Method and system for stereo based vehicle pose estimation

1. A method implemented on a computer having at least one processor, a storage, and a communication platform for estimating a pose of an object in autonomous driving, comprising:

obtaining stereo images of a scene from at least two cameras;

detecting an object present in the scene based on the stereo images;

identifying a plurality of feature points of the object;

determining at least depth information associated with the plurality of feature points;

obtaining a configuration of the plurality of feature points;

estimating an orientation of the object based on the configuration of the plurality of feature points and the depth information; and

determining a pose of the object based on the orientation and additional features of the object.

2. The method of claim 1 , wherein when the object corresponds to a vehicle, the plurality of feature points include:

a left most point of the object as appearing in the stereo images;

a right most point of the object as appearing in the stereo images; and

a closest point on the object that has a shortest distance from an ego vehicle having the at least two cameras installed thereon.

3. The method of claim 2 , wherein the step of the estimating the orientation comprises:

in response to the left most, right most, and closest point forming a straight line,

measuring a distance between the left most and the right most points,

determining a first orientation as the orientation of the object if the distance satisfies a criterion, and

determining a second orientation as the orientation of the object if the distance does not satisfy the criterion; and

in response to the left most, right most, and the closest point not forming a straight line, determining a third orientation as the orientation of the object.

4. The method of claim 3 , wherein the criterion is dynamically determined in accordance with the depth information.

5. The method of claim 3 , wherein the orientation of the object represents two alternative poses of the object and the pose of the object is to be determined based on the additional features.

6. The method of claim 5 , wherein the step of determining the pose comprises:

determining the additional features to be detected from the object based on the orientation;

detecting the additional features from at least one of the stereo images; and

selecting one of the alternative poses of the object as the pose of the object based on the additional features.

7. The method of claim 6 , wherein the additional features include at least one of:

visual features observable from at least one of a frontal view, a rear view, and a side view of the object;

motion features associated with the object; and

a combination thereof.

8. The method of claim 6 , wherein the step of determining the additional features comprises:

in response to the orientation corresponding to the first orientation, detecting visual features observable between the left most and right most points representing one of the front or the rear view of the object;

in response to the orientation corresponding to the second orientation, detecting visual features observable between the left most and right most points representing the side view of the object; and

in response to the orientation corresponding to the third orientation, detecting visual features observable both between the left most point and the closest point and between the closest point and the right most point, representing at least one of the side view and one of the frontal view and the back view of the object.

9. The method of claim 7 , wherein the motion related features include a directionality of movement of the object, wherein the directionality includes one of towards, away, left, and right.

10. A machine readable non-transitory medium having information stored thereon for estimating a pose of an object in autonomous driving, wherein the information, when read by the machine, causes the machine to perform:

obtaining stereo images of a scene from at least two cameras;

detecting an object present in the scene based on the stereo images;

identifying a plurality of feature points of the object;

determining at least depth information associated with the plurality of feature points;

obtaining a configuration of the plurality of feature points;

estimating an orientation of the object based on the configuration of the plurality of feature points and the depth information; and

determining a pose of the object based on the orientation and additional features of the object.

11. The medium of claim 10 , wherein when the object corresponds to a vehicle, the plurality of feature points include:

a left most point of the object as appearing in the stereo images;

a right most point of the object as appearing in the stereo images; and

a closest point on the object that has a shortest distance from an ego vehicle having the at least two cameras installed thereon.

12. The medium of claim 11 , wherein the step of the estimating the orientation comprises:

in response to the left most, right most, and closest point forming a straight line,

measuring a distance between the left most and the right most points,

determining a first orientation as the orientation of the object if the distance satisfies a criterion, and

determining a second orientation as the orientation of the object if the distance does not satisfy the criterion; and

in response to the left most, right most, and the closest point not forming a straight line, determining a third orientation as the orientation of the object.

