System and method for tracking objects

• US 8,818,702 B2
• Filed: 11/09/2010
• Issued: 08/26/2014
• Est. Priority Date: 11/09/2010
• Status: Active Grant

• Current Assignee: GM Global Technology Operations LLC (General Motors Company)
• Original Assignee: GM Global Technology Operations LLC (General Motors Company)
• Inventors: Zeng, Shuqing, Li, Yuanhong
• Primary Examiner(s): Tran, Khoi
• Assistant Examiner(s): MOYER, DALE S
• Application Number:
  US12/942,456
• Publication Number:
  US 20120116662A1
• Time in Patent Office: 1,386 Days
• Field of Search: 340435-438, 340/903, 340/988, 382/103, 382106-107, 382/154, 701300-302
• US Class Current: 701/301
• CPC Class Codes: G05D 1/024 in combination with a laser lasers per se H01S
• CPC Class Codes: G05D 1/027 comprising intertial navigation means, e.g. azimuth detector inertial navigation G01C21/16; inertial navigation combined with non-inertial navigation instruments G01C21/165

Abstract

A method for tracking an object, by generating kinematic models corresponding to the object using a computerized object-tracking system, is disclosed. The method includes receiving, by a data collection module, scan data associated with an object, and generating by the module, using the scan data, a new frame F, associated with a new time t+dt and including new points X. A clustering module identifies a new group G of new points X_i of the new frame F. A data association module associates the new group G with the object based on previous information associated with the object. A tracking module determines a new kinematic model M corresponding to the object based on the new group G.

Claims

1. A non-transitory computer-readable storage device comprising instructions that, when executed by a processor, cause the processor to perform operations, for tracking from a vehicle, an object by generating kinematic models corresponding to the object, comprising:

receiving scan data associated with the object;

generating, using the scan data, a new frame F, associated with a new time t+dt and including new points X;

identifying a new group G of new points X_i, of the new points X of the new frame F, where i=1, . . . , N, with N being a positive integer corresponding to a number of the new points;

associating the new group G with the object based on previous information associated with the object; and

determining, based on the new group G, a new kinematic model M corresponding to the object;

wherein;

determining, based on the new group G, the new kinematic model M corresponding to the object includes generating a predicted group G_p;

determining the new kinematic model M includes iteratively matching the predicted group G_p to the new group G;

iteratively matching the predicted group G_p to the new group G includes determining an initial translation adjustment vector Δ

T and iteratively adjusting a translation vector T using the initial translation adjustment vector Δ

T; and

determining the initial translation adjustment vector Δ

T is performed using at least one of;

a localized point-matching technique; and

a maximum-likelihood technique.

2. The computer-readable storage device of claim 1, wherein:

the operations further comprise processing the new frame F prior to identifying the new group G of new points X_i; and

processing the new frame F includes filtering the new frame F to represent the new frame F by vehicle-reference coordinates.

3. The computer-readable storage device of claim 1, wherein:

the operations further comprise processing the new frame F prior to identifying the new group G of new points X_i; and

processing, the new frame F includes filtering the new frame F to represent the new frame F by ground-reference coordinates.

4. The computer-readable storage device of claim 1, wherein:

generating the predicted group G_p is performed as a function of a previous kinematic model representing the object, the new group G, and a translation vector T;

the predicted group G_p corresponds to a future time t+δ

t, and is determined according to;

G<sub>t+δ

t</sub>^p=G_t+T_t|<sub>δ

t</sub>;

G^P represents the predicted group G_p;

G^P<sub>t+δ

t</sub> represents the predicted group in association with the future time t+δ

t;

T_t|<sub>δ

t</sub> represents the translation vector T;

the translation vector T_t|<sub>δ

t</sub> includes an x-direction translation component dx and a y-direction translation component dy;

the translation components are determined as a function of a velocity V_t and a velocity direction θ

t of the previous kinematic model representing the object;

the translation components dx, dy are related to the velocity V_t and the velocity direction θ

t of the previous kinematic model representing the object according to;

5. The computer-readable storage device of claim 1, wherein the localized point-matching technique includes identifying neighboring prediction points X_j, the neighboring prediction points X_j being points of the predicted group G_p located within a specified distance of each of the new points X_i of the new group G.

6. The computer-readable storage device of claim 1, wherein the maximum-likelihood technique includes determining the translation adjustment vector Δ

T so as to maximize a likelihood that the predicted group G_p matches the new group G, including matching each of the new points X_i to a neighboring predicted point X_j of the predicted group G_p, such that an aggregate of distances between each new point X_i and neighboring prediction points X_j is minimized.

7. The computer-readable storage device of claim 6, wherein the maximum-likelihood technique further includes adding the translation adjustment vector Δ

T to the initial translation vector T to produce an iterated translation vector T′

.

8. The method of claim 7, wherein the maximum-likelihood technique includes determining whether the iterated translation vector T′

has converged so as to match the predicted group G_p to the new group G.

