Object movement estimator using one-dimensional optical flow

• US 5,717,792 A
• Filed: 12/15/1995
• Issued: 02/10/1998
• Est. Priority Date: 09/13/1993
• Status: Expired due to Fees

• Current Assignee: Covey Leadership Center, Inc. (Franklin Covey Company)
• Original Assignee: Massachusetts Institute of Technology
• Inventors: Harris, John G., Ancona, Nicola, Poggio, Tomaso
• Primary Examiner(s): Couso, Jose L.
• Assistant Examiner(s): Bella, Matthew C.
• Application Number:
  US08/573,044
• Time in Patent Office: 788 Days
• Field of Search: 382/278, 382/279, 382/107, 348/413, 348/415, 348/416, 364/424.01, 364/424.02, 364/449
• US Class Current: 382/278
• CPC Class Codes: G06T 7/20 Analysis of motion motion estimation for coding, decoding, compressing or decompressing digital video signals H04N19/43, H04N19/51

Abstract

An object movement estimator (400) predicts optical flow by estimating changes in sensed images using one-dimensional correlations. Sensed image portions are stored at a first instant of time and compared with sensed image portions at a second instant of time. An optimal flow is estimated by performing a one-dimensional correlation of the sensed image portions at the first and second instants of time, and a time-to-contact is estimated based on the predicted optical flow thus obtained.

Claims

1. A system for predicting a rotational speed of a first object relative to a second object, comprising:

sensing means affixed to said first object for sensing a two-dimensional plurality of image portions of said second object, said sensing means including a first group of sensors and a second group of sensors, the first group of sensors not being collinear with the second group of sensors;

a first memory coupled to said sensing means for storing said sensed image portions at a first instant of time;

a second memory coupled to said sensing means for storing said sensed image portions at a second instant of time;

a correlator coupled to said first and second memories for estimating one-dimensional optical flows by performing correlations on said sensed image portions at said first instant of time and said sensed image portions at said second instant of time, a first one of said correlations corresponding to a first component of said image portions incident on said first group of sensors and a second one of said correlations corresponding to a second component of said image portions incident on said second group of sensors; and

an estimator, coupled to said correlator, for generating an estimate of said rotational speed in response to results of said correlations by said correlator.

5. A system as in claim 1, wherein said first and second groups of sensors comprise two orthogonally configured rows of optical detectors.

9. A system as in claim 1, said correlator further comprising a gaussian filter, said gaussian filter adapted to accept as input said sensed image portions stored in said first and second memories and to produce therefrom filtered image portion signals for application to said correlator, the correlator operating according to the convolution formula:

space="preserve" listing-type="equation">E=(Gσ

.sub.y (y)*E)*G"σ

.sub.x (x)

where

E is an array of said filtered image portion signals,

E is an array of said sensed image portions stored in said first and second memories,

G is the gaussian filter transfer function,

G" is a second derivative of G,

σ

_y is a standard deviation of the gaussian filter in the y dimension, and

σ

_x is a standard deviation of the gaussian filter in the x dimension.

2. A system for predicting a rotational speed of a first object relative to a second object, comprising:

a plurality of sensing means affixed to said first object, each of said sensing means sensing a one-dimensional plurality of image portions of said second object, at least two of said sensing means not being collinear;

a first memory coupled to said plurality of sensing means for storing said sensed image portions at a first instant of time;

a second memory coupled to plurality of sensing means for storing said sensed image portions at a second instant of time;

a correlator coupled to said first and second memories for estimating one-dimensional optical flows by performing correlations on said sensed image portions at said first instant of time and said sensed image portions at said second instant of time, each of such correlations being responsive to one of said image portions incident on a corresponding one of said sensing means; and

an estimator, coupled to said correlation means, for generating an estimate of said rotational speed in response to results of said correlation by said correlation means.

6. A system as in claim 2, wherein said sensing means comprises a plurality of rows of optical detectors, said plurality of rows defining a plane.

