Method for determining the length of a queue

• US 9,424,474 B2
• Filed: 06/13/2013
• Issued: 08/23/2016
• Est. Priority Date: 06/20/2012
• Status: Active Grant

• Current Assignee: Xovis AG
• Original Assignee: Xovis AG
• Inventors: Studer, Christian, Herrli, Markus, Wuethrich, Martin, Studer, David
• Primary Examiner(s): Chu, Randolph I
• Application Number:
  US14/409,881
• Publication Number:
  US 20150324647A1
• Time in Patent Office: 1,167 Days
• Field of Search: None
• US Class Current: 1/1
• CPC Class Codes: G06T 1/60 Memory management
• CPC Class Codes: G06T 5/40 using histogram techniques
• CPC Class Codes: G06T 7/292 Multi-camera tracking
• CPC Class Codes: G06T 7/90 Determination of colour characteristics
• 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 40/103 Static body considered as a whole, e.g. static pedestrian or occupant recognition

Abstract

Method for determining the length of a queue of objects in a predefined region having at least one entrance and at least one exit, in the course of which errors in the acquisition of objects entering or exiting the region are corrected during the determination of the length of the queue. In a first step, a specific entry signature E of each object entering the predefined region through the at least one entrance is determined with the aid of at least one first image sensor. Thereafter, the specific entry signature E is stored in an entry list of a calculation unit, each entry signature E being provided with an index value i reflecting the temporal sequence of the entries. In addition, a value L reflecting the length of the queue of the objects is increased by one. Furthermore, a specific exit signature A of an object exiting the predefined region through the at least one exit is determined with the aid of at least one second image sensor, the specific exit signature A being stored in an exit list of a calculation unit as exit signature A₀ with index value j=0. The index value of exit signatures A_j already present in the exit list is respectively increased by one, and exit signatures A_j which have an index value j that is greater than a predeterminable value M are deleted from the exit list. Subsequently, similarity values C_n for −N≦n≦+N are calculated with the aid of the calculation unit, N being a predeterminable value which defines a specific number N of entry signatures of objects which, before or after the object with the entry signature E_L which next exits the predefined region on the basis of the length of the queue L, will exit or have already exited the predefined region. In the process, similarity metrics U_mn are respectively calculated between an exit signature A_m and an entry signature E_L+m+n, M running from 0 to M, and added to C_n. The value L is decremented by the value one. Subsequently, the similarity values C_n are compared with the similarity value C₀. The value L is corrected by adding the index value x given the presence of a value C_x with index value n=x that indicates a higher similarity than the similarity value C₀.

Claims

1. Method for determining the length of a queue of objects in a predefined region having at least one entrance and at least one exit, comprising the steps of:

a) determining with the aid of at least one first image sensor a specific entry signature E of each object entering the predefined region through the at least one entrance;

b) storing the specific entry signature E in an entry list of a calculation unit, each entry signature E being provided with an index value i reflecting the temporal sequence of the entries, and a value L reflecting the length of the queue of the objects being increased by one;

c) determining with the aid of at least one second image sensor a specific exit signature A of an object exiting the predefined region through the at least one exit;

d) storing the specific exit signature A in an exit list of a calculation unit as exit signature A₀ with index value j=0, the index value of exit signatures A_j already present in the exit list respectively being increased by one, and the exit signatures A_j which have an index value j that is greater than a predetermined value M being deleted from the exit list;

e) calculating similarity values C_n for −

N≦

n≦

+N with the aid of the calculation unit, N being a predetermined value which defines a specific number N of entry signatures of objects which, before or after the object with the entry signature E_L whose exit is next expected on the basis of the length of the queue L, will exit or have already exited the predefined region, similarity metrics U_mn respectively being calculated between an exit signature A_m and an entry signature E_L+m+n, m running from 0 to M, and added to C_n;

f) decrementing the value L by the value one;

g) correcting the value L by comparing the similarity values C_n with the similarity value C₀, and correcting the value L by adding the index value x given the presence of a value C_x with index value n=x that indicates a higher similarity than the similarity value C₀.

2. Method according to claim 1, wherein each similarity value C_n is added to a corresponding histogram value K_n of a histogram list, and the new histogram value K_n is stored in the histogram list, the value L being corrected only once a histogram value K_x with index value n=x indicates a higher similarity than the histogram value K₀, by correcting the value L by adding the index value x.

3. Method according to claim 1, wherein the calculation of the similarity values C_n and the correction of the value L are performed not after each acquired exit signature A, but respectively after a predetermined number of acquired exit signatures A.

4. Method according to claim 1, wherein a correction of the value L is performed only if a similarity value C_x or a histogram value K_x indicates a higher similarity which is at least higher by a predetermined value Q than the similarity value C₀ or the histogram value K₀, respectively.

5. Method according to claim 1, wherein the objects are people.

6. Method according to claim 1, wherein entry signatures E_i and exit signatures A acquired by the image sensors are color signatures of the objects.

7. Method according to claim 6, wherein the similarity metrics U_mn are calculated using a distance vector between the exit signature A_m and the entry signature E_i in a multidimensional color space.

8. Method according to claim 1, wherein the similarity metrics U_mn are determined by an earth-movers-distance algorithm.

9. Method according to claim 1, wherein on the basis of the determined length of the queue L of the objects in the predefined region an estimated dwell time is calculated for objects newly entering the predefined region.

10. Method according to claim 1, wherein in addition to the entry signature E_i the entry time T_i is stored for each object entering the predefined region, and an average dwell time of the objects in the predefined region is calculated using the stored entry times T_i and the exit times of the objects.

11. Method according to claim 1, wherein differences of the entry signatures E_i and the exit signatures A_m are compensated by a correction algorithm on the basis of different lighting conditions at the at least one entrance and at the at least one exit.

12. Device for determining the length of a queue of objects in a predefined region having at least one entrance and at least one exit in accordance with a method according to claim 1, comprising at least two image sensors, the image sensors being arranged in such a way that at least one first image sensor detects objects entering the predefined region through the at least one entrance, and at least one second image sensor detects objects exiting the predefined region through the at least one exit, and comprising at least one calculation unit which calculates the length of the queue L of the objects in the predefined region using the detected entering and exiting objects, the calculation unit or the image sensors being configured in such a way that a specific signature is determined for each entering and exiting object.

13. Device according to claim 12, wherein the image sensors are digital cameras.

14. Device according to claim 12, wherein the at least one calculation unit is integrated in at least one of the image sensors.

15. Use of a method according to claim 1 to operate a device in order to determine the length L of a queue of objects in a predefined region.

16. Method according to claim 3, wherein the calculation of the similarity values C_n and the correction of the value L are performed after five, ten or fifteen acquired exit signatures A.

17. Method according to claim 4, wherein Q indicates at least a 1.5 times higher similarity.

18. Method according to claim 6, wherein entry signatures E_i and exit signatures A acquired by the image sensors are color mean values of the objects.

19. Method according to claim 9, wherein the estimated dwell time is output by a display device at the entrance.

20. Method according to claim 10, wherein the estimated dwell time is output by a display device at the entrance.