Object classification using two-dimensional projection

• US 9,042,661 B2
• Filed: 09/30/2010
• Issued: 05/26/2015
• Est. Priority Date: 09/30/2010
• Status: Active Grant

• Current Assignee: Analogic Corporation
• Original Assignee: Analogic Corporation
• Inventors: Pavlovich, Julia, Ying, Zhengrong, Naidu, Ram
• Primary Examiner(s): Li, Ruiping
• Application Number:
  US13/876,131
• Publication Number:
  US 20130202217A1
• Time in Patent Office: 1,699 Days
• Field of Search: 382/218, 382/103, 382/199, 382/170
• US Class Current: 382/218
• CPC Class Codes: G06T 11/008 Specific post-processing after tomographic reconstruction, e.g. voxelisation, metal artifact correction
• CPC Class Codes: G06V 10/76 based on eigen-space representations, e.g. from pose or different illumination conditions; Shape manifolds
• CPC Class Codes: G06V 2201/05 Recognition of patterns representing particular kinds of hidden objects, e.g. weapons, explosives, drugs

Abstract

One or more systems and/or techniques are provided to identify objects comprised in a compound object without segmenting three-dimensional image data of the potential compound object. Two-dimensional projections of a potential compound object (e.g., Eigen projections) are examined to identify the presence of known objects. The projections are compared to signatures, such as morphological characteristics, of one or more known objects. If it is determined based upon the comparison that there is a high likelihood that the compound object comprises a known object, a portion of the projection is masked, and it is compared again to the signature to determine if this likelihood has increased. If it has, a sub-object of the compound object may be classified based upon characteristics of the known object (e.g., the compound object may be classified as a potential threat item if the known object is a threat item).

Claims

1. A method for classifying a potential compound object, comprising:

generating a first fidelity score for a projection of the potential compound object, where the first fidelity score represents a likelihood that the potential compound object comprises a comparator characteristic of a known object;

if the first fidelity score meets a first threshold, generating a masked projection of the potential compound object by excluding portions of the projection of the potential compound object that comprise outliers to the comparator characteristic, comprising;

identifying a distance between an edge of the potential compound object in the projection and a central point of the comparator characteristic; and

masking out one or more portions of the potential compound object in the projection that meet a distance threshold for removal; and

generating a second fidelity score for the masked projection of the potential compound object using the comparator characteristic of the known object, at least some of the method implemented at least in part via a processing unit.

2. The method of claim 1, comprising generating a boundary image of the potential compound object in the projection prior to the generating a first fidelity score.

3. The method of claim 1, comprising:

generating an Eigen projection of the potential compound object; and

generating an Eigen edge map image of the Eigen projection of the potential compound object.

4. The method of claim 1, wherein the comparator characteristic comprises a morphological characteristic.

5. The method of claim 4, the generating a first fidelity score comprising:

comparing a morphological characteristic of the potential compound object in the projection with the morphological characteristic of the known object; and

computing the first fidelity score for the potential compound object based on the comparison.

6. The method of claim 1, the generating a first fidelity score comprising:

overlaying an image of a shape characteristic over an image of the potential compound object in the projection; and

adjusting a size of the image of the shape characteristic so that the size of the image of the shape characteristic is to scale with the image of the potential compound object in the projection.

7. The method of claim 6, the generating a masked projection of the potential compound object comprising:

finding a central area of the image of the shape characteristic;

dividing the adjusted image of the shape characteristic into sectors relative to the central area; and

masking out one or more portions of the image of the potential compound object in the projection that lie in sectors comprising a distance threshold between the central area and an edge in the image of the potential compound object.

8. The method of claim 1, the masking out one or more portions of the potential compound object in the projection comprising masking out a portion of the potential compound object in the projection that comprises a largest outlier portion from the central point.

9. The method of claim 1, comprising refitting the comparator characteristic to the masked projection.

10. The method of claim 9, if the second fidelity score meets a second threshold, different than the first threshold, classifying the masked projection based on the known object.

11. The method of claim 1, comprising classifying a dominant portion of the potential compound object using the comparator characteristic if the second fidelity score for the masked projection of the potential compound object meets a second threshold, different than the first threshold.

12. A system for classifying a potential compound object, comprising:

one or more processing units; and

memory comprising instructions that when executed by at least one of the one or more processing units perform operations, comprising;

generating a first fidelity score for a projection of the potential compound object that represents a likelihood that one or more portions of the potential compound object comprise a comparator characteristic of a known object, comprising;

overlaying an image of a shape characteristic over an image of the potential compound object in the projection; and

adjusting a size of the image of the shape characteristic so that the size of the image of the shape characteristic is to scale with the image of the potential compound object in the projection;

generating a masked projection of the potential compound object by masking out one or more portions of the projection of the potential compound object that comprise outliers to the comparator characteristic, if the first fidelity score meets a first threshold, comprising;

finding a central area of the image of the shape characteristic;

dividing the adjusted image of the shape characteristic into sectors relative to the central area; and

masking out a portion of the image of the potential compound object in the projection that lies in a sector comprising a distance threshold between the central area and an edge in the image of the potential compound object; and

generating a second fidelity score for the masked projection of the potential compound object using the comparator characteristic of the known object.

13. The system of claim 12, comprising generating the projection of the potential compound object using three-dimensional image data of the potential compound object.

14. The system of claim 13, the generating the projection comprising mapping the projection to Eigen space.

15. The system of claim 12, comprising generating an edge map image of the projection of the potential compound object and the generating a first fidelity score comprising generating the first fidelity score using the edge map image.

16. The system of claim 15, wherein the edge map image is an Eigen edge map.

17. The system of claim 12, the generating a first fidelity score comprising generating the first fidelity score for an unmasked potential compound object in the projection.

18. The system of claim 12, wherein the comparator characteristic is derived from a library of known comparator characteristics.

19. A method for classifying a potential compound object identified in image data from a radiographic imaging apparatus, comprising:

generating an Eigen edge map of an Eigen space representation of the potential compound object, where the Eigen edge map comprises merely a boundary representation of the potential compound object in Eigen space;

comparing a signature of the Eigen edge map with a signature of a known object characteristic to generate a first likelihood score;

if the first likelihood score meets a first threshold, removing portions of the Eigen edge map that meet a removal threshold based on the known object characteristic to generate a masked Eigen edge map, comprising;

identifying a distance between an edge of the potential compound object in the Eigen edge map and a central point of the signature of the known object characteristic; and

masking out one or more portions of the potential compound object in the Eigen edge map that meet a distance threshold for removal; and

comparing a signature of the masked Eigen edge map with the signature of the known object characteristic to generate a second likelihood score.

20. The method of claim 19, comprising converting three-dimensional image data of the potential compound object into the Eigen space representation of the potential compound object.

21. The method of claim 19, comprising classifying a dominant portion of the potential compound object based on the known object characteristic if the second likelihood score meets a second threshold.

22. The method of claim 19, wherein the known object characteristic comprises a morphological characteristic.