9. The computer-readable storage device of claim 1, wherein:

if the magnitude of the translation adjustment vector Δ

T is above a predetermined threshold, the iterated translation vector T′

is further iterated; and

if the magnitude of the translation adjustment vector Δ

T is within the predetermined threshold, the iterated translation vector T′

is used to determine parameters of the new kinematic model M, including using components di′

, dy′

of the iterated translation vector T′

to determine a velocity V_t and an angle of velocity θ

t for the new kinematic model M according to relationships represented as;

10. A non-transitory computer-readable storage device comprising instructions that, when executed by a processor, cause the processor to perform operations, for tracking from a vehicle, an object by generating kinematic models corresponding to the object, comprising:

receiving scan data associated with the object;

generating, using the scan data, a new frame F, associated with a new time t+dt and including new points X;

identifying a new group G of new points X_i, of the new points X of the new frame F, where i=1, . . . , N, with N being a positive integer corresponding to a number of the new points;

associating the new group G with the object based on previous information associated with the object; and

determining, based on the new group G, a new kinematic model M corresponding to the object;

wherein;

determining, based on the new group G, the new kinematic model M corresponding to the object includes generating a predicted group G_p;

determining the new kinematic model M includes iteratively matching the predicted group G_p to the new group G;

iteratively matching the predicted group G_p to the new group G includes determining an initial translation adjustment vector Δ

T, and iteratively adjusting a translation vector T using the initial translation adjustment vector Δ

T;

if the magnitude of the translation adjustment vector Δ

T is above a predetermined threshold the iterated translation vector T′

is further iterated; and

if the magnitude of the translation adjustment vector Δ

T is within the predetermined threshold, the iterated translation vector T′

is used to determine parameters of the new kinematic model M, including using components dx′

, dy′

of the iterated translation vector T′

to determine a velocity V_t and an angle of velocity θ

t for the new kinematic model M according to relationships represented as;

11. The computer-readable storage device of claim 10, wherein:

the operations further comprise processing the new frame F prior to identifying the new group G of new points X_i of the new frame F; and

processing the new frame F includes filtering the new frame F to represent the new frame F by vehicle-reference coordinates.

12. The computer-readable storage device of claim 10, wherein:

the operations further comprise processing the new frame F prior to identifying the new group G of new points X_i of the new frame F; and

processing the new frame F includes filtering the new frame F to represent the new frame F by ground-reference coordinates.

13. The computer-readable storage device of claim 10, wherein:

generating the predicted group G_p is performed as a function of a previous kinematic model representing the object, the new group G, and a translation vector T;

the predicted group G_p corresponds to a future time t+δ

t, and is determined according to;

G<sub>t+δ

t</sub>^p=G_t+T_t|<sub>δ

t</sub>;

G^P represents the predicted group G_p;

G^P<sub>t+δ

t</sub> represents the predicted group in association with the future time t+δ

t;

T_t|<sub>δ

t</sub> represents the translation vector T;

the translation vector T_t|<sub>δ

t</sub> includes an x-direction translation component dx and a y-direction translation component dy;

the translation components are determined as a function of a velocity V_t and a velocity direction θ

t of the previous kinematic model representing the object;

the translation components dx, dy are related to the velocity V_t and the velocity direction θ

t of the previous kinematic model representing the object according to;

14. The computer-readable storage device of claim 10, wherein:

determining the initial translation adjustment vector Δ

T is performed using a localized point-matching technique; and

the localized point-matching technique includes identifying neighboring prediction points X_j, the neighboring prediction points X_j being points of the predicted group G_p located within a specified distance of each of the new points X_i of the new group G.

15. The computer-readable storage device of claim 10, wherein:

determining the initial translation adjustment vector Δ

T is performed using a maximum-likelihood technique; and

the maximum-likelihood technique includes determining the translation adjustment vector Δ

T so as to maximize a likelihood that the predicted group G_p matches the new group G, including matching each of the new points X_i to a neighboring predicted point X_j of the predicted group G_p, such that an aggregate of distances between each new point X_i and neighboring prediction points X_j is minimized.

16. The computer-readable storage device of claim 15, wherein the maximum-likelihood technique further includes adding the translation adjustment vector Δ

T to the initial translation vector T to produce an iterated translation vector T′

.

17. The method of claim 16, wherein the maximum-likelihood technique includes determining whether the iterated translation vector T′

has converged so as to match the predicted group G_p to the new group G.

18. A non-transitory computer-readable storage device comprising instructions that, when executed by a processor, cause the processor to perform operations, for tracking from a vehicle, an object by generating kinematic models corresponding to the object, object, comprising:

receiving scan data associated with the object;

generating, using the scan data, a new frame F, associated with a new time t+dt and including new points X;

identifying a new group G of new points X_i, of the new points X of the new frame F, where i=1, . . . , N, with N being a positive integer corresponding to a number of the new points;

associating the new group G with the object based on previous information associated with the object; and

determining, based on the new group G, a new kinematic model M corresponding to the object;

wherein;

determining, based on the new group G, the new kinematic model M corresponding to the object includes generating a predicted group G_p;

generating the predicted group G_p is performed as a function of a previous kinematic model representing the object, the new group G, and a translation vector T;

the predicted group G_p corresponds to a future time t+δ

t, and is determined according to;

G<sub>t+δ

t</sub>^p=G_t+T_t|<sub>δ

t</sub>;

G^P represents the predicted group G_p;

G^P<sub>t+δ

t</sub> represents the predicted group in association with the future time t+δ

t;

T_t|<sub>δ

t</sub> represents the translation vector T;

the translation vector T_t|<sub>δ

t</sub> includes an x-direction translation component dx and a y-direction translation component dy;

the translation components are determined as a function of a velocity V_t and a velocity direction θ

t of the previous kinematic model representing the object;

the translation components dx, dy are related to the velocity V_t and the velocity direction θ

t of the previous kinematic model representing the object according to;

19. The computer-readable storage device of claim 18, wherein:

the operations further comprise processing the new frame F prior to identifying the new group G of new points X_i of the new frame F; and

processing the new frame F includes the filtering the new frame F to represent the new frame F by vehicle-reference coordinates.

20. The computer-readable storage device of claim 18, wherein:

the operations further comprise processing the new frame F prior to identifying the new group G of new points X_i of the new frame F; and

processing the new frame F includes the filtering the new frame F to represent the new frame F by ground-reference coordinates.