Collaborative feature extraction system for three dimensional datasets

• US 8,811,748 B2
• Filed: 05/21/2012
• Issued: 08/19/2014
• Est. Priority Date: 05/20/2011
• Status: Active Grant

• Current Assignee: Autodesk, Inc.
• Original Assignee: Autodesk, Inc.
• Inventors: Morris, Aaron C.
• Primary Examiner(s): Koziol, Stephen R
• Assistant Examiner(s): Lee, Jonathan S
• Application Number:
  US13/476,676
• Publication Number:
  US 20120294532A1
• Time in Patent Office: 820 Days
• Field of Search: None
• US Class Current: 382/195
• CPC Class Codes: G01S 17/89 for mapping or imaging
• CPC Class Codes: G01S 7/4802 using analysis of echo signal for target characterisation; Target signature; Target cross-section
• CPC Class Codes: G06F 16/29 Geographical information databases

Abstract

A collaborative feature extraction system uses crowdsourced feedback to improve its ability to recognize objects in three-dimensional datasets. The system accesses a three-dimensional dataset and presents images showing possible objects to a group of users, along with potential identifiers for the objects. The users provide feedback as to the accuracy of the identifiers, and the system uses the feedback to adjust parameters for candidate identifiers to improve its recognition of three-dimensional assets in future iterations.

Claims

1. A collaborative feature extraction system, comprising:

a knowledge base containing a first dataset, the first dataset comprising parameters that are associated with one or more assets;

a second dataset, the second dataset comprising three-dimensional data representing a geographic area and one or more candidate objects within the geographic area;

a processor in electronic communication with the knowledge base; and

a computer-readable medium containing programming instructions that, when executed, cause the processor to;

analyze the three-dimensional data to identify one of the candidate objects as a selected one of the assets by comparing the second dataset with one or more of the parameters in the first dataset;

output, to multiple users, a first image representing the identified candidate object and a user prompt;

in response to the user prompt, receive user responses from each of the multiple users indicating whether the selected asset matches the identified candidate object; and

based on receipt of user responses from a threshold number of users that indicate the selected asset matches the identified candidate object and when the identified candidate object has attributes that are at an edge of a parameter range for the selected asset, update the knowledge base by expanding the parameter range associated with the selected asset.

2. The system of claim 1, wherein the instructions that, when executed, cause the processor to identify the candidate object as the selected asset comprise instructions to:

compare a measurement of points in the second dataset for the identified candidate object with a measurement of points parameter for a subset of the assets to yield a first result;

compare a dimension in the second dataset for the identified candidate object with a dimensional parameter for a plurality of the assets to yield a second result; and

identify the identified candidate object as the selected asset based at least in part on the first result and the second result.

3. The system of claim 1, wherein:

the instructions that, when executed, cause the processor to identify the candidate object as the selected asset comprise instructions to;

compare a color value in the second dataset for the identified candidate object with a color parameter for a plurality of the assets to yield a color comparison result, and

identify the identified candidate object as the asset based at least in part on the color comparison result; and

the instructions that, when executed, cause the processor to update the knowledge base comprise instructions to adjust a color parameter associated with the selected asset based on the color value for the identified candidate object.

4. The system of claim 1, wherein the instructions that, when executed, cause the processor to identify the candidate object as the selected asset comprise instructions to:

compare an intensity value in the second dataset for the candidate object with an intensity parameter for a plurality of the assets to yield an intensity comparison result; and

identify the candidate object as the selected asset based at least in part on the intensity comparison result.

5. The system of claim 4, wherein the instructions that, when executed, cause the processor to update the knowledge base comprise instructions to adjust an intensity parameter associated with the selected asset based on the intensity value for the identified candidate object.

6. The system of claim 1, wherein the instructions that, when executed, cause the processor to output an image representing the identified candidate object comprise instructions to:

determine, based on the three-dimensional data for the identified candidate object, a plurality of views of the identified candidate object; and

output, to one or more of the multiple users, a second image representing the identified candidate object, wherein the first image comprises one of the views and the second image comprises a different one of the views.

7. The system of claim 1, wherein the computer-readable medium also comprises instructions to generate a report that identifies a plurality of candidate objects that users identified as assets.

8. The system of claim 1, wherein:

the knowledge base also comprises global positioning system data representing a location of the geographic area; and

the three-dimensional data comprises light detection and ranging system data including distance information, color information, and intensity information for a plurality of points within the geographic area.

9. The system of claim 1, wherein the instructions that, when executed, cause the processor to identify the candidate object as the selected asset comprise instructions to:

identify a group of data points in the second dataset having time stamps that are no more than a first threshold magnitude apart, as well as location values that are no more than a first threshold magnitude apart;

identify, from within the group, a subset of data points having attributes that match parameters of the selected asset.

10. The system of claim 1, wherein the instructions that, when executed, cause the processor to output the first image representing the identified candidate object comprise instructions to:

determine a plurality of components within the identified candidate object;

enhance a first one of the components to distinguish the first component from a second one of the components; and

output, to one or more of the multiple users, the first image so that the first component is visually distinguished form the second component.