10. A system as in claim 2, said correlator further comprising a gaussian filter, said gaussian filter adapted to accept as input said sensed image portions stored in said first and second memories and to produce therefrom filtered image portion signals for application to said correlator, the correlator operating according to the convolution formula:

space="preserve" listing-type="equation">E=(Gσ

.sub.y (y)*E)*G"σ

.sub.x (x)

where

E is an array of said filtered image portion signals,

E is an array of said sensed image portions stored in said first and second memories,

G is the gaussian filter transfer function,

G" is a second derivative of G,

σ

_y is a standard deviation of the gaussian filter in the y dimension, and

σ

_x is a standard deviation of the gaussian filter in the x dimension.

12. A system as in claim 6, said correlator further comprising a gaussian filter, said gaussian filter adapted to accept as input said sensed image portions stored in said first and second memories and to produce therefrom filtered image portion signals for application to said correlator, the correlator operating according to the convolution formula:

space="preserve" listing-type="equation">E=(Gσ

.sub.y (y)*E)*G"σ

.sub.x (x)

where

E is an array of said filtered image portion signals,

E is an array of said sensed image portions stored in said first and second memories,

G is the gaussian filter transfer function,

G" is a second derivative of G,

σ

_y is a standard deviation of the gaussian filter in the y dimension, and

σ

_x is a standard deviation of the gaussian filter in the x dimension.

3. A system for predicting a translational velocity of a first object relative to a second object, comprising:

sensing means affixed to said first object for sensing a two-dimensional plurality of image portions of said second object, said sensing means including a first group of sensors and a second group of sensors, the first group of sensors not being collinear with the second group of sensors;

a first memory coupled to said sensing means for storing said sensed image portions at a first instant of time;

a second memory coupled to said sensing means for storing said sensed image portions at a second instant of time;

a correlator coupled to said first and second memories for estimating one-dimensional optical flows by performing correlations on said sensed image portions at said first instant of time and said sensed image portions at said second instant of time, a first one of said correlations corresponding to a first component of said image portions incident on said first group of sensors and a second one of said correlations corresponding to a second component of said image portions incident on said second group of sensors; and

an estimator, coupled to said correlator, for generating an estimate of said translational velocity in response to results of said correlation by said correlation means.

7. A system as in claim 3, wherein said first and second groups of sensors comprise two orthogonally configured rows of optical detectors.

11. A system as in claim 3, said correlator further comprising a gaussian filter, said gaussian filter adapted to accept as input said sensed image portions stored in said first and second memories and to produce therefrom filtered image portion signals for application to said correlator, the correlator operating according to the convolution formula:

space="preserve" listing-type="equation">E=(Gσ

.sub.y (y)*E)*G"σ

.sub.x (x)

where

E is an array of said filtered image portion signals,

E is an array of said sensed image portions stored in said first and second memories,

G is the gaussian filter transfer function,

G" is a second derivative of G,

σ

_y is a standard deviation of the gaussian filter in the y dimension, and

σ

_x is a standard deviation of the gaussian filter in the x dimension.

4. A system for predicting a translational velocity of a first object relative to a second object, comprising:

a plurality of sensing means affixed to said first object, each of said sensing means sensing a one-dimensional plurality of image portions of said second object, at least two of said sensing means not being collinear;

a first memory coupled to said plurality of sensing means for storing said sensed image portions at a first instant of time;

a second memory coupled to plurality of sensing means for storing said sensed image portions at a second instant of time;

a correlator coupled to said first and second memories for estimating one-dimensional optical flows by performing correlations on said sensed image portions at said first instant of time and said sensed image portions at said second instant of time, each of such correlations being responsive to one of said image portions incident on a corresponding one of said sensing means; and

an estimator, coupled to said correlator, for generating an estimate of said translational velocity in response to results of said correlations by said correlator.

8. A system as in claim 4, wherein said sensing means comprises a plurality of rows of optical detectors, said plurality of rows defining a plane.

