Clutter rejection through edge integration

• US 6,081,605 A
• Filed: 03/08/1993
• Issued: 06/27/2000
• Est. Priority Date: 03/08/1993
• Status: Expired due to Fees

• Current Assignee: the united states of america as represented by the secretary of the navy
• Original Assignee: the united states of america as represented by the secretary of the navy
• Inventors: Roth, Duane, Morrison, Larry M.
• Primary Examiner(s): Couso, Jose L.
• Application Number:
  US08/028,013
• Time in Patent Office: 2,668 Days
• Field of Search: 382/1, 382/30, 382/34, 382/41, 382/42, 382/48, 358/105, 358/107, 358/108
• US Class Current: 382/103
• CPC Class Codes: G06V 10/255 Detecting or recognising potential candidate objects based on visual cues, e.g. shapes
• CPC Class Codes: G06V 10/30 Noise filtering

Abstract

A device and method for performing clutter rejection through edge integran. The invention uses optical flow estimation to produce target centering information which is used to control an image source so that a detected moving target is maintained in the center of an image containing the target and accompanying clutter while continuous, real-time integration of the target edges is performed. The output image produced contains an enhanced target and the target to clutter ratio is, therefore, enhanced.

Claims

1. an image processing device for discriminating a target from clutter in an image containing a moving target, said device comprising:

a means for detecting a moving target in an image input, said means additionally having a target centering input;

a means, connected to said detecting means, for extracting the edge content of said target;

a means, connected to said edge extracting means for integrating the extracted edges of said image over time to reject clutter, said integrating means providing an output image;

a means, connected to said integrating means, for estimating the optical flow of said target; and

a means, connected to receive as an input the output of said estimating means and having its output connected as an input to said detection means, for centering the target in said image to permit the integrating means to increase the target to clutter ratio.

2. The image processing device of claim 1 wherein said means for detecting a moving target in an image input is a spatio-temporal filter.

3. The image processing device of claim 2 wherein said spatio-temporal filter, comprises:

means for Gaussian filtering a current image input;

means for Gaussian filtering a delayed version of said current image input; and

means for subtracting the current filtered image from the delayed filtered image.

4. The image processing device of claim 1 wherein said means for extracting the edge content of said target is a two-dimensional bandpass filter.

5. The device of claim 4 wherein said two-dimensional bandpass filter, comprises:

a means for producing a Gaussian kernel with broad support;

a means for producing a Gaussian kernel with narrow support; and

a means for subtracting said broad support Gaussian kernel from said narrow support Gaussian kernel.

6. The image processing device of claim 1 wherein said means for integrating the edges of the image over time is a leaky integrator performing clutter rejection and is of the following form:

space="preserve" listing-type="equation">y[n]=A*y[n-1]+B*x[n]

where,

y[n]=the out signal;

x[n]=the input signal;

A and B are parameters to be chosen.

7. The device of claim 6 wherein A and B are chosen to sum to 1.0.

8. The image processing device of claim 1 wherein said means for estimating the optical flow of the target, comprises:

means for calculating and storing a target template containing the greatest amount of energy;

means for correlating the calculated template with each successive image in accordance with ##EQU2## where, O(i,j)=the result of the correlation;

T(m,n)=the target template;

I(i,j)=the input image patch;

i and j are the image indexes;

m and n are the template indexes;

M Horizontal diminsion of window in pixels

N Vertical dimension of window in pixels

means for identifying the maximum correlation point for the template in use, said maximum point defining optical flow; and

means for periodically recalculating the template to reflect changes in the target shape.

9. The image processing device of claim 1 wherein said means for centering the target in the image to permit the integrating means to increase the target to clutter ratio is a feedback device to control the image source selected from the group consisting of a seeker and a focal plane array, so that the target initially detected by said target detection means is centered in the image continuously to permit continuous integration by said integration means.

10. The method of image processing for discriminating a target from clutter in an image containing a moving target, comprising the steps of:

a) detecting a moving target in an input image in conjunction with a target centering input;

b) extracting the edge content of said target;

c) integrating the extracted edges of said image over time to reject clutter and providing an output image;

d) estimating the optical flow of said target; and

e) producing a target centering input from said optical flow information for said detection step.

11. The method of claim 10 wherein said detecting step includes spatio-temporal filtering.

12. The method of claim 11 wherein said spatio-temporal filtering includes the steps of:

a) Gaussian filtering a current image input;

b) Gaussian filtering a delayed version of said current image input; and

c) subtracting the current filtered image from the delayed filtered image.

13. The method of claim 10 wherein said extracting step comprises the steps of:

a) producing a Gaussian kernel with broad support;

b) producing a Gaussian kernel with narrow support; and

c) subtracting said broad support Gaussian kernel from said narrow support Gaussian kernel.

14. The method of claim 10 wherein said integrating step comprises performance of leaky integration for clutter rejection and is of the form:

space="preserve" listing-type="equation">y[n]=A*y[n-1]+B*x[n]

where,

y[n]=the out signal;

x[n]=the input signal;

A and B are parameters to be chosen.

15. The method of claim 14 wherein parameters A and B are chosen to sum to 1.0.

16. The method of claim 10 wherein the step of estimating the optical flow of the target, comprises the steps of:

a) segmenting the image;

b) choosing the segment with the greatest amount of energy to be a target template;

c) calculating and storing said target template;

d) correlate the calculated template with each successive image in accordance with ##EQU3## where, O(i,j)=the result of the correlation;

T(m,n)=the target template;

I(i,j)=the input image patch;

i and j are the image indexes;

m and n are the template indexes;

M Horizontal diminsion of window in pixels

N Vertical dimension of window in pixels

e) identify the maximum correlation point for the template in use, said maximum point defining optical flow; and

f) periodically recalculating the template to reflect changes in the target shape.