Automated target shape detection for vehicle muon tomography

• US 7,945,105 B1
• Filed: 04/07/2008
• Issued: 05/17/2011
• Est. Priority Date: 04/07/2008
• Status: Active Grant

• Current Assignee: Decision Sciences International Corporation
• Original Assignee: Decision Sciences International Corporation
• Inventors: Jaenisch, Holger M.
• Primary Examiner(s): NGUYEN, KIET TUAN
• Application Number:
  US12/099,072
• Time in Patent Office: 1,135 Days
• Field of Search: 250/306, 250/307, 382/249, 382/128, 382/254
• US Class Current: 382/249
• CPC Class Codes: G01N 9/24 by observing the transmission of wave or particle radiation through the material
• CPC Class Codes: G01V 5/20 Detecting prohibited goods, e.g. weapons, explosives, hazardous substances, contraband or smuggled objects
• CPC Class Codes: G06F 18/00 Pattern recognition
• CPC Class Codes: G06V 10/443 by matching or filtering
• CPC Class Codes: G06V 10/457 by analysing connectivity, e.g. edge linking, connected component analysis or slices
• CPC Class Codes: G06V 20/52 Surveillance or monitoring of activities, e.g. for recognising suspicious objects recognising microscopic objects G06V20/69
• CPC Class Codes: G06V 2201/05 Recognition of patterns representing particular kinds of hidden objects, e.g. weapons, explosives, drugs

Abstract

Techniques, apparatus and systems for muon tomography vehicle imaging use autonomous processing of 3-dimensional muon tomography vehicle images based on Data Modeling techniques and various applications including analyzing vehicle voxel data such as muon vehicle images to detect potential threat objects and then to further discriminate the identified potential threat objects by shape.

Claims

1. A computer-automated method for processing muon vehicle imaging data of a vehicle under inspection located in a vehicle inspection region within a muon tomography vehicle imaging system, comprising:

processing muon vehicle imaging data obtained from the vehicle inspection region to obtain a histogram of the muon vehicle imaging data at different positions in the vehicle inspection region;

separating the muon vehicle imaging data into bins based on the histogram;

removing a subset of the muon vehicle imaging data in a mode bin that has higher frequencies of occurrence than remaining muon vehicle imaging data from the muon vehicle imaging data to retain the remaining muon vehicle imaging data for further processing;

using a vehicle assembly recognizer Data Model for identifying vehicle components from the remaining muon vehicle imaging data;

applying the vehicle assembly recognizer Data Model to process the remaining muon vehicle imaging data to produce an image of the vehicle inspection region by removing vehicle components;

using a target identification Data Model for identifying a target object from the remaining muon vehicle imaging data;

applying the target identification Data Model to process data of the image produced after applying the vehicle assembly recognizer Data Model to produce an image of the target object; and

processing the image of the target object to determine a location and a shape of the target object in the vehicle under inspection.

2. The method of claim 1, wherein:

the processing of the image of the target object to determine a location and a shape of the target object in the vehicle under inspection comprises;

extracting cross-section slices from the image of the target object, each cross-section slice having at least one data point in the image of the target object;

performing a 2-dimensional primitive shape detection algorithm to each slice to determine a 2-dimensional primitive shape of the target object in each slice; and

processing 2-dimensional primitive shapes of the target object in the extracted cross-section slices to construct a 3-dimensional primitive shape representative of the shape of the target object in the vehicle under inspection.

3. The method of claim 2, wherein:

the performing of the 2-dimensional primitive shape detection algorithm to each slice to determine the 2-dimensional primitive shape of the target object in each slice comprises;

applying a plurality of primitive shape Data Models to process each slice to determine whether one or more of primitive shapes respectively corresponding to the primitive shape Data Models be present in each slice.

4. The method of claim 3, wherein:

the primitive shape Data Models comprise a Data Model for a square shape, a Data Model for a circle shape, and a Data Model for a rectangle shape.

5. The method of claim 4, comprising:

dividing each slice into a plurality of neighborhood regions; and

processing each neighborhood region based on the primitive shape Data Models to determine presence of each primitive shape.

6. The method of claim 5, comprising:

using ratios of numbers of identified primitive shapes to numbers of neighborhood regions respectively corresponding to the identified primitive shapes to determine the shape of the target object in the vehicle under inspection.

