Adaptive bayes image correlation

• US 7,974,475 B1
• Filed: 08/20/2009
• Issued: 07/05/2011
• Est. Priority Date: 08/20/2009
• Status: Expired due to Fees

• Current Assignee: Thomas Cecil Minter
• Original Assignee: Thomas Cecil Minter
• Inventors: Minter, Thomas Cecil
• Primary Examiner(s): PERUNGAVOOR, SATHYANARAYA V
• Application Number:
  US12/583,395
• Time in Patent Office: 684 Days
• Field of Search: 382/209
• US Class Current: 382/209
• CPC Class Codes: G06F 18/24155 Bayesian classification
• CPC Class Codes: G06V 10/7515 Shifting the patterns to accommodate for positional errors
• CPC Class Codes: G06V 10/764 using classification, e.g. of video objects

Abstract

This invention relates generally to a system and method for correlating two images for the purpose of identifying a target in an image where templates are provided a priori only for the target. Information on other objects in the image being searched may be unavailable or difficult to obtain. This invention treats the design of target matching-templates and target matched-filters for image correlation as a statistical pattern recognition problem. By minimizing a suitable criterion, a target matching-template or a target matched-filter is estimated which approximates the optimal Bayes discriminant function in a least-squares sense. Both Bayesian image correlation methods identify the target with minimum probability of error while requiring no prior knowledge of other objects in the image being searched. The system and method is adaptive in that it can be re-optimizing (adapted) to recognize the target in a new search image using only information from the new image.

Claims

1. A method for classifying and locating a target in a digital search image, comprising the steps of:

receiving target reference images descriptive of said target and an estimate of the a priori probability of said target in said digital search image;

forming a plurality of labeled target measurement vectors from said target reference images;

receiving said digital search image;

forming a plurality of unlabeled measurement vectors from said digital search image;

executing a training stage using said a priori probability of said target in said digital search image, said labeled target measurement vectors, and said of unlabeled measurement vectors from said digital search image, including a step of least squares estimation of a Bayes parameter weighting vector;

forming a two dimensional Bayes target matching-template from said Bayes parameter weighting vector;

correlating said two-dimensional Bayes target matching-template with said digital search image using a spatial domain template-matching image correlation method; and

executing an adaptive Bayes classifier to identify objects as either said target or as not-target and determining said target location in said digital search image.

2. The method of claim 1 wherein said step of forming a plurality of labeled target measurement vectors from said target reference images includes a step of transforming the contents of each two-dimensional said target reference image into a said labeled target measurement vector.

3. The method of claim 1 wherein said step of forming a plurality of unlabeled measurement vectors from said digital search image includes a step of scanning said digital search image with a two-dimensional window, equal in size to said target reference images, and transforming the contents of each said two-dimensional widow into a said unlabeled measurement vector.

4. The method of claim 1 wherein said step of correlating said two-dimensional Bayes target matching-template with said digital search image using a spatial domain template-matching image correlation method includes a step of computing a correlation measure which is the value of the target posterior probability at a plurality of locations in said digital search image.

5. The method of claim 1 wherein said step of executing an adaptive Bayes classifier to identify objects as either said target or as not-target includes a step of evaluating said value of target posterior probability using the adaptive Bayes decision rule to identify said object as either said target or as said not-target.

6. The method of claim 1 wherein said step of determining said target location in said digital search image includes the step of identifying locations in said digital search image where said value of target posterior probability peaks for said objects identified as a said target using said adaptive Bayes decision rule.

13. The method of claim 1 wherein said step of correlating said two-dimensional Bayes target matching-template with said digital search image using a frequency domain matched-filter filter image correlation method includes a step of computing a correlation measure which is the value of the target posterior probability at a plurality of locations in said digital search image.

7. A method for classifying and locating a target in a digital search image, comprising the steps of:

receiving target reference images descriptive of said target and an estimate of the a priori probability of said target in said digital search image;

forming a plurality of labeled target measurement vectors from said target reference images;

receiving said digital search image;

forming a plurality of unlabeled measurement vectors from said digital search image;

executing a training stage using said a priori probability of said target in said digital search image, said labeled target measurement vectors, and said of unlabeled measurement vectors from said digital search image, including a step of least squares estimation of a Bayes parameter weighting vector;

forming a two dimensional Bayes target matching-template from said Bayes parameter weighting vector;

correlating said two-dimensional Bayes target matching-template with said digital search image using a frequency domain matched-filter image correlation method; and

executing an adaptive Bayes classifier to identify objects as either said target or as not-target and determining said target location in said digital search image.

8. The method of claim 7 wherein said step of forming a plurality of labeled target measurement vectors from said target reference images includes a step of transforming the contents of each two-dimensional said target reference image into a said labeled target measurement vector.

9. The method of claim 7 wherein said step of forming a plurality of unlabeled measurement vectors from said digital search image includes a step of scanning said digital search image with a two-dimensional window, equal in size to said target reference images, and transforming the contents of each said two-dimensional widow into a said unlabeled measurement vector.

10. The method of claim 7 wherein said step of correlating said two-dimensional Bayes target matching-template with said digital search image using a frequency domain matched-filter includes a step of applying a two-dimensional time domain truncation function to said Bayes target weighting-template.

11. The method of claim 7 wherein said step of correlating said two-dimensional Bayes target matching-template with said digital search image using a frequency domain matched-filter filter image correlation method includes a step of padding the truncated said Bayes target weighting-template and said digital search image with zeroes.

12. The method of claim 7 wherein said step of correlating said two-dimensional Bayes target matching-template with said digital search image using a frequency domain matched-filter filter image correlation method includes a step of applying a fast Fourier transform to the padded and truncated said Bayes target weighting-template and the padded said digital search image.

14. The method of claim 7 wherein said step of executing an adaptive Bayes classifier to identify objects as either said target or as not-target includes a step of evaluating said value of target posterior probability using the adaptive Bayes decision rule to identify said object as either said target or as said not-target.

15. The method of claim 7 wherein said step of determining said target location in said digital search image includes the step of identifying a location in said digital search image where said value of target posterior probability peaks for said objects identified as a said target using said adaptive Bayes decision rule.