13. The medium of claim 12 , wherein the criterion is dynamically determined in accordance with the depth information.

14. The medium of claim 12 , wherein the orientation of the object represents two alternative poses of the object and the pose of the object is to be determined based on the additional features.

15. The medium of claim 14 , wherein the step of determining the pose comprises:

determining the additional features to be detected from the object based on the orientation;

detecting the additional features from at least one of the stereo images; and

selecting one of the alternative poses of the object as the pose of the object based on the additional features.

16. The medium of claim 15 , wherein the additional features include at least one of:

visual features observable from at least one of a frontal view, a rare rear view, and a side view of the object;

motion features associated with the object; and

a combination thereof.

17. The medium of claim 15 , wherein the step of determining the additional features comprises:

in response to the orientation corresponding to the first orientation, detecting visual features observable between the left most and right most points representing one of the front or the rear view of the object;

in response to the orientation corresponding to the second orientation, detecting visual features observable between the left most and right most points representing the side view of the object; and

in response to the orientation corresponding to the third orientation, detecting visual features observable both between the left most point and the closest point and between the closest point and the right most point, representing at least one of the side view and one of the frontal view and the back view of the object.

18. The medium of claim 16 , wherein the motion related features include a directionality of movement of the object, wherein the directionality includes one of towards, away, left, and right.

19. A system for estimating a pose of an object in autonomous driving, comprising:

a stereo image pre-processor configured for obtaining stereo images of a scene from at least two cameras;

an object detection unit configured for detecting an object present in the scene based on the stereo images;

a critical feature point extractor configured for

identifying a plurality of feature points of the object, and

obtaining a configuration of the plurality of feature points;

an object depth estimator configured for determining at least depth information associated with the plurality of feature points;

an object orientation estimator configured for estimating an orientation of the object based on the configuration of the plurality of feature points and the depth information; and

a pose ambiguity resolution unit configured for determining a pose of the object based on the orientation and additional features of the object.

20. The system of claim 19 , wherein when the object corresponds to a vehicle, the plurality of feature points include:

a left most point of the object as appearing in the stereo images;

a right most point of the object as appearing in the stereo images; and

a closest point on the object that has a shortest distance from an ego vehicle having the at least two cameras installed thereon.

21. The system of claim 20 , wherein the object orientation estimator is configured to perform:

in response to the left most, right most, and closest point forming a straight line,

measuring a distance between the left most and the right most points,

determining a first orientation as the orientation of the object if the distance satisfies a criterion, and

determining a second orientation as the orientation of the object if the distance does not satisfy the criterion; and

in response to the left most, right most, and the closest point not forming a straight line, determining a third orientation as the orientation of the object.

22. The system of claim 21 , wherein the criterion is dynamically determined in accordance with the depth information.

23. The system of claim 21 , wherein the orientation of the object represents two alternative poses of the object and the pose of the object is to be determined based on the additional features.

24. The system of claim 23 , wherein the step of determining the pose comprises:

determining the additional features to be detected from the object based on the orientation;

detecting the additional features from at least one of the stereo images; and

selecting one of the alternative poses of the object as the pose of the object based on the additional features.

25. The system of claim 24 , wherein the additional features include at least one of:

visual features observable from at least one of a frontal view, a rare rear view, and a side view of the object;

motion features associated with the object; and

a combination thereof.

26. The system of claim 24 , wherein the step of determining the additional features comprises:

in response to the orientation corresponding to the first orientation, detecting visual features observable between the left most and right most points representing one of the front or the rear view of the object;

in response to the orientation corresponding to the second orientation, detecting visual features observable between the left most and right most points representing the side view of the object; and

in response to the orientation corresponding to the third orientation, detecting visual features observable both between the left most point and the closest point and between the closest point and the right most point, representing at least one of the side view and one of the frontal view and the back view of the object.

27. The system of claim 25 , wherein the motion related features include a directionality of movement of the object, wherein the directionality includes one of towards, away, left, and right.