Granted Patent B2

US 8,644,557 · App. 13/296,989 · Granted Feb 4, 2014

Abstract

Claims (59)

1. An apparatus of estimating a position of a moving vehicle, the apparatus comprising:

a feature point matching unit which generates vectors connecting feature points of a previous image frame and feature points of a current image frame, corresponding to the feature points of the previous image frame, and determines spatial correlations between the feature points of the current image frame;

a clustering unit which configures at least one motion cluster by grouping at least one vector among the vectors based on the spatial correlations; and

a noise removal unit which removes noise from each motion cluster of the at least one motion cluster,

wherein the position of the moving vehicle is estimated based on the at least one motion cluster.

2. The apparatus of claim 1 , further comprising a position estimation unit which estimates the position of the moving vehicle based on each motion cluster from which the noise has been removed,

wherein the current image frame and the previous image frame are obtained by one single image obtaining unit, installed on the moving vehicle, at two different time points.

3. The apparatus of claim 2 , further comprising a three-dimensional (3D) coordinate estimation unit which determines 3D coordinates of the image obtaining unit based on each motion cluster from which the noise has been removed,

wherein the position estimation unit estimates the position of the moving vehicle based on the 3D coordinates of the image obtaining unit.

4. The apparatus of claim 1 , further comprising;

a three-dimensional (3D) coordinate estimation unit which, based on each motion cluster from which the noise has been removed, determines 3D coordinates of one single image obtaining unit, installed on the moving vehicle, which obtains the current image frame and the previous image frame at two different time points; and

a position estimation unit which estimates the position of the moving vehicle based on the 3D coordinates of the image obtaining unit.

5. The apparatus of claim 4 , wherein the clustering unit configures a plurality of motion clusters by grouping at least one vector among the vectors based on the spatial correlations,

wherein the 3D coordinate estimation unit determines a 3D coordinate of the image obtaining unit for each of the plurality of motion clusters from which the noise has been removed, and

wherein the position estimation unit estimates the position of the moving vehicle based on the 3D coordinate of the image obtaining unit for each of the plurality of motion clusters from which the noise has been removed.

6. The apparatus of claim 1 , wherein the noise of each motion cluster comprises a vector which is inconsistent with the other vectors in the each motion cluster.

7. The apparatus of claim 1 , wherein the feature point matching unit comprises:

a feature point extraction unit which tracks the feature points of the current image, and generates the vectors connecting the corresponding feature points between the two image frames by detecting optical flow of each pixel; and

a correlation determination unit which determines the spatial correlations.

8. The apparatus of claim 7 , wherein the feature point extraction unit extracts the feature points using a Kanade-Lucas-Tomasi (KLT) tracker and a Harris corner detector,

wherein the correlation determination unit triangularizes the feature points on a two-dimensional (2D) plane using a Delaunay triangulation technique,

wherein the clustering unit configures the at least one motion cluster by applying a Markov random field (MRF) optimization algorithm using graph-cut in a feature space, and

wherein the noise removal unit removes the noise by applying random sample consensus (RANSAC) to each motion cluster of the at least one motion cluster.

9. The apparatus of claim 1 , further comprising;

a calibration unit which determines a calibration matrix using internal parameters of an image obtaining unit, installed on the moving vehicle, which obtains the current image frame and the previous image frame at two different time points;

a three-dimensional (3D) coordinate estimation unit which determines an essential matrix based on the calibration matrix and a fundamental matrix determined by the noise removal unit with respect to each motion cluster, and determines 3D coordinates of the image obtaining unit based on the essential matrix; and

a position estimation unit which estimates the position of the moving vehicle based on the 3D coordinates of the image obtaining unit.

