Person counting method and device for same

• US 9,971,941 B2
• Filed: 10/22/2014
• Issued: 05/15/2018
• Est. Priority Date: 02/24/2014
• Status: Active Grant

• Current Assignee: SK Telecom Co., Ltd.
• Original Assignee: SK Telecom Co., Ltd.
• Inventors: Yang, Seungji
• Primary Examiner(s): Chu, Randolph I
• Application Number:
  US15/108,784
• Publication Number:
  US 20160321507A1
• Time in Patent Office: 1,301 Days
• Field of Search: None
• CPC Class Codes: G06F 18/22 Matching criteria, e.g. proximity measures
• CPC Class Codes: G06T 2207/30196 Human being; Person
• CPC Class Codes: G06T 2207/30242 Counting objects in image
• CPC Class Codes: G06T 7/246 using feature-based methods, e.g. the tracking of corners or segments
• CPC Class Codes: G06T 7/60 Analysis of geometric attributes
• CPC Class Codes: G06V 20/53 Recognition of crowd images, e.g. recognition of crowd congestion
• CPC Class Codes: G06V 40/20 Movements or behaviour, e.g. gesture recognition recognition of facial expressions G06V40/16

Abstract

A method for counting a number of people includes detecting a foreground contour from a current frame inputted, performing a head detection by defining a baseline virtually at a point perpendicular to a center point of a camera-captured area, by defining a search area around the baseline, and by detecting, as as candidate heads, circular contours to be recognized as heads of people at a point where the foreground contour crosses the baseline, tracking the candidate heads respectively in subsequent frames until the candidate heads pass through the search area, and counting up the number of people by the candidate heads that respectively pass through the search area.

Claims

1. A method performed by an image analyzing apparatus for counting a number of people, the image analyzing apparatus comprising one or more application specific integrated circuits (ASICs) to implement a microprocessor to execute the method comprising:

detecting a foreground contour from a current frame inputted;

performing a head detection, comprising;

defining a baseline virtually at a point perpendicular to a center point of a camera-captured area,

defining a search area around the baseline, and

detecting, as candidate heads, circular contours to be recognized as heads of people at a point where the foreground contour crosses the baseline;

tracking the candidate heads respectively in subsequent frames until the candidate heads pass through the search area; and

outputting the number of people by counting up the recognized heads of people based on a tracking list and the tracked candidate heads that respectively pass through the search area,

wherein the ASICs implements the microprocessor to further execute

generating the tracking list containing candidate heads, and

wherein the tracking of the candidate heads comprises;

determining whether a candidate head detected by the tracking is spatially overlapped with at least one of the candidate heads on the tracking list; and

if not spatially overlapped, newly inserting the nonoverlapped candidate heads into the tracking list.

2. The method of claim 1,

wherein the tracking of the candidate heads comprises:

applying a fast template matching method to the candidate heads on the tracking list and thereby determining a point bearing a strong similarity to the current frame as a new center point for each of the candidate heads.

3. The method of claim 1,

wherein the tracking of the candidate heads comprises:

determining the search area for performing a process of matching an arbitrary candidate head on the tracking list with the current frame.

4. The method of claim 1, wherein

the tracking of the candidate heads comprises:

calculating a height (R_h) of the search area based on an upper area (R_u) of the search area with respect to respective candidate heads and a lower area (R_d) of the search area with respect to the respective candidate heads,

setting a width (R_w) of the search area to a value equal to the height (R_h) of the search area, and

determining the search area based on the height (R_h) and the width (R_w) of the search area.

5. The method of claim 4, further comprising generating a tracking list containing candidate heads, and

wherein the tracking of the candidate heads comprises:

calculating the height (R_h) of the search area based on a first weight (w_α

), a second weight (w_β

), an average moving speed (S(t)) of a t-th candidate head on the tracking list, and up-and-down moving directions (Δ

y(t)) with respect to the respective candidate heads.

6. The method of claim 5, wherein the tracking of the candidate heads comprises:

calculating the average moving speed (S(t)) of the t-th candidate head on the tracking list based on a number (N) of total trajectories and x-y coordinates (x_n, y_n) of a candidate head of an n-th trajectory.