7. The method of claim 3, wherein:

the performing of the 2-dimensional primitive shape detection algorithm to each slice to determine the 2-dimensional primitive shape of the target object in each slice comprises;

determining a covariance for a surface orientation for each slice.

8. The method of claim 1, comprising:

determining an optimal bin number for the bins.

9. A computer system for processing muon vehicle imaging data of a vehicle volume under inspection within a muon tomography vehicle imaging system, comprising:

means for processing muon vehicle imaging data obtained from the vehicle volume to obtain a histogram of the muon vehicle imaging data at different positions in a vehicle inspection region;

means for separating the muon vehicle imaging data into bins based on the histogram;

means for removing a subset of the muon vehicle imaging data in a mode bin that has higher frequencies of occurrence than remaining muon vehicle imaging data from the muon vehicle imaging data to retain the remaining muon vehicle imaging data for further processing;

means for using a background recognizer Data Model for identifying background structure from the remaining muon vehicle imaging data;

means for applying the background recognizer Data Model to process the remaining muon vehicle imaging data to produce an image of the vehicle inspection region by removing identified background structure;

means for using a target identification Data Model for identifying a target object from the remaining muon vehicle imaging data;

means for applying the target identification Data Model to process data of the image produced after applying the background recognizer Data Model to produce an image of the target object; and

means for processing the image of the target object to determine a location and a shape of the target object in the vehicle volume.

10. The system as in claim 9, wherein:

the background recognizer Data Model is a vehicle assembly recognizer Data Model for identifying vehicle components as the background structure.

11. A computer program product, encoded on a computer-readable medium, operable to cause data processing apparatus to perform operations for processing muon vehicle imaging data of a vehicle under inspection located in a vehicle inspection region within a muon tomography vehicle imaging system, the operations comprising:

processing muon vehicle imaging data obtained from the vehicle inspection region to obtain a histogram of the muon vehicle imaging data at different positions in the vehicle inspection region;

separating the muon vehicle imaging data into bins based on the histogram;

removing a subset of the muon vehicle imaging data in a mode bin that has higher frequencies of occurrence than remaining muon vehicle imaging data from the original muon vehicle imaging data to retain the remaining muon vehicle imaging data for further processing;

using a vehicle assembly recognizer Data Model for identifying vehicle components from the remaining muon vehicle imaging data;

applying the vehicle assembly recognizer Data Model to process the remaining muon vehicle imaging data to produce an image of the vehicle inspection region by removing vehicle components;

using a target identification Data Model for identifying a target object from the remaining muon vehicle imaging data;

applying the target identification Data Model to process data of the image produced after applying the vehicle assembly recognizer Data Model to produce an image of the target object; and

processing the image of the target object to determine a location and a shape of the target object in the vehicle under inspection.

12. A muon tomography vehicle imaging system, comprising:

a first set of muon detectors located on a first side of a vehicle inspection region holding a vehicle for inspection, the muon detectors measuring incident muons;

a second set of muon detectors located on a second side of the vehicle inspection region opposite to the first side to measure outgoing muons exiting the vehicle inspection region; and

a muon tomography control module that receives detector outputs from the first and the second sets of muon detectors and processes the detector outputs to produce a 3-dimensional image of the vehicle inspection region, the muon tomography control module comprising a suspect object detection module for processing muon vehicle imaging data of the 3-dimensional image, the suspect object detection module operable to;

process muon vehicle imaging data obtained from the vehicle inspection region to obtain a histogram of the muon vehicle imaging data at different positions in the vehicle inspection region;

separate the muon vehicle imaging data into bins based on the histogram;

remove a subset of the muon vehicle imaging data in a mode bin that has higher frequencies of occurrence than remaining muon vehicle imaging data from the original muon vehicle imaging data to retain the remaining muon vehicle imaging data for further processing;

using a vehicle assembly recognizer Data Model for identifying vehicle components from the remaining muon vehicle imaging data;

apply the vehicle assembly recognizer Data Model to process the remaining muon vehicle imaging data to produce an image of the vehicle inspection region by removing vehicle components;

using a target identification Data Model for identifying a target object from the remaining muon vehicle imaging data;

apply the target identification Data Model to process data of the image produced after applying the vehicle assembly recognizer Data Model to produce an image of the target object; and

process the image of the target object to determine a location and a shape of the target object in the vehicle under inspection.