10. The apparatus of claim 8 , wherein the clustering unit configures a plurality of motion clusters by grouping at least one vector among the vectors based on the spatial correlations,

wherein the 3D coordinate estimation unit determines a 3D coordinate of the image obtaining unit for each of the plurality of motion clusters from which the noise has been removed, and

wherein the position estimation unit estimates the position of the moving vehicle based on the 3D coordinate of the image obtaining unit for each of the plurality of motion clusters from which the noise has been removed.

11. A method of estimating a position of a moving vehicle, the method comprising:

generating vectors connecting feature points of a previous image frame and feature points of a current image frame corresponding to the feature points of the previous image frame;

determining spatial correlations between the feature points of the current image frame;

configuring at least one motion cluster by grouping the vectors based on the spatial correlations; and

removing noise from each motion cluster of the at least one motion cluster.

12. The method of claim 11 , further comprising estimating the position of the moving vehicle based on each motion cluster from which the noise has been removed,

wherein the current image frame and the previous image frame are obtained by one single image obtaining unit, installed on the moving vehicle, at two different time points.

13. The method of claim 12 , further comprising determining three-dimensional (3D) coordinates of the image obtaining unit based on each motion cluster from which the noise has been removed,

wherein the estimating the position of the moving vehicle based on each motion cluster from which the noise has been removed is performed based on the 3D coordinates of the image obtaining unit.

14. The method of claim 11 , further comprising;

determining three-dimensional (3D) coordinates of one single image obtaining unit, installed on the moving vehicle, which obtains the current image frame and the previous image frame at two different time points, based on each motion cluster from which the noise has been removed; and

estimating the position of the moving vehicle based on the 3D coordinates of the image obtaining unit.

15. The method of claim 14 , wherein the configuring at least one motion cluster comprises configuring a plurality of motion clusters by grouping at least one vector among the vectors based on the spatial correlations,

wherein the determining the 3D coordinates comprises determining a 3D coordinate of the image obtaining unit for each of the plurality of motion clusters from which the noise has been removed, and

wherein the estimating the position of the moving vehicle is performed based on the 3D coordinate of the image obtaining unit for each of the plurality of motion clusters from which the noise has been removed.

16. The method of claim 11 , wherein the noise of each motion cluster comprises a vector which is inconsistent with the other vectors in the each motion cluster.

17. The method of claim 11 , wherein the generating the vectors comprises:

tracking the feature points of the current image frame corresponding to the feature points of the previous image frame; and

generating the vectors connecting the corresponding feature points between the two image frames by detecting optical flow of each pixel.

18. The method of claim 11 , wherein the feature points are extracted using a Kanade-Lucas-Tomasi (KLT) tracker and a Harris corner detector,

wherein the determining the spatial correlations comprises determining the spatial correlations by triangularizing the feature points on a two-dimensional (2D) plane using a Delaunay triangulation technique,

wherein the configuring the motion cluster comprises configuring the motion cluster by applying a Markov random field (MRF) optimization algorithm using graph-cut in a feature space, and

wherein the removing the noise comprises removing an outlier by applying random sample consensus (RANSAC) to each motion cluster.

19. The method of claim 11 , further comprising determining a calibration matrix using internal parameters of an image obtaining unit, installed on the moving vehicle, which obtains the current image frame and the previous image frame at two different time points;

determining an essential matrix based on the calibration matrix and a fundamental matrix determined with respect to each motion cluster, and determines three-dimensional (3D) coordinates of the image obtaining unit based on the essential matrix; and

estimating the position of the moving vehicle based on the 3D coordinates of the image obtaining unit.

20. The method of claim 19 , wherein the configuring at least one motion cluster comprises configuring a plurality of motion clusters by grouping at least one vector among the vectors based on the spatial correlations,

wherein the determining the 3D coordinates comprises determining a 3D coordinate of the image obtaining unit for each of the plurality of motion clusters from which the noise has been removed, and

wherein the estimating the position of the moving vehicle is performed based on the 3D coordinate of the image obtaining unit for each of the plurality of motion clusters from which the noise has been removed.

Assignments (6)