17. The method of claim 1, wherein the element of identifying the identified candidate object as the selected asset comprises:

identify a group of data points in the second dataset having time stamps that are no more than a first threshold magnitude apart, as well as location values that are no more than a first threshold magnitude apart;

identify, from within the group, a subset of data points having attributes that match parameters of the selected asset.

11. A method comprising, by a processor:

accessing a knowledge base containing a first dataset, the first dataset comprising parameters that are associated with one or more assets;

accessing a second dataset, the second dataset comprising three-dimensional data representing a geographic area and one or more candidate objects within the geographic area;

analyzing the three-dimensional data to identify one of the candidate objects as a selected one of the assets by comparing the second dataset with one or more of the parameters in the first dataset;

outputting, to multiple users, a first image representing the identified candidate object, a representation of selected asset and a user prompt;

receiving user responses from each of the multiple users indicating whether the selected asset matches the identified candidate object; and

based on receipt of user responses from a threshold number of users that indicate the selected asset matches the identified candidate object and when the identified candidate object has attributes that are at an edge of a parameter range for the selected asset, updating the knowledge base by expanding the parameter range associated with the selected asset.

12. The method of claim 11, wherein the element of identifying the identified candidate object as the selected asset comprises:

comparing a measurement of points in the second dataset for the identified candidate object with a measurement of points parameter for a plurality of the assets to yield a first result;

comparing a dimension value in the second dataset for the identified candidate object with a dimensional parameter for a plurality of the assets to yield a second result; and

identifying the identified candidate object as the selected asset based at least in part on the first result and the second result.

13. The method of claim 11, wherein:

the element of identifying the identified candidate object as the selected asset comprises;

comparing a color value in the second dataset for the identified candidate object with a color parameter for a plurality of the assets to yield a color comparison result, and

identifying the identified candidate object as the selected asset based at least in part on the color comparison result; and

the element of updating the knowledge base comprises adjusting a color parameter associated with the selected asset based on the color value for the identified candidate object.

14. The method claim 11, wherein:

the element of identifying the identified candidate object as the selected asset comprises;

comparing an intensity value in the second dataset for the identified candidate object with an intensity parameter for a plurality of the assets to yield an intensity comparison result, and

identifying the identified candidate object as the selected asset based at least in part on the intensity comparison result; and

the element of updating the knowledge base comprises instructions to adjust an intensity parameter associated with the selected asset based on the intensity value for the identified candidate object.

15. The method claim 11, wherein the element of outputting an image representing the identified candidate object comprises:

determining, based on the three-dimensional data for the selected asset, a plurality of views of the identified candidate object; and

outputting, to one or more of the multiple users, a second image representing the identified candidate object, wherein the first image comprises one of the views and the second image comprises a different one of the views.

16. The method of claim 11, wherein:

the knowledge base also comprises global positioning system data representing a location of the geographic area; and

the three-dimensional data comprises light detection and ranging system data including distance information, color information, and intensity information for a plurality of points within the geographic area.

18. The method of claim 11, wherein element of outputting the first image representing the identified candidate object comprises:

determining, a plurality of components within the identified candidate object;

enhancing a first one of the components to distinguish the first component from a second one of the components; and

outputting, to one or more of the multiple users, the first image so that the first component is visually distinguished from the second component.

19. A non-transitory computer-readable medium containing programming instructions for causing a processor to:

access a knowledge base containing a first dataset, the first dataset comprising parameters that are associated with one or more assets;

access a second dataset, the second dataset comprising three-dimensional data representing a geographic area and one or more candidate objects within the geographic area;

analyze the three-dimensional data to identify one of the candidate objects as a selected one of the assets by comparing the second dataset with one or more of the parameters in the first dataset;

output, to multiple users, a first image representing the identified candidate object and a user prompt;

receive a plurality of user responses from each of the multiple users indicating whether the selected asset matches the identified candidate object; and

based on receipt of user responses from a threshold number of users that indicate the selected asset matches the identified candidate object and when the identified candidate object has attributes that are at an edge of a parameter range for the selected asset, update the knowledge base by expanding the parameter range associated with the selected asset so that either or both one of an intensity parameter and a color parameter is adjusted.

20. The computer-readable medium of claim 19, wherein:

the instructions for causing the processor to identify the identified candidate object as the selected asset comprise instructions to;

comparing a color value in the second dataset for the identified candidate object with a color parameter for a plurality of the assets to yield a color comparison result,

compare an intensity value in the second dataset for the identified candidate object with an intensity parameter for a plurality of the assets yield an intensity comparison result, and

identify the identified candidate object as the selected asset based at least in part on the color comparison result and the intensity comparison result; and

the instructions for causing the processor to update the knowledge base comprise instructions to;

adjust a color parameter associated with the selected asset based on the color value for the identified candidate object, and

adjust an intensity parameter associated with the selected asset based on the intensity value for the identified candidate object.