Landmark detection apparatus and method for intelligent system

• US 7,606,416 B2
• Filed: 11/17/2004
• Issued: 10/20/2009
• Est. Priority Date: 11/17/2003
• Status: Expired due to Fees

• Current Assignee: Samsung Electronics Co. Ltd.
• Original Assignee: Samsung Electronics Co. Ltd.
• Inventors: Park, Ki-cheol, Han, Woo-Sup, Kwon, Woong, Roh, Kyung-shik, Oh, Yeon-taek, Shim, Young-bo
• Primary Examiner(s): Bali; Vikkram
• Assistant Examiner(s): YEH, EUENG NAN
• Application Number:
  US10/989,401
• Publication Number:
  US 20050107947A1
• Time in Patent Office: 1,798 Days
• Field of Search: 382/162, 382/171, 701/207
• US Class Current: 382/171
• CPC Class Codes: G05D 1/0246 using a video camera in combination with image processing means
• CPC Class Codes: G06V 10/462 Salient features, e.g. scale invariant feature transforms [SIFT]
• CPC Class Codes: G06V 10/56 relating to colour
• CPC Class Codes: G06V 10/758 Involving statistics of pixels or of feature values, e.g. histogram matching
• CPC Class Codes: G06V 20/10 Terrestrial scenes scenes under surveillance with static cameras G06V20/52; scenes perceived from the exterior of a vehicle G06V20/56; scenes perceived from the interior of a vehicle G06V20/59

Abstract

A landmark detection apparatus and method, the apparatus including a first detection unit that generates N first sample blocks in a first sampling region using a first weighted sampling method according to a first degree of dispersion, and that performs a first landmark detection by comparing a feature of each first sample block a feature of a landmark model, where the first sample region is set to the entirety of a current frame image; and a second detection unit that generates N second sample blocks in a second sampling region using a second weighted sampling method according to a second degree of dispersion, and that performs a second landmark detection by comparing a feature of each second sample block the feature of the landmark model, where the second sampling region is set to an area less than the entirety of the current frame image.

Claims

1. A landmark detection apparatus for an intelligent system, the apparatus comprising:

a first detection unit that generates a group of N first sample blocks in a first sampling region using a first weighted sampling method according to a first degree of dispersion, and that performs a first landmark detection by comparing a feature of each first sample block of the group of N first sample blocks to a feature of a landmark model, wherein the first sampling region is set to the entirety of a frame image; and

a second detection unit that generates a group of N second sample blocks in a second sampling region using a second weighted sampling method according to a second degree of dispersion, and that performs a second landmark detection by comparing a feature of each second sample block of the group of N second sample blocks to the feature of the landmark model, wherein the second sampling region is set to an area less than the entirety of the frame image,

wherein each of the first degree of dispersion and the second degree of dispersion is adaptively adjusted according to an accumulative probability in a current frame image, and then applied to a next frame image;

and wherein the first detection unit and the second detection unit are implemented by a processor.

2. The apparatus of claim 1, wherein the first detection unit:

calculates a matching probability between each first sample block of the group of N first sample blocks and the landmark model based on a comparison of a color histogram of each first sample block of the group of N first sample blocks and a color histogram of the landmark model;

estimates a position of a detected landmark; and

generates another group of N first sample blocks to be applied to a next frame image in a weighted sampling method according to the calculated matching probability.

3. The apparatus of claim 2, wherein the first detection unit updates the color histogram of the landmark model using:

the color histograms of the first sample blocks of the group of N first sample blocks in the frame image;

the calculated matching probabilities between each first sample block of the group of N first sample blocks and the landmark model, and a feature of the landmark model used in the frame image.

4. The apparatus of claim 1, wherein the second detection unit:

calculates a matching probability between each second sample block of the group of N second sample blocks and the landmark model based on a comparison of a color histogram of each second sample block of the group of N second sample blocks and a color histogram of the landmark model;

estimates a position of a detected landmark; and

generates another group of N second sample blocks to be applied to a next frame image in a weighted sampling method according to the calculated matching probability.

5. The apparatus of claim 4, wherein the second detection unit updates the color histogram of the landmark model using:

the color histograms of the second sample blocks of the group of N second sample blocks in the frame image;

the calculated matching probabilities between each second sample block of the group of N second sample blocks and the landmark model; and

a feature of the landmark model used in the frame image.