7. The method of claim 5, wherein the tracking of the candidate heads comprises:

calculating the up-and-down moving directions (Δ

y(t)) with respect to the respective candidate heads based on a number (N) of total trajectories and a y coordinate (y_n) of a candidate head of an n-th trajectory.

8. The method of claim 4, further comprising generating a tracking list containing candidate heads, and

wherein the tracking of the candidate heads comprises:

calculating the lower area (R_d) of the search area with respect to respective candidate heads based on a second weight (w_β

), the average moving speed (S(t)) of the t-th candidate head on the tracking list, and the upper area (R_u) of the search area with respect to the respective candidate heads.

9. The method of claim 8, wherein the tracking of the candidate heads comprises:

calculating the average moving speed (S(t)) of the t-th candidate head on the tracking list based on a number (N) of total trajectories and x-y coordinates (x_n, y_n) of a candidate head of an n-th trajectory.

10. The method of claim 8, wherein the tracking of the candidate heads comprises:

calculating the height (R_h) of the search area based on a first weight (w_α

), a second weight (w_β

), an average moving speed (S(t)) of a t-th candidate head on the tracking list, and up-and-down moving directions (Δ

y(t)) with respect to the candidate heads.

11. The method of claim 1, wherein the detecting of the foreground contour comprises:

extracting a background contour from a initial frame containing only a background, extracting an entire contour of the current frame, and removing the background contour from the entire contour to acquire the foreground contour.

12. The method of claim 11, wherein the detecting of the foreground contour comprises:

extracting the foreground contour by using an adaptive thresholding process based on a frame integral image.

13. The method of claim 1, wherein the performing of the head detection comprises:

extracting the circular contours by performing a circular Hough transform based on the foreground contour.

14. The method of claim 13, wherein the performing of the head detection comprises:

performing the circular Hough transform by using sizes of the circular contours between a predetermined minimum value and a predetermined maximum value, as prior information.

15. The method of claim 1, wherein the counting up comprises:

counting up the candidate heads one by one passing through the search area by taking account of directions of the circular contours.

16. An image analyzing apparatus comprising one or more application specific integrated circuits (ASICs) to implement:

a foreground contour detecting unit configured to detect a foreground contour from a current frame inputted;

a head detecting unit configured to

define a baseline virtually at a point perpendicular to a center point of a camera-captured area,

define a search area around the baseline, and

detect, as candidate heads, circular contours to be recognized as heads of people at a point where the foreground contour crosses the baseline;

a head tracking unit configured to track the candidate heads respectively in subsequent frames until the candidate heads pass through the search area; and

a counting unit configured to outputting a number of people by counting up the recognized heads of people based on a tracking list and the tracked candidate heads that respectively pass through the search area,

wherein the ASICs further implements the head tracking unit configured to

generate the tracking list containing candidate heads, and

wherein the ASICs further implements the head tracking unit configured to track the candidate heads by

determining whether a candidate head detected by the tracking is spatially overlapped with at least one of the candidate heads on the tracking list; and

inserting, if not spatially overlapped, the nonoverlapped candidate heads into the tracking list.

17. The image analyzing apparatus of claim 16, wherein the ASICs further implements the head tracking unit configured to track the candidate heads by

applying a fast template matching method to the candidate heads on the tracking list and thereby determining a point bearing a strong similarity to the current frame as a new center point for each of the candidate heads.

18. The image analyzing apparatus of claim 16, wherein the ASICs further implements the head tracking unit configured to track the candidate heads by

determining the search area for performing a process of matching an arbitrary candidate head on the tracking list with the current frame.

19. The image analyzing apparatus of claim 16, wherein the ASICs further implements the head tracking unit configured to track the candidate heads by

calculating a height (R_h) of the search area based on an upper area (R_u) of the search area with respect to respective candidate heads and a lower area (R_d) of the search area with respect to the respective candidate heads,

setting a width (R_w) of the search area to a value equal to the height (R_h) of the search area, and

determining the search area based on the height (R_h) and the width (R_w) of the search area.