Anomalous stationary object detection and reporting

• US 10,257,466 B2
• Filed: 06/29/2017
• Issued: 04/09/2019
• Est. Priority Date: 06/29/2012
• Status: Active Grant

• Current Assignee: Intellective Ai, Inc.
• Original Assignee: Omni AI, Inc.
• Inventors: Xu, Gang, Cobb, Wesley Kenneth
• Primary Examiner(s): Atala, Jamie J
• Assistant Examiner(s): Nguyen, Kathleen V
• Application Number:
  US15/636,872
• Publication Number:
  US 20180084225A1
• Time in Patent Office: 649 Days
• Field of Search: 348134
• CPC Class Codes: G06F 18/2433 Single-class perspective, e.g. one-against-all classification; Novelty detection; Outlier detection
• CPC Class Codes: G06V 10/764 using classification, e.g. of video objects
• CPC Class Codes: H04N 7/18 Closed-circuit television [CCTV] systems, i.e. systems in which the video signal is not broadcast
• CPC Class Codes: H04N 7/188 Capturing isolated or intermittent images triggered by the occurrence of a predetermined event, e.g. an object reaching a predetermined position signal generation from motion picture films H04N5/253

Abstract

Techniques are disclosed for analyzing a scene depicted in an input stream of video frames captured by a video camera. The techniques include receiving data for an object within the scene and determining whether the object has remained substantially stationary within the scene for at least a threshold period. If the object is determined to have remained stationary for at least the threshold period, a rareness score is calculated for the object to indicate a likelihood of the object being stationary to an observed degree at an observed location. The rareness score may use a learning model to take into account previous stationary and/or non-stationary behavior of objects within the scene. In general, the learning model may be updated based on observed stationary and/or non-stationary behaviors of the objects. If the rareness score meets reporting conditions, the stationary object event may be reported.

Claims

1. A computer-implemented method, comprising:

analyzing, via one or more processors, an input stream of captured video frames, the captured video frames defining at least one scene, the captured video frames including a corresponding bounding box(es) for objects within the captured video frames;

determining, via the one or more processors, whether an object within the captured video frames has remained at a specified location in the at least one scene for at least a threshold period;

determining, via the one or more processors, whether the object has remained substantially stationary within the at least one scene during the threshold period;

upon determining that the object has remained substantially stationary and remained within the at least one scene for at least the threshold period, determining, via the one or more processors, a rareness score for the object, the rareness score determined based at least in part on a number of intersection pixels of the object'"'"'s corresponding bounding box(es), union of pixels covered by the object'"'"'s corresponding bounding box(es) during tracking period, and a change in size of the object'"'"'s corresponding bounding box(es); and

generating, via the one or more processors, an alert upon determining the rareness score exceeds a threshold.

2. The computer-implemented method of claim 1, wherein the rareness scored is further determined based at least on learned stationary and non-stationary count maps indicating observed frequencies of stationary and non-stationary objects in regions of the at least one scene.

3. The computer-implemented method of claim 2, further comprising:

updating the stationary and non-stationary count maps based on at least one of the observed stationary and non-stationary behavior of the object.

4. The computer-implemented method of claim 3, wherein updating includes updating pixel count value in area covered by the object in at least one of the stationary count map or the non-stationary count map.

5. The computer-implemented method of claim 1, wherein the rareness score is further determined based at least in part on a sum of stationary and non-stationary counts of all objects over all pixels.

6. The computer-implemented method of claim 1, further comprising:

assigning a value for the alert, the value indicating a measure of severity of the alert.

7. The computer-implemented method of claim 1, further comprising:

storing a list of substantially stationary objects no longer visible in the at least one scene;

upon determining the object appears from behind another object, comparing the object to the substantially stationary objects in the list; and

upon determining the object matches one of the substantially stationary objects in the stored list, adding time the object is substantially stationary before leaving the at least one scene, time between the object leaving the at least one scene and appearing from behind the other object, and time during which the object is substantially stationary after appearing from behind the other object, to obtain the time the object has remained at the observed location in the at least one scene.

8. The computer-implemented method of claim 1, further comprising:

storing a list of objects previously associated with rareness scores that exceeded the threshold value; and

if the object is determined to have remained stationary and remained within the at least one scene for at least the threshold period, comparing pixel locations of the object with pixel locations of objects in the stored list,

wherein the alert is generated or the object is reported only if pixel locations of the object do not match the pixel locations of any of the objects in the stored list.