according to claim 1, wherein the maximizing of the value of the quality function is performed using Newton'"'"'s method.

7. A method according to claim 1, wherein the method is used to:

generate a first point cloud based on sensor data received from a 3D LIDAR device;

generate a second point cloud based on sensor data received from a 2D LIDAR device, and

use a Kernelized Renyi Distance function on the first and the second point clouds to determine the extrinsic calibration parameters for the 2D LIDAR device.

8. A method according to claim 1, where, in the step of maximizing the value of a quality function for the point cloud, the method evaluates bi-direction data only once to reduce computation time.

9. A method according to claim 1, where, in the step of maximizing the value of a quality function for the point cloud, the method evaluates an entropy contribution between a first said point of the point cloud and neighbouring said points in the point cloud to reduce computation time.

10. A method according to claim 1, wherein the point cloud data is stored in a tree data structure for processing.

11. A method according to claim 10, wherein the tree data structure comprises a k-d tree.

12. Transportable apparatus including at least one sensor and a processor configured to execute a method according to claim 1.

13. A vehicle including transportable apparatus according to claim 12.

14. A method of calibrating at least one sensor mounted on a transportable apparatus, the method including:

receiving, by a computer processor, data representing pose history of the transportable apparatus, the pose history data representing at least two different locations of the transportable apparatus within a global frame;

receiving, by the computer processor, sensor data from at least one sensor mounted on the transportable apparatus, the sensor data representing coordinates of an object in a sensor frame that is different than the global frame, the sensor data being obtained while the transportable apparatus is located at each of the at least two different locations of the transportable apparatus in the pose history data;

estimating, by the computer processor, coordinates of the object in the global frame based on the sensor data received from the at least one sensor;

generating, by the computer processor, point cloud data by combining the estimated coordinates of the object in the global frame with corresponding point covariences derived from the pose history data;

maximizing, by the computer processor, a value of a quality function for the point cloud;

determining, by the computer processor, extrinsic calibration parameters for the at least one sensor using the maximized quality function, and

calibrating the at least one sensor using the determined extrinsic calibration parameters.

15. A non-transient computer program product comprising computer code that when executed by one or more processors causes a process for determining extrinsic calibration parameters for at least one sensor mounted on a transportable apparatus, the process comprising:

receiving data representing pose history of the transportable apparatus with respect to a global frame;

receiving sensor data from at least one sensor mounted on the transportable apparatus;

generating at least one point cloud datum using the sensor data received from the at least one sensor, each point in a said point cloud having a corresponding point covariance derived from the pose history data;

maximizing a value of a quality function for the at least one point cloud; and

using the maximized quality function to determine extrinsic calibration parameters for the at least one sensor.

16. A computer program product according to claim 15, wherein the value of the quality function depends on pairwise distances between said points in a said point cloud (X^) and associated said point covariance (Σ

) of each said point derived from the pose history data.

17. A computer program product according to claim 15, wherein the quality function (E) is expressed as E(Θ

|Z, Y), where Z represents the data received from one said sensor, the sensor data representing a set of measurements over a period of time, and where Y represents a set of poses of the transportable apparatus from the pose history over the period of time, and where Θ

provides a most likely estimate for the extrinsic calibration parameters for the sensor.

18. A computer program product according to claim 15, wherein the quality function comprises an entropy-based point cloud quality metric and the maximizing of the value of the quality function is performed using Newton'"'"'s method.

19. A computer program product according to claim 15, wherein the process is used to:

generate a first point cloud based on sensor data received from a 3D LIDAR device;

generate a second point cloud based on sensor data received from a 2D LIDAR device, and

use a Kernelized Renyi Distance function on the first and the second point clouds to determine the extrinsic calibration parameters for the 2D LIDAR device.

20. A computer program product according to claim 15, wherein the point cloud data is stored in a tree data structure for processing, and the tree data structure comprises a k-d tree.