Method for tracking moving object

• US 8,243,990 B2
• Filed: 12/31/2008
• Issued: 08/14/2012
• Est. Priority Date: 11/17/2008
• Status: Active Grant

• Current Assignee: Industrial Technology Research Institute
• Original Assignee: Industrial Technology Research Institute
• Inventors: Huang, Chung-Hsien, Shih, Ming-Yu
• Primary Examiner(s): Park, Chan S
• Assistant Examiner(s): YENTRAPATI, AVINASH
• Application Number:
  US12/346,845
• Publication Number:
  US 20100124358A1
• Time in Patent Office: 1,322 Days
• Field of Search: 382/100, 382/103, 382/107, 382/118, 382/115, 382/282, 382/286, 382/291
• US Class Current: 382/103
• CPC Class Codes: G06T 2207/10016 Video; Image sequence
• CPC Class Codes: G06T 2207/10024 Color image
• CPC Class Codes: G06T 2207/20021 Dividing image into blocks, subimages or windows
• CPC Class Codes: G06T 2207/30196 Human being; Person
• CPC Class Codes: G06T 2207/30232 Surveillance
• CPC Class Codes: G06T 7/251 involving models
• CPC Class Codes: G06T 7/254 involving subtraction of images
• CPC Class Codes: G06T 7/277 involving stochastic approaches, e.g. using Kalman filters
• CPC Class Codes: G06V 20/52 Surveillance or monitoring of activities, e.g. for recognising suspicious objects recognising microscopic objects G06V20/69
• CPC Class Codes: G06V 40/23 Recognition of whole body movements, e.g. for sport training

Abstract

A method for tracking a moving object is provided. The method detects the moving object in a plurality of continuous images so as to obtain space information of the moving object in each of the images. In addition, appearance features of the moving object in each of the images are captured to build an appearance model. Finally, the space information and the appearance model are combined to track a moving path of the moving object in the images. Accordingly, the present invention is able to keep tracking the moving object even if the moving object leaves the monitoring frame and returns again, so as to assist the supervisor in finding abnormal acts and making following reactions.

Claims

1. A method for tracking a moving object, comprising:

detecting the moving object in a plurality of continuous images to obtain space information of the moving object in each of the images;

tracking the moving path of the moving object according to the space information, and accumulating a elapsed time of the moving object staying in the images;

capturing an appearance feature of the moving object in each of the images to build an appearance model of the moving object if the elapsed time is greater than a first predetermined time;

associating the appearance model of the moving object to a plurality of appearance models within a database to judge whether the appearance model of the moving object is recorded in the database, the associating comprising;

calculating a distance between the appearance model of the moving object and each of the appearance models in the database, judging whether the distance is less than a threshold, and renewing the appearance model in the database according to the appearance model of the moving object if the distance of the appearance model is less than the threshold, the calculating comprising;

calculating a first distance between the appearance models built at two different time points based on the moving object in the images, so as to build a first distance distribution;

calculating a second distance between the appearance models of two moving objects in the images, so as to build a second distance distribution; and

obtaining a boundary of the first distance distribution and the second distance distribution to serve as the threshold for distinguishing the appearance model; and

combining the space information and the appearance model of the moving object to track a moving path of the moving object in the images.

2. The method for tracking a moving object as claimed in claim 1, wherein the step of detecting the moving object in a plurality of continuous images further comprises:

subtracting a background image from the images to detect the moving object.

3. The method for tracking a moving object as claimed in claim 1, wherein after the step of detecting the moving object in a plurality of continuous images, the method further comprises:

marking a plurality of connection regions in the images; and

estimating a minimum rectangular region capable of encircling the connection regions.

4. The method for tracking a moving object as claimed in claim 3, wherein the steps of detecting the moving object in a plurality of continuous images further comprise:

judging whether the moving object is a tracking object; and

filtering the moving object not belonged to the tracking object.

5. The method for tracking a moving object as claimed in claim 4, wherein the steps of judging whether the moving object is the tracking object comprise:

judging whether an area of the rectangular region is greater than a first predetermined value; and

judging the moving object encircled by the rectangular region to be the tracking object if the area is greater than the first predetermined value.

