Method and device for the detection of moving objects in a video image sequence

• US 20130108109A1
• Filed: 11/01/2012
• Published: 05/02/2013
• Est. Priority Date: 11/02/2011
• Status: Active Grant

• Current Assignee: Airbus Defence and Space GmbH
• Original Assignee: EADS Deutschland GmbH (Airbus SE)
• Inventors: STAHL, Christoph, SEIDEL, Heiko, LEUCK, Holger
• Granted Patent: US 9,014,426 B2
• Time in Patent Office: Days
• US Class Current: 382/103
• CPC Class Codes: G06T 7/254 involving subtraction of images

Abstract

Techniques for the detection of moving objects in a video image sequence are provided. Distinctive image variation points in a difference image are determined. An image variation threshold value is established and image variation points are determined as those points in the difference image in which the absolute image brightness value of which exceeds the image variation threshold value. The quality of the image variation points are determined based on at least one predetermined quality criterion. If the quality criterion is met, the determined image variation points determined are determined as the distinctive image variation points determined. Otherwise, the establishment of an image variation threshold value and determination of the quality of image variation points is repeated with a changed image variation threshold value.

Claims

1. A method for the detection of moving objects in a video image sequence, comprising the steps:

a) determining, by a processor, distinctive feature points in one image of a pair of consecutive images of the video image sequence;

b) determining, by the processor, a mathematical transformation for imaging one of the two images of the pair of images onto the other of the two images of the pair of images, using the distinctive feature points determined in step a);

c) determining, by the processor, a difference image as a difference between the images of the pair of images transformed onto one another by means of transformation determined in step b);

d) determining, by the processor, distinctive image variation points in the difference image determined in step c);

e) determining, by the processor, object points from the distinctive image variation points determined in step d), wherein step d) comprises the following steps

d1) establishing an image variation threshold value and determining image variation points as those points in the difference image determined in step c) with an absolute image brightness value that exceeds the image variation threshold value,

d2) analyzing the quality of the image variation points determined in step d1) based on at least one predetermined quality criterion,

d3) if the quality criterion is met, establishing the image variation points determined in step d1) as the distinctive image variation points determined in step d), and otherwise repeating steps d1) and d2) with a different image variation threshold value.

2. The method according to claim 1, wherein step a) comprises the following steps:

a1) detecting feature points in one of the two images of the pair of images, respectively in a vicinity of grid points of a pre-established grid that is laid over the one of the two images of the pair of images,

a2) determining the distinctive feature points as a selection of the feature points detected in step a1) using at least one predetermined selection criterion.

3. The method according to claim 1, wherein step b) comprises the following steps:

b1) determining displacement vectors for each of the distinctive feature points determined in step a) by determining corresponding distinctive feature points in the other of the two images of the pair of images,

b2) determining displacement vectors to be used further as a selection of the displacement vectors determined in step b1), based on at least one predetermined selection criterion,

b3) calculating the mathematical transformation on the basis of the displacement vectors selected in step b2).

4. The method according to claim 1, wherein step c) comprises the following steps:

c1) applying the mathematical transformation onto one another the two images of the pair of images,

c2) determining the difference image by a pixel-by-pixel subtraction of image brightness values of the two images.

5. The method according to claim 4, wherein the image variation threshold value initially established in step d1) is calculated depending on the image brightness values occurring in the difference image.

6. The method according to claim 1, wherein in step d2) the number of image variation points in the form of isolated pixels in the difference image not exceeding a predetermined maximum number is used as the quality criterion.

7. The method according to claim 1, wherein the quality criterion used in step d2), is changeable by receipt of a user input.

8. The method according to claim 1, wherein in step e) the object points are determined respectively as image regions of image variation points lying closely together determined in step d).

9. A device for the detection of moving objects in a video image sequence, comprising:

an imaging sensor device configured to generate the video image sequence; and

a program-controlled computer device configured to

a) determine distinctive feature points in one image of a pair of consecutive images of the video image sequence;

b) determine a mathematical transformation for imaging one of the two images of the pair of images onto the other of the two images of the pair of images, using the distinctive feature points determined in a);

c) determine a difference image as a difference between the images of the pair of images transformed onto one another by means of transformation determined in b);

d) determe distinctive image variation points in the difference image determined in c);

e) determine object points from the distinctive image variation points determined in d),

wherein the program-controlled computer device is configured to performed the determination in d) by

d1) establishing an image variation threshold value and determining image variation points as those points in the difference image determined in step c) with an absolute image brightness value that exceeds the image variation threshold value,

d2) analyzing the quality of the image variation points determined in step d1) based on at least one predetermined quality criterion,

d3) if the quality criterion is met, establishing the image variation points determined in step d1) as the distinctive image variation points determined in step d), and otherwise repeating steps d1) and d2) with a different image variation threshold value.