13. A method of predicting a rotational speed of a first object relative to a second object, the method comprising:

sensing and storing a two-dimensional plurality of image portions of said second object as viewed from said first object at a first instant of time, using a first group of sensors and a second group of sensors, the first group of sensors not being collinear with the second group of sensors;

sensing said two-dimensional plurality of image portions at a second instant of time, using said first group of sensors and said second group of sensors;

estimating optical flows in one dimension by performing one-dimensional correlations on said sensed two-dimensional plurality of image portions at said first and second instants of time, a first one of said correlations corresponding to a first component of said image portions incident on said first group of sensors and a second one of said correlations corresponding to a second component of said image portions incident on said second group of sensors; and

estimating said rotational speed in response to fie results of said estimating optical flows.

17. A method as in claim 13, wherein estimating optical flows includes determining optical flow vectors at each of said sensed image portions, and wherein estimating said rotational speed includes determining a tangential component metric of said optical flow vectors.

14. A method of predicting a rotational speed of a first object relative to a second object, the method comprising:

sensing and storing a two-dimensional plurality of image sets over a two-dimensional imaging area, each of said image sets including a one-dimensional plurality of image portions of said second object as viewed from said first object at a first instant of time, using a first group of sensors and a second group of sensors, the first group of sensors not being collinear with the second group of sensors;

sensing said two-dimensional plurality of image portions at a second instant of time, using said first group of sensors and said second group of sensors;

estimating one-dimensional optical flows over the two-dimensional imaging area by performing one-dimensional correlations on each of said sensed image portions at said first and second instants of time, a first one of said correlations corresponding to a first component of said image portions incident on said first group of sensors and a second one of said correlations corresponding to a second component of said image portions incident on said second group of sensors; and

estimating said rotational speed in response to the results of said estimating optical flows.

18. A method as in claim 14, wherein estimating optical flows includes determining optical flow vectors at each of said sensed image portions, and wherein estimating said rotational speed includes determining a tangential component metric of said optical flow vectors.

15. A method of predicting a translational velocity of a first object relative to a second object, the method comprising:

sensing and storing a two-dimensional plurality of image portions of said second object as viewed from said first object at a first instant of time, using a first group of sensors and a second group of sensors, the first group of sensors not being collinear with the second group of sensors;

sensing said two-dimensional plurality of image portions at a second instant of time, using said first group of sensors and said second group of sensors;

estimating optical flows in one dimension by performing one-dimensional correlations on said sensed two-dimensional plurality of image portions at said first and second instants of time, a first one of said correlations corresponding to a first component of said image portions incident on said first group of sensors and a second one of said correlations corresponding to a second component of said image portions incident on said second group of sensors; and

estimating said translational velocity in response to the results of said estimating optical flows.

19. A method as in claim 15, wherein estimating optical flows includes determining optical flow vectors at each of said sensed image portions, and wherein estimating said translational velocity includes determining a unidirectional component metric of said optical flow vectors.

16. A method of predicting a translational velocity of a first object relative to a second object, the method comprising:

sensing and storing a two-dimensional plurality of image sets over a two-dimensional imaging area, each of said image sets including a one-dimensional plurality of image portions of said second object as viewed from said first object at a first instant of time, using a first group of sensors and a second group of sensors, the first group of sensors not being collinear with the second group of sensors;

sensing said two-dimensional plurality of image portions at a second instant of time, using said first group of sensors and said second group of sensors;

estimating one-dimensional optical flows over the two-dimensional imaging area by performing one-dimensional correlations on each of said sensed image portions at said first and second instants of time, a first one of said correlations corresponding to a first component of said image portions incident on said first group of sensors and a second one of said correlations corresponding to a second component of said image portions incident on said second group of sensors; and

estimating said translational velocity in response to the results of said estimating optical flows.

20. A method as in claim 16, wherein estimating optical flows includes determining optical flow vectors at each of said sensed image portions, and wherein estimating said translational velocity includes determining a unidirectional component metric of said optical flow vectors.