Automatic scene correlation and identification

1. A method of determining location of a target image within a two-dimensional input image by employing a three-dimensional reference model, the method comprising the steps of:

extracting object edge points in the input image;

given estimated seeker line-of-sight orientation and range data, projecting a three-dimensional target reference model to obtain a two-dimensional reference image;

applying a weighted-direction-cross-product Hough Transform to the object edge points and points of the two-dimensional reference image;

classifying the input image as containing a target image or not;

if the input image contains the target image, identifying a location of the target image within the two-dimensional input image;

computing confidence measure as to the identified location of the target image within the two-dimensional input image; and

performing confidence accumulation over a plurality of instances of application of the method.

2. The method of claim 1 wherein the extracting step comprises clutter-rejection.

3. The method of claim 1 wherein the projecting step comprises the steps of:

decompressing a compressed three-dimensional target reference model;

projecting the three-dimensional target reference model to obtain a two-dimensional reference image, given estimated seeker line-of-sight orientation and range data; and

continuously adjusting extent of a ground footprint of the projected two-dimensional reference image based upon estimated seeker-to-ground range.

4. The method of claim 1 wherein the applying step comprises the steps of:

accumulating Hough correlation counts into Hough planes for individual Canny edge directions;

super-pixeling individual Hough planes;

generating the filtered weighted-direction-cross-product Hough correlation surface in super-pixel domain;

selecting an area of interest in the original pixel domain;

generating the filtered weighted-direction-cross-product Hough correlation surface in original pixel domain; and

detecting a peak of the correlation surface as an aim-point.

5. The method of claim 1 wherein the classifying and computing steps comprise employing feature extraction and computing quality measures as to the identified location selected from the group consisting of scene classification and confidence measures.

6. The method of claim 1 wherein the applying step comprises the step of combining Hough planes by pair-wise AND logic.

7. The method of claim 6 wherein the combining step comprises multiplying correlation counts from each pair of Hough planes.

8. The method of claim 7 wherein the combining step comprises performing a weighted sum over the pair-wise products from all combinations of Hough planes in pair.

9. Computer software for determining location of a target image within a two-dimensional input image by employing a three-dimensional reference model, said software comprising code embodied in a storage medium, said code comprising:

code extracting object edge points in the input image;

code projecting a three-dimensional target reference model to obtain a two-dimensional reference image given estimated seeker line-of-sight orientation and range data;

code applying a weighted-direction-cross-product Hough Transform to the object edge points and points of the two-dimensional reference image;

code classifying the input image as containing a target image or not;

code identifying a location of the target image within the two-dimensional input image if the input image contains the target image;

code computing confidence measure as to the identified location of the target image within the two-dimensional input image; and

code performing confidence accumulation over a plurality of instances of application of the method.

10. The software of claim 9 wherein said extracting code comprises clutter-rejection code.

11. The software of claim 9 wherein said projecting code comprises the following codes:

code decompressing a compressed three-dimensional target reference model;

code projecting the three-dimensional target reference model to obtain a two-dimensional reference image, given estimated seeker line-of-sight orientation and range data; and

code continuously adjusting extent of a ground footprint of the projected two-dimensional reference image based upon estimated seeker-to-ground range.

12. The software of claim 9 wherein said applying code comprises the following codes:

code accumulating Hough correlation counts into Hough planes for individual Canny edge directions;

code super-pixeling individual Hough planes;

code generating the filtered weighted-direction-cross-product Hough correlation surface in super-pixel domain;

code selecting an area of interest in the original pixel domain;

code generating the filtered weighted-direction-cross-product Hough correlation surface in area of interest in the original pixel domain; and

code detecting a peak of the correlation surface as an aim-point.

13. The software of claim 9 wherein said classifying and computing codes comprise code for feature extraction and code for computing quality measures as to the identified location consisting of code of scene classification and code for confidence measures.

14. The software of claim 9 wherein said applying code comprises the code of combining Hough planes by pair-wise AND logic.

15. The software of claim 14 wherein said combining step comprises code multiplying correlation counts from each pair of Hough planes.

16. The software of claim 15 wherein said combining code comprises code performing a weighted sum over the pair-wise products from all combinations of Hough planes in pair.