SYSTEM AND METHOD FOR CLASSIFICATION OF OBJECTS FROM 3D RECONSTRUCTION

• US 20140254324A1
• Filed: 09/11/2013
• Published: 09/11/2014
• Est. Priority Date: 03/11/2013
• Status: Active Grant

• Current Assignee: Navy The USA As Represented By The Secretary of The
• Original Assignee: The Government Of The United States Of America, As Represented By The Secretary Of The Navy
• Inventors: Dubberley, John R., Lin, Bruce Y.
• Granted Patent: US 9,405,959 B2
• Time in Patent Office: Days
• US Class Current: 367/88
• CPC Class Codes: G01S 15/8902 Side-looking sonar
• CPC Class Codes: G01S 7/539 using analysis of echo signal for target characterisation; Target signature; Target cross-section
• CPC Class Codes: G06T 17/05 Geographic models
• CPC Class Codes: G06V 10/60 relating to illumination properties, e.g. using a reflectance or lighting model
• CPC Class Codes: G06V 20/64 Three-dimensional objects

Abstract

System and method for detecting and classifying man-made objects on the seafloor using 3D reconstruction techniques. Enhanced sea floor object detection with classification is provided that is as good as provided by short range optical imagery. This approach eliminates the step of passing off identification to humans, and enhances the speed, accuracy, and safety of present operations in mine detection and neutralization.

Claims

1. An automatic computer method for detecting and classifying manmade objects comprising:

automatically creating, by a computer, a filter to detect differences in characteristics between natural and manmade objects;

automatically receiving, by the computer, sonar intensity data, the sonar intensity data including objects and background;

automatically selecting, by the computer, manmade objects from the objects and background based on the filter;

automatically separating, by the computer, the selected manmade objects from a background by segmentation;

automatically creating, by the computer, a 3D reconstruction of the selected manmade objects by melding multiple scans of information about the selected manmade objects;

automatically creating, by the computer, feature vectors based on the 3D reconstruction; and

automatically classifying, by the computer, the feature vectors into types based on support vector machine technology and clustering technology.

2. The method as in claim 1 wherein creating the 3D reconstruction comprises:

automatically detecting pixel to pixel variation in seafloor digital images indicating differentiation between platonic and non-platonic shapes.

3. The method as in claim 1 wherein creating the 3D reconstruction comprises:

automatically detecting pixel to pixel variation in seafloor digital images indicating differentiation between objects based on sea encrustation on the objects.

4. The method as in claim 1 wherein creating the 3D reconstruction comprises:

automatically detecting pixel to pixel variation in seafloor digital images indicating differentiation based on facet boundaries of the objects.

5. The method as in claim 1 wherein creating a 3D reconstruction comprises:

automatically extracting height characteristics from the sonar intensity data by performing interferometry between two sonar arrays of the sonar intensity data.

6. The method as in claim 1 wherein creating a 3D reconstruction comprises:

automatically detecting pixel to pixel intensity variation across the sonar intensity data; and

automatically implying a curvature from the variation.

7. The method as in claim 6 wherein implying a curvature comprises:

automatically using a shape-from-shadow technique.

8. The method as in claim 1 wherein creating feature vectors comprises:

automatically computing lengths and widths of the objects in pixels based on the number of pixels along the length and width; and

automatically multiplying the length pixel number and the width pixel number by a length pixel size and a width pixel size, respectively.

9. The method as in claim 1 wherein creating feature vectors comprises:

automatically determining a greatest reflection from the object; and

automatically basing the feature vectors on the greatest reflection.

10. The method as in claim 1 wherein creating feature vectors comprises:

automatically computing the difference between pixel intensity of the brightest pix in the object and the darkest pixel in the shadow.

11. A computer system for detecting and classifying manmade objects comprising:

a filter processor executing on a special purpose computer creating a filter to detect differences in characteristics between natural and manmade objects, the filter processor receiving sonar intensity data, the sonar intensity data including objects and background;

an object processor executing on the special purpose computer selecting manmade objects from the objects and background based on the filter;

a 3D reconstruction processor executing on the special purpose computer creating a 3D reconstruction of the selected manmade objects by melding multiple scans of information about the selected manmade objects;

a segmentation processor on the special purpose computer separating the 3D reconstruction from the background;

a feature vector creator executing on the special purpose computer creating feature vectors based on the 3D reconstruction; and

a classifier executing on the special purpose computer classifying the feature vectors into types based support vector machine technology and clustering technology.

12. The system as in claim 11 further comprising:

a pixel processor executing on the special purpose computer detecting pixel to pixel variation in seafloor digital images indicating differentiation between platonic and non-platonic shapes.

13. The system as in claim 11 further comprising:

a pixel processor executing on the special purpose computer detecting pixel to pixel variation in seafloor digital images indicating differentiation between objects based on sea encrustation on the objects.

14. The system as in claim 11 further comprising:

a pixel processor executing on the special purpose computer detecting pixel to pixel variation in seafloor digital images indicating differentiation based on facet boundaries of the objects.

15. The system as in claim 11 further comprising:

a pixel processor executing on the special purpose computer extracting height characteristics from the sonar intensity data by performing interferometry between two sonar arrays of the sonar intensity data.

16. The system as in claim 11 further comprising:

a pixel processor executing on the special purpose computer detecting pixel to pixel intensity variation across the sonar intensity data, and implying a curvature from the variation.

17. The system as in claim 16 wherein implying a curvature comprises:

using a shape-from-shadow technique.

18. The system as in claim 11 wherein the feature vector creator further comprises:

computer code executing on the special purpose computer;

computing lengths and widths of the objects in pixels based on the number of pixels along the length and width; and

multiplying the length pixel number and the width pixel number by a length pixel size and a width pixel size, respectively.

19. The system as in claim 11 wherein the feature vector creator further comprises:

computer code executing on the special purpose computer;

determining a greatest reflection from the object; and

basing the feature vectors on the greatest reflection.

20. The system as in claim 11 wherein the feature vector creator further comprises:

computer code executing on the special purpose computer computing the difference between pixel intensity of a brightest pixel in the object and a darkest pixel in the shadow.