13. The system of claim 12, wherein:

the processing of the image of the target object to determine a location and a shape of the target object in the vehicle under inspection comprises;

extracting cross-section slices from the image of the target object, each cross-section slice having at least one data point in the image of the target object;

performing a 2-dimensional primitive shape detection algorithm to each slice to determine a 2-dimensional primitive shape of the target object in each slice; and

processing 2-dimensional primitive shapes of the target object in the extracted cross-section slices to construct a 3-dimensional primitive shape representative of the shape of the target object in the vehicle under inspection.

14. The system of claim 13, wherein:

the performing of the 2-dimensional primitive shape detection algorithm to each slice to determine the 2-dimensional primitive shape of the target object in each slice comprises;

applying a plurality of primitive shape Data Models to process each slice to determine whether one or more of primitive shapes respectively corresponding to the primitive shape Data Models be present in each slice.

15. The system of claim 14, wherein:

the primitive shape Data Models comprise a Data Model for a square shape, a Data Model for a circle shape, and a Data Model for a rectangle shape.

16. The system of claim 15, comprising:

dividing each slice into a plurality of neighborhood regions; and

processing each neighborhood region based on the primitive shape Data Models to determine presence of each primitive shape.

17. The system of claim 12, comprising:

determining an optimal bin number for the bins.

18. A computer-automated method for processing muon vehicle imaging data from a muon tomography vehicle imaging system, comprising:

applying Data Models to process muon tomography vehicle image data to isolate one or more target objects from other objects in a vehicle volume under inspection by the muon tomography vehicle imaging system;

applying primitive shape recognition to cross-section slices of an image of the vehicle volume containing data points of the one or more target objects to identify predetermined primitive shapes; and

processing identified primitive shapes to construct a 3-dimensional image of each target object.

19. The method as in claim 18, wherein:

the applying Data Models to process muon tomography vehicle image data to isolate one or more target objects from other objects comprises;

processing muon vehicle imaging data obtained from the vehicle volume to obtain a histogram of the muon vehicle imaging data at different positions in the vehicle volume;

separating the muon vehicle imaging data into bins based on the histogram;

removing a subset of the muon vehicle imaging data in a mode bin that has higher frequencies of occurrence than remaining muon vehicle imaging data from the muon vehicle imaging data to retain the remaining muon vehicle imaging data for further processing;

applying a background recognizer Data Model for identifying the other objects to process the remaining muon vehicle imaging data to produce an image of the vehicle volume by removing the identified other objects; and

applying a target identification Data Model for identifying the one or more target objects to process data of the image produced after applying the background recognizer Data Model to produce an image of the one or more target objects.

20. A computer-implemented method for using natural cosmic muons as a radiation source to obtain muon tomographic images of a vehicle under inspection, comprising:

obtaining a cosmic muon tomography vehicle image of a vehicle under inspection;

processing vehicle voxel data of each obtained muon vehicle image to compute histogram of the vehicle voxel data of the obtained muon vehicle image;

applying statistical processing based on the histogram to the vehicle voxel data to use a Data Model to monitor one or more changes in the obtained muon vehicle image that trigger a tip-off condition indicative of presence of a target object in the obtained muon vehicle image; and

determining a shape of the target object.

21. The method as in claim 20, wherein:

the shape of the target object is determined by;

extracting cross-section slices from the image of the target object, each cross-section slice having at least one data point in the image of the target object;

performing a 2-dimensional primitive shape detection algorithm to each slice to determine a 2-dimensional primitive shape of the target object in each slice; and

processing 2-dimensional primitive shapes of the target object in the extracted cross-section slices to construct a 3-dimensional primitive shape representative of the shape of the target object in the vehicle under inspection.

22. The method of claim 21, wherein:

the performing of the 2-dimensional primitive shape detection algorithm to each slice to determine the 2-dimensional primitive shape of the target object in each slice comprises;

applying a plurality of primitive shape Data Models to process each slice to determine whether one or more of primitive shapes respectively corresponding to the primitive shape Data Models be present in each slice.

23. The method as in claim 20, comprising:

for a voxel that triggers the tip-off condition, applying the Data Model to selected neighboring voxels for the voxel to perform a shape detection in the selected neighboring voxels.

24. The method as in claim 23, comprising:

generating an overall shape of the target object represented by voxels that trigger the tip-off condition.

25. The method as in claim 23, wherein:

the shape detection includes operating Data Model shape detectors for detecting different primitive shapes to determine presence or absence of the primitive shapes.