Device, method, and computer for image-based counting of objects passing through a counting section in a specified direction

• US 9,418,300 B2
• Filed: 11/20/2009
• Issued: 08/16/2016
• Est. Priority Date: 01/13/2009
• Status: Active Grant

• Current Assignee: Robert Bosch GmbH
• Original Assignee: Robert Bosch GmbH
• Inventors: Ebling, Julia, Krueger, Christof, Warzelhan, Jan Karl
• Primary Examiner(s): Ustaris, Joseph
• Assistant Examiner(s): Ren, Zhubing
• Application Number:
  US12/922,229
• Publication Number:
  US 20110037852A1
• Time in Patent Office: 2,461 Days
• US Class Current: 1/1
• CPC Class Codes: G06V 20/53 Recognition of crowd images, e.g. recognition of crowd congestion

Abstract

For operators of public or commercial buildings, such as supermarkets, it is interesting to determine how many persons, in particular potential buyers, enter the building each day. In other areas, as well, entering and exiting moving objects are counted: it is typical, for example, that cars driving into and out of a parking garage are counted. The invention relates to a device 1 for counting objects in a monitoring area, having a counting module 9, wherein the counting module comprises an object counter increasing by one counting value when one of the objects completely passes through a counting section 6 in a counting direction 4 as a counting object, having at least one monitoring camera 2 for capturing the monitoring area, wherein the monitoring camera 2 is designed for outputting an image data stream of the monitoring area, and having an extraction module 10 designed for extracting moving image areas in a current image of the image data flow, wherein the moving image areas can represent the objects or partial segments thereof, wherein the counting module 9 is designed for increasing the object counter 9 by a partial value for each current image for one of the moving image areas 3, the direction of motion thereof corresponding to the counting direction 4 and thus representing a potential counting object or a partial segment thereof, said partial value being less than the counting value.

Claims

1. A device (1) for counting objects in a surveillance region, comprising

a counting module (9), wherein the counting module includes an object counter that is increased by one counting value, on average, when one of the objects, as the object to be counted, completely passes through a counting section (6) in a counting direction (4), the counting section (6) defining a length that the corresponding object must cover between entering and exiting a region of an image in order to be counted, and

at least one surveillance camera (2) for recording the surveillance region, wherein the surveillance camera (2) outputs an image data stream of the surveillance region,

an extraction module (10), including a flow module (11) for determining an optical flow field having flow vectors for extracting moving image regions in a current image of the image data stream, wherein the moving image regions Gail-represent the objects or segments thereof, and

an estimating module (14) for estimating a subsection of the length that the corresponding object to be counted has covered in the current image depending on the direction of motion or velocity of the corresponding moving image region;

wherein the counting module (9) increases the object counter (9) by a partial value for each current image for one of the moving image regions (3), a direction of motion thereof corresponding to the counting direction (4) and thus representing a potential object to be counted, or a segment thereof,

wherein the partial value is less than the counting value,

wherein the subsection is estimated by projecting the flow vectors onto the counting direction (4) and/or the counting section (6), and

wherein a size of the partial value is determined depending on the subsection covered and/or a size of the segment of the object to be counted.

2. The device (1) according to claim 1, wherein the counting module (9) sums a partial value for each new and/or updated flow field having a moving image region that corresponds to an object to be counted.

3. The device (1) according to claim 1, wherein the objects to be counted are non-individualized and/or untracked and/or anonymized.

4. The device (1) according to claim 1, further comprising a segmenting module (12) which segments the moving image regions and/or the flow regions into object regions.

5. The device (1) according to claim 4, further comprising an object identification device (13) which identifies the object class of the segmented object regions.

6. The device (1) according to claim 4, wherein the subsection is assigned to one of the segmented object regions.

7. A method, implemented by a device including a memory, for counting objects in a surveillance region in a counting direction (4) along a counting section (6), the method comprising the steps of:

increasing an object counter by one counting value when one of the objects, as the object to be counted, completely passes through the counting section (6) in the counting direction (4),

extracting moving image regions from a current image of the surveillance region including determining an optical flow field having flow vectors,

defining a length for the counting section (6), which the corresponding object must cover between entering and exiting a region of an image to be counted,

increasing the object counter by a partial value for the current image for one of the moving image regions, where a direction of motion thereof corresponds to a counting direction (4) and thus represents a potential object to be counted or a segment thereof, wherein the partial value is less than the one counting value,

estimating a subsection of the length that the corresponding object to be counted has covered in the current image depending on the direction of motion or velocity of the moving image region by projecting the flow vectors onto the counting direction (4) and/or the counting section (6).

8. The method defined by claim 7, further comprising determining a size of the partial value based the subsection covered and/or a size of the segment of the object to be counted.

9. A computer program product comprising program code means, which program code means is stored on a non-transitory computer-readable data carrier and performs all steps of the method according to claim 7, when the program code means is run on a computer and/or the device (1).