6. The apparatus of claim 1, wherein the first degree of dispersion is greater than the second degree of dispersion.

7. The apparatus of claim 1, wherein the second sampling region is set closely around a detected landmark detected in the first landmark detection.

8. A landmark detection method for an intelligent system performed on a processor, the method comprising using the processor to perform a process comprising:

generating a group of N first sample blocks in a first sampling region using a first weighted sampling method according to a first degree of dispersion, and performing a first landmark detection by comparing a feature of each first sample block of the group of N first sample blocks to a feature of a landmark model, wherein the first sampling region is set to the entirety of a frame image; and

generating a group of N second sample blocks in a second sampling region using a second weighted sampling method according to a second degree of dispersion, and performing a second landmark detection by comparing a feature of each second sample block of the group of second sample blocks to the feature of the landmark model, wherein the second sampling region is set to an area less than the entirety of the frame image,

wherein each of the first degree of dispersion and the second degree of dispersion is adaptively adjusted according to an accumulative probability in a current frame image, and then applied to a next frame image.

9. The method of claim 8, wherein the first degree of dispersion is greater than the second degree of dispersion.

10. The method of claim 8, wherein the first degree of dispersion or the second degree of dispersion is increased or decreased according to a result of comparing the accumulative probability in the frame image with a predetermined threshold value, and applied to the next frame image.

11. The method of claim 8, wherein the first degree of dispersion or the second degree of dispersion is determined by a result of comparing an actual degree of dispersion in the frame image with a predetermined highest degree of dispersion, and applied to the next frame image.

12. The method of claim 8, wherein, after generating the group of N first sample blocks and the group of N second sample blocks, the feature of the landmark model is updated using:

the features of the first sample blocks of the group of N first sample blocks and the second sample blocks of the group of N second sample blocks in the frame image;

matching probabilities for the first sample blocks of the group of N first sample blocks and the second sample blocks of the group of N second sample blocks and the landmark model, and feature amounts of the landmark model used in the frame image.

13. The method of claim 8, wherein the feature is a color histogram.

14. The method of claim 8, further comprising:

generating color histograms of each first sample block of the group of N first sample blocks;

generating a color histogram of the landmark model;

calculating a similarity between the color histogram of each first sample block of the group of N first sample blocks and the color histogram of the landmark model; and

calculating a matching probability between each first sample block of the group of N first sample blocks and the landmark model based on the calculated similarity.

15. The method of claim 14, further comprising updating the color histogram of the landmark model using:

the color histograms of the first sample blocks of the group of N first sample blocks in the frame image, the calculated matching probabilities between each first sample block of the group of N first sample blocks and the landmark model, and the feature of the landmark model used in the frame image.

16. The method of claim 14, further comprising adjusting the first degree of dispersion using the accumulative probability of the current frame image.

17. The method of claim 8, further comprising:

generating color histograms of each second sample block of the group of N second sample blocks;

generating a color histogram of the landmark model;

calculating a similarity between the color histogram of each second sample block of the group of N second sample blocks and the color histogram of the landmark model;

calculating a matching probability between each second sample block of the group of N second sample blocks and the landmark model based on the calculated similarity.

18. The method of claim 17, further comprising updating the color histogram of the landmark model using:

the color histograms of the second sample blocks of the group of N second sample blocks in the frame image, the calculated matching probabilities between each first sample block of the group of N second sample blocks and the landmark model, and the feature of the landmark model used in the frame image.

19. The method according to claim 8, wherein the second sampling region is set closely around a detected landmark detected in the first landmark detection.

20. A tangible computer-readable recording medium for storing a program for performing a landmark detection method for an intelligent system, the method comprising:

generating a group of N first sample blocks in a first sampling region using a first weighted sampling method according to a first degree of dispersion, and performing a first landmark detection by comparing a feature of each first sample block of the group of N first sample blocks to a feature of a landmark model, wherein the first sampling region is set to the entirety of a current frame image; and

generating a group of N second sample blocks in a second sampling region using a second weighted sampling method according to a second degree of dispersion, and performing a second landmark detection by comparing a feature of each second sample block of the group of second sample blocks to the feature of the landmark model, wherein the second sampling region is set to an area less than the entirety of the current frame image,

wherein each of the first degree of dispersion and the second degree of dispersion is adaptively adjusted according to an accumulative probability in a current frame image, and then applied to a next frame image.

21. The computer-readable recording medium according to claim 20, wherein the second sampling region is set closely around a detected landmark detected in the first landmark detection.