Method and system for detecting moving objects

Patent metadata

Field	Value
Publication number	US-9852341-B2
Application number	US-201314765546-A
Country	US
Kind code	B2
Filing date	Feb 4, 2013
Priority date	Feb 4, 2013
Publication date	Dec 26, 2017
Grant date	Dec 26, 2017

Abstract

A moving objects detection method is disclosed. The method may include: identifying a plurality of feature points based on a plurality of video frames; selecting from the plurality of feature points to form a first and a second groups of feature points based on correlations between the plurality of feature points; and identifying in at least one video frame two segments based on the first and the second groups of feature points, respectively, as detected moving objects, where a correlation between two feature points may include a distance component and a movement difference component, where the distance component is related to a distance between the two feature points, and the movement difference component is related to a difference between corresponding movements of the two feature points. A moving objects detection system is also provided.

First claim

We claim:

1. A method for detecting moving objects, comprising:
identifying a plurality of feature points based on a plurality of video frames;
selecting from the plurality of feature points first and second groups of feature points based on correlations between the plurality of feature points;
identifying in at least one video frame two segments based on the first and the second groups of feature points, respectively, as detected moving objects, where a correlation between two feature points comprises a distance component and a movement difference component, where the distance component is related to a distance between the two feature points, and the movement difference component is related to a difference between corresponding movements of the two feature points;
determining that the first and the second groups have a common feature point;
identifying a first base point from the first group that represents the feature points of the first group;
identifying a second base point from the second group that represents the feature points of the second group; and
classifying the common feature point into the first group in response to determining that the first base point of the first group has a larger correlation with the common feature point than the second base point of the second group.

2. The method of claim 1 , where a correlation between two feature points is calculated as:

C
12

=

exp
⁡

(

-

ds
12

σ
dis

-

dv
12

σ
v

)

where C 12 represents a correlation between two feature points P 1 and P 2 ;
ds 12 represents a distance between the two feature points;
dv 12 represents a movement difference of the two feature points; and
σ dis and σ v represent statistical variances of distance and movement, respectively.

3. The method of claim 2 , where, if ds 12 is greater than σ dis , or if dv 12 is greater than two times of σ v , the correlation C 12 is set to zero.

4. The method of claim 3 , where σ dis is set to a value in the range of √{square root over (W 2 +H 2 )}/15 to √{square root over (W 2 +H 2 )}/10, where W represents a width of the video frames, and H represents a height of the video frames.

5. The method of claim 3 , where σ v is set to a value in the range of 6 to 10.

6. The method of claim 1 wherein:
the first base point from the first group of feature points has the greatest sum of correlations with the other feature points of the first group; and
the second base point from the second group of feature points has the greatest sum of correlations with the other feature points of the second group.

7. The method of claim 1 further comprising:
calculating for each feature point of the first group a sum of correlations between the feature point and the other feature points of the first group;
identifying from the first group a predetermined proportion of feature points having the greatest sums of correlations to obtain a third group of feature points;
calculating for each feature point of the third group a sum of correlations between the feature point and the other feature points of the third group;
identifying for the first group a first base point which has the greatest sum of correlations with the other feature points of the third group;
calculating for each feature point of the second group a sum of correlations between the feature point and the other feature points of the second group;
identifying from the second group the predetermined proportion of feature points having the greatest sums of correlations to obtain a fourth group of feature points;
calculating for each feature point of the fourth group a sum of correlations between the feature point and the other feature points of the fourth group;
identifying for the second group a second base point which has the greatest sum of correlations with the other feature points of the fourth group; and
classifying the common feature point into one of the first and the second groups whose base point has a larger correlation with the common feature point to obtain updated first and second groups of feature points.

8. The method of claim 7 , where the predetermined proportion is 50%.

9. The method of claim 1 , further comprising:
identifying a dominant plane using RANSAC based homography;
calculating a fundamental matrix based on feature points on the dominant plane;
determining that movements of the first and the second groups of feature points conform to the fundamental matrix; and
classifying an object identified based on the group of feature points as a static object.

10. A moving objects detection system comprising:
a camera; and
a processing device configured to:
identify a plurality of feature points based on a plurality of video frames captured by the camera;
select from the plurality of feature points first and second groups of feature points based on correlations between the plurality of feature points;
identify in at least one video frame two segments based on the first and the second groups of feature points, respectively, as detected moving objects, where a correlation between two feature points comprises a distance component and a movement difference component, where the distance component is related to a distance between the two feature points, and the movement difference component is related to a difference between corresponding movements of the two feature points;
determine that the first and the second groups have a common feature point;
identify a first base point from the first group that represents the feature points of the first group;
identify a second base point from the second group that represents the feature points of the second group; and
classify the common feature point into the first group in response to determining that the first base point of the first group has a larger correlation with the common feature point than the second base point of the second group.

11. The moving objects detection system of claim 10 , where a correlation between two feature points is calculated as:

C
12

=

exp
⁡

(

-

ds
12

σ
dis

-

dv
12

Assignees

• Harman Int Ind
• Harman Int Industries Incorporation

Inventors

• Zhou Yun
• Hong Chuyang
• Zhang Yankun
• Weyrich Norman

Classifications

• G06T2207/30252

  Vehicle exterior; Vicinity of vehicle · CPC title

• G06T2207/30261

  Obstacle · CPC title

• G06T2207/20076

  Probabilistic image processing · CPC title

• G06T2207/10016

  Video; Image sequence · CPC title

• G06T2207/30196

  Human being; Person · CPC title