SYSTEMS AND METHODS FOR EVALUATING ROBUSTNESS

• US 20100086221A1
• Filed: 10/01/2009
• Published: 04/08/2010
• Est. Priority Date: 10/03/2008
• Status: Active Grant

• Current Assignee: 3M Innovative Properties Company (3M Company)
• Original Assignee: 3M Innovative Properties Company (3M Company)
• Inventors: Stankiewicz, Brian J., Misemer, David K., Linzie, Brian L., Casner, Glenn E., Anderson, Nathan J., Brooks, Brian E., Gardner, Timothy J., Canavan, Michael Kelly
• Granted Patent: US 8,442,328 B2
• Time in Patent Office: Days
• US Class Current: 382/224
• CPC Class Codes: G06Q 10/067 Enterprise or organisation modelling
• CPC Class Codes: G06Q 10/101 Collaborative creation, e.g. joint development of products or services
• CPC Class Codes: G06Q 30/0206 Price or cost determination based on market factors
• CPC Class Codes: G06V 10/462 Salient features, e.g. scale invariant feature transforms [SIFT]

Abstract

Systems and methods for evaluating the robustness of objects within a scene or a scene itself.

Claims

1. A computer system comprising:

a visual attention module operative to;

(a) receive scene input;

(b) apply a visual attention model to the scene input; and

,

(c) predict regions within the scene input that tend to attract visual attention; and

,

a robustness assessment module operative to interact with the visual attention module to determine a degree to which at least one of the identified regions is robust or the scene is robust.

2. The system of claim 1, wherein robustness is a measure of the sensitivity of predictions output from a visual attention model to either of, or some combination of:

(a) variability within the visual properties of the scene;

or,

(b) variability within the visual attention model that is applied by the visual attention module.

3. The system of claim 2, wherein the robustness assessment module is further operative to introduce variability into visual properties of the scene.

4. The system of claim 3, wherein the robustness assessment module introduces variability into visual properties of the scene by modifying the scene input.

5. The system of claim 4, wherein modifying the scene input comprises systematically or randomly selecting pixels within the image to modify, and systematically or randomly perturbing the properties of those selected pixels.

6. The system of claim 2, wherein the robustness assessment module is further operative to introduce variability into the visual attention model that is applied by the visual attention module.

7. The system of claim 6, wherein the variability is introduced into the visual attention model by modifying operating parameters of the visual attention model.

8. The system of claim 7, wherein modifying the operating parameters is done systematically or randomly.

9. The system of claim 7, wherein the operating parameters that are modified are those that control the sensitivity of the visual attention model to any of the following:

color, contrast, motion, shape, spatial frequency, orientation.

10. The system of claim 2, wherein the robustness assessment module is further operative to both:

(a) introduce variability into the visual properties of the scene; and

,

(b) introduce variability into the visual attention model that is applied by the visual attention module.

11. The system of claim 10, wherein introducing variability into the visual properties of the scene comprises introducing visual distractors into the scene.

12. The system of claim 10, further comprising:

a database module operative to record how the identified regions change in view of variability introduced into visual properties of a scene, or variability introduced within the visual attention model that is applied by the visual attention module, the database module also operative to record the type of variability.

13. The system of claim 12, wherein the robustness assessment module repeatedly:

(a) introduces variability into the visual properties of the scene thus generating a modified scene, and;

(b) invokes the visual attention module and provides the modified scene as input.

14. The system of claim 12, wherein the robustness assessment module repeatedly:

(a) introduces variability into the visual attention model that is applied by the visual attention module, and;

(b) invokes the visual attention module to use the modified visual attention model.

15. The system of claim 1, further comprising:

a user interface module operative to display information indicative of the degree to which at least one of the identified regions is robust.

16. The system of claim 15, wherein the information comprises a graphical representation of the scene overlaid with heat maps associated with individual regions of the scene, the gradients of the heat maps correlated to the robustness of individual regions of the scene.

17. The system of claim 15, wherein the information comprises a number.

18. The system of claim 1, wherein the robustness assessment module is additionally provided with a duration of time that particular regions are viewable, and additionally uses this information to determine whether the region is robust.

19. The system of claim 1, wherein the visual attention model is an empirical model.

20. The system of claim 2, wherein the robustness assessment module is further operative to quantify the degree of robustness as a number, and provide this number to a computer program.

21. A computer-implemented method comprising:

receiving scene input;

applying a visual attention model to the scene input to predict regions within the scene input that tend to attract visual attention;

changing the scene so as to produce a modified scene;

applying the visual attention model to the modified scene input to predict regions within the scene input that tend to attract visual attention; and

,

based on the regions identified from the application of the visual attention model to the scene input and the modified scene, determining the degree to which one of the identified regions, or the scene input, is robust.

22. The computer-implemented method of claim 21, wherein robust is a measure of the sensitivity of predictions from a visual attention model to variability within the visual properties of the scene.

23. The computer-implemented method of claim 22, wherein changing the scene comprises systematically or randomly selecting pixels within the image to modify, and systematically or randomly perturbing the properties of those selected pixels.

24. The computer-implemented method of claim 22, wherein changing the scene comprises introducing visual distractors into the scene.

25. The computer-implemented method of claim 22, wherein changing the scene comprises modifying regions that comprise objects within the scene.

26. The computer implemented method of claim 22, wherein changing the scene comprises modifying the size, color, or shape of regions that comprise objects within the scene.

27. The computer-implemented method of claim 22, wherein changing the scene comprises adjusting the brightness or contrast of the scene.

28. The computer-implemented method of claim 22, further comprising:

presenting to a user, via a user interface, information indicative of the degree to which one of the identified regions, or the scene input, is robust.

29. The computer-implemented method of claim 28, wherein the information is a number.

30. The computer-implemented method of claim 21, further comprising:

providing a number indicative of the degree to which one of the identified regions, or the scene input, is robust to computer program.

31. A computer-implemented method comprising:

receiving scene input;

applying a visual attention model to the scene input to predict regions within the scene input that tend to attract visual attention;

changing the visual attention model so as to produce a modified visual attention model;

applying the modified visual attention model to the scene input to predict regions within the scene input that tend to attract visual attention; and

,

based on the regions identified from the application of the visual attention model and the modified visual attention model to the scene input, determining the degree to which one of the identified regions, or the scene input, is robust.

32. The computer-implemented method of claim 31, wherein robust is a measure of the sensitivity of predictions from a visual attention model to variability within the visual attention model.

33. The computer-implemented method of claim 32, wherein changing the visual attention model comprises modifying operating parameters of the visual attention model.

34. The computer implemented method of claim 33, wherein the operating parameters are modified randomly or systematically.

35. The computer-implemented method of claim 34, wherein the operating parameters are those that control the sensitivity of the visual attention model to any of, or combination of, the following:

color, contrast, motion, shape, special frequency, and orientation.

36. The computer-implemented method of claim 31, further comprising:

providing output indicative of the determined degree of robustness.

37. The computer-implemented method of claim 36, wherein the output is a number.

38. The computer-implemented method of claim 37, wherein the output is provided to a computer program.