Pedestrian tracking and counting method and device for near-front top-view monitoring video

• US 20150317797A1
• Filed: 09/03/2013
• Published: 11/05/2015
• Est. Priority Date: 11/28/2012
• Status: Active Grant

• Current Assignee: Xi'An Zhongxing New Software Co., Ltd. (ZTE Corporation)
• Original Assignee: ZTE Corporation
• Inventors: Lu, Ping, Zeng, Gang, Lu, Jian
• Granted Patent: US 9,773,317 B2
• Time in Patent Office: Days
• US Class Current: 1/1
• CPC Class Codes: G06T 2207/10016 Video; Image sequence
• CPC Class Codes: G06T 2207/20224 Image subtraction
• CPC Class Codes: G06T 2207/30196 Human being; Person
• CPC Class Codes: G06T 2207/30232 Surveillance
• CPC Class Codes: G06T 2207/30242 Counting objects in image
• CPC Class Codes: G06T 7/11 Region-based segmentation
• CPC Class Codes: G06T 7/174 involving the use of two or more images
• CPC Class Codes: G06T 7/194 involving foreground-background segmentation
• CPC Class Codes: G06T 7/251 involving models
• CPC Class Codes: G06T 7/62 of area, perimeter, diameter or volume
• CPC Class Codes: G06T 7/66 of image moments or centre of gravity
• CPC Class Codes: G06V 20/53 Recognition of crowd images, e.g. recognition of crowd congestion

Abstract

Provided are a pedestrian tracking and counting method and device for a near-front top-view monitoring video, wherein the method includes that a video image under a current monitoring scene is acquired, the acquired video image is compared with a background image, and when it is determined that the video image is a foreground image, each blob in the foreground image is segmented and combined to acquire a target blob representing an individual pedestrian, and tracking and counting are performed according to the centre-of-mass coordinate of each target blob in a detection area to acquire the number of pedestrians under the current monitoring scene. Thus the accuracy of a counting result can be improved.

Claims

1. A pedestrian tracking and counting method for a near-front top-view monitoring video, comprising:

acquiring a video image under a current monitoring scene;

comparing the video image with a background image, and when it is determined that the video image is a foreground image, segmenting and combining each blob in the foreground image to acquire a target blob representing an individual pedestrian;

performing tracking and counting according to a centre-of-mass coordinate of each target blob in a detection area to acquire the number of pedestrians under the current monitoring scene.

2. The method according to claim 1, wherein segmenting each blob in the foreground image comprises:

segmenting a blob having a width of a circumscribed rectangle of the blob greater than twice a threshold T_W of a width of an individual by equal width according to T_W;

segmenting a blob having a height of a circumscribed rectangle of the blob greater than twice a threshold T_H of a height of an individual by equal height according to T_H;

deleting a blob having an area smaller than a predetermined threshold, wherein the blob is acquired after the segmenting.

3. The method according to claim 1, wherein combining each blob in the foreground image comprises:

combining two blobs having a centre-of-mass horizontal distance not exceeding T_W and a centre-of-mass longitudinal distance not exceeding T_H;

dividing a width of a circumscribed rectangle of each acquired combined blob by T_W and then performing rounding to acquire N_W, and dividing a height of the circumscribed rectangle by T_H and then performing rounding to acquire N_H;

determining whether N_W and N_H of each combined blob are smaller than 2 respectively;

if greater than 2, then splitting the combined blob;

if smaller than 2, then retaining the combined blob; and

repeating the three foregoing steps until all blobs are combined.

4. The method according to claim 1, after it is determined that the video image is a foreground image, the method further comprising:

updating a background image and a background amplitude with different updating rate coefficients according to an image of a current frame.

5. The method according to claim 1, wherein performing tracking comprises:

tracking each target blob in the detection area using a nearest proximity matching algorithm, wherein a cost function applied by the nearest proximity matching algorithm is a Euclidean distance.

6. A pedestrian tracking and counting device for a near-front top-view monitoring video, comprising:

a video image acquiring unit, configured to acquire a video image under a current monitoring scene;

a target blob acquiring unit, configured to compare the video image with a background image, and to, when it is determined that the video image is a foreground image, segment and combine each blob in the foreground image to acquire a target blob representing an individual pedestrian; and

a pedestrian number counting unit, configured to perform tracking and counting according to a centre-of-mass coordinate of each target blob in a detection area to acquire the number of pedestrians under the current monitoring scene.

7. The device according to claim 6, wherein the target blob acquiring unit is configured to segment a blob having a width of a circumscribed rectangle of the blob greater than twice a threshold T_W of a width of an individual by equal width according to T_W;

segment a blob having a height of a circumscribed rectangle of the blob greater than twice a threshold T_H of a height of an individual by equal height according to T_H;

delete a blob having an area smaller than a predetermined threshold, wherein the blob is acquired after the segmenting.

8. The device according to claim 6, wherein the target blob acquiring unit is further configured to combine two blobs having a centre-of-mass horizontal distance not exceeding T_W and a centre-of-mass longitudinal distance not exceeding T_H;

divide a width of a circumscribed rectangle of each acquired combined blob by T_W and then perform rounding to acquire N_W, and divide the height of the circumscribed rectangle by T_H and then perform rounding to acquire N_H;

determine whether N_W and N_H of each combined blob are smaller than 2 respectively;

if greater than 2, then split the combined blob;

if smaller than 2, then retain the combined blob; and

repeat the three foregoing steps until all blobs are combined.

9. The device according to claim 6, further comprising:

an updating unit, configured to update a background image and a background amplitude with different updating rate coefficients according to an image of a current frame.

10. The device according to claim 6, wherein the pedestrian number counting unit is configured to track each target blob in the detection area using a nearest proximity matching algorithm, wherein a cost function applied by the nearest proximity matching algorithm is a Euclidean distance.