6. The method for tracking a moving object as claimed in claim 4, wherein the steps of judging whether the moving object is the tracking object comprise:

judging whether an aspect ratio of the rectangular region is greater than a second predetermined value; and

judging the moving object encircled by the rectangular region to be the tracking object if the aspect ratio is greater than the second predetermined value.

7. The method for tracking a moving object as claimed in claim 3, wherein the steps of capturing the appearance feature of the moving object in each of the images to build the appearance model of the moving object comprise:

dividing the rectangular region into a plurality of blocks, and capturing a color distribution of each of the blocks;

obtaining a median of the color distribution from each of the blocks according to a recursive method to build a binary tree for describing the color distribution; and

selecting the color distributions of branches of the binary tree to serve as a feature vector of the appearance model of the moving object.

8. The method for tracking a moving object as claimed in claim 1, wherein the moving object is a pedestrian.

9. The method for tracking a moving object as claimed in claim 7, wherein the step of dividing the rectangular region into the blocks comprises dividing the rectangular region into a head block, a body block and a lower limb block according to a proportion of 2:

4;

4.

10. The method for tracking a moving object as claimed in claim 9, wherein the step of capturing the color distribution of each of the blocks comprises neglecting the color distribution of the head block.

11. The method for tracking a moving object as claimed in claim 10, wherein the color distribution comprises a color feature in a RGB color space or in a hue, saturation, illumination (HSI) color space.

12. The method for tracking a moving object as claimed in claim 10, wherein the steps of combining the space information and the appearance model of the moving object to track the moving path of the moving object in the images comprise:

calculating a spatial related prior probability of the corresponding moving object in two adjacent images according to the space information;

calculating a likelihood of the corresponding moving object in the two adjacent ages according to the appearance information; and

combining the prior probability and the likelihood to a Bayesian tracker to judge the moving path of the moving object in the adjacent images.

13. The method for tracking a moving object as claimed in claim 10, wherein the steps of

associating the appearance model of the moving object to a plurality of appearance models within the database to judge whether the appearance model of the moving object is recorded in the database comprise:

only recording the elapsed time of the moving object into the database if the appearance model of the moving object is already recorded in the database; and

recording the elapsed time and the appearance model of the moving object into the database if the appearance model of the moving object is not recorded in the database.

14. The method for tracking a moving object as claimed in claim 13, wherein the step of renewing the appearance model in the database according to the appearance model of the moving object comprises:

selecting one of the appearance models with the distance being less than the threshold and most similar to the appearance model of the moving object to renew the appearance model in the database.

15. The method for tracking a moving object as claimed in claim 13, wherein the steps of building the first distance distribution and the second distance distribution comprise:

respectively calculating a mean and a standard deviation of the first distance distribution and the second distance distribution; and

representing data of the first distance distribution and the second distance distribution by a normal distribution according to the mean and the standard deviation, so as to build the first distance distribution and the second distance distribution.

16. The method for tracking a moving object as claimed in claim 10, wherein after the step of recording the elapsed time and the appearance model of the moving object into the database, the method further comprises:

analysing a time sequence of the moving object in the database to determine whether the moving object is accorded to a wander event.

17. The method for tracking a moving object as claimed in claim 16, wherein the steps of analysing the time sequence of the moving object in the database to determine whether the moving object is accorded to a wander event comprise:

judging whether a time for the moving object continuously appearing in the images is greater than a second predetermined time; and

judging the moving object to be accorded to the wander event if the time for the moving object continuously appearing in the images is greater than the second predetermined time.

18. The method for tracking a moving object as claimed in claim 16, wherein the steps of analysing the time sequence of the moving object in the database to determine whether the moving object is accorded to a wander event comprise:

judging whether a time interval for the moving object leaving the images is less than a third predetermined time; and

judging the moving object to be accorded to the wander event if the time interval for the moving object leaving the images is less than the third predetermined time.