System and method for detecting potential mugging event via trajectory-based analysis

1. A computer-implemented method for automatically detecting an occurrence of a potential mugging interaction event associated with two or more people concurrently present in a surveilled area using a video camera directed towards the surveilled area, the method comprising:

a) acquiring a video stream from the video camera, the video stream including a temporal sequence of video frames including the surveilled area within a FOV (field-of-view) associated with the video camera;

b) detecting and tracking two or more people within a common temporal sequence of video frames included in the video stream, and generating a trajectory of each person tracked within the common temporal sequence of video frames;

c) processing the trajectories of the tracked people to extract one or more trajectory interaction features (TIFs) associated with the trajectories of the two or more people tracked within the common temporal sequence of video frames,

wherein the TIFs include one or more of a position, a velocity, a travel orientation/direction, an acceleration, and a relative distance associated with the two or more people within the common temporal sequence of video frames; and,

d) applying predefined heuristics to the extracted TIFs to detect a potential mugging interaction event associated with the predefined heuristics has occurred involving at least two people of the two or more people tracked within the common temporal sequence of video frames,

wherein the predefined heuristics include the calculation of an evidence vector state, the evidence vector state calculated as a function of a proximity threshold, a change of travel orientation/direction threshold, an acceleration threshold, and a velocity threshold associated with the two or more people tracked within the common temporal sequence of video frames.

2. The computer-implemented method for automatically detecting the occurrence of a potential mugging interaction event associated with two or more people according to claim 1 , further comprising:

e) collecting evidence of the detected potential mugging interaction event in step d), the evidence including one or more of the common temporal sequence of video frames, an indicator associated with the common temporal sequence of video frames, one or more of extracted TIFs associated with the trajectories of the two or more people within the common temporal sequence of video frames, a time duration and start/end time associated with the detected potential mugging interaction event, a calculated probability of the occurrence of the potential mugging interaction event, and a classification of at least one of the two or more people within the common temporal sequence of video frames.

3. The computer-implemented method for automatically detecting the occurrence of a potential mugging interaction event associated with two or more people according to claim 2 , further comprising:

f) communicating an alert to an operatively associated central system, the alert indicating one or more of the common temporal sequence of video frames, an indicator associated with the common temporal sequence of video frames, one or more of extracted TIFs associated with the trajectories of the two or more people within the common temporal sequence of video frames, the time duration and start/end time associated with the detected potential mugging interaction event, the calculated probability of the occurrence of the potential mugging interaction event, and the classification of at least one of the two or more people within the common temporal sequence of video frames.

4. A computer-implemented method for automatically detecting the occurrence of a potential mugging interaction event associated with two or more people concurrently present in a surveilled area using a video camera directed towards the surveilled area, the method comprising:

a) acquiring a video stream from the video camera, the video stream including a temporal sequence of video frames including the surveilled area within a FOV (field-of-view) associated with the video camera;

b) detecting and tracking two or more people within a common temporal sequence of video frames included in the video stream, and generating a trajectory of each person tracked within the common temporal sequence of video frames;

c) processing the trajectories of the tracked people to extract one or more trajectory interaction features (TIFs) associated with the trajectories of the two or more people tracked within the common temporal sequence of video frames; and

d) applying predefined heuristics to the extracted TIFs to detect a potential mugging interaction event associated with the predefined heuristics has occurred involving at least two people of the two or more people tracked within the common temporal sequence of video frames,

wherein the predefined heuristics detect the potential mugging interaction event has occurred if the two or more people tracked are within a predefined proximity threshold relative to each other, at least one of the two or more people in proximity change a travel direction by a predefined change of travel orientation/direction, and at least one of the two or more people increase travel speed by a predefined travel speed threshold subsequent to being within the predefined proximity threshold relative.

5. A computer-implemented method for automatically detecting the occurrence of a potential mugging interaction event associated with two or more people concurrently present in a surveilled area using a video camera directed towards the surveilled area, the method comprising:

a) acquiring a video stream from the video camera, the video stream including a temporal sequence of video frames including the surveilled area within a FOV (field-of-view) associated with the video camera;

b) detecting and tracking two or more people within a common temporal sequence of video frames included in the video stream, and generating a trajectory of each person tracked within the common temporal sequence of video frames;

c) processing the trajectories of the tracked people to extract one or more trajectory interaction features (TIFs) associated with the trajectories of the two or more people tracked within the common temporal sequence of video frames; and

d) applying predefined heuristics to the extracted TIFs to detect a potential mugging interaction event associated with the predefined heuristics has occurred involving at least two people of the two or more people tracked within the common temporal sequence of video frames,

wherein the predefined heuristics detect the potential mugging interaction event has occurred if the two or more people tracked are within a predefined proximity threshold relative to each other and a speed profile pattern associated with at least one of the people tracked indicates an increase of travel speed above a predetermined speed threshold after a relatively short duration speed slow down within the predefined proximity threshold.

6. A video system for automatically detecting an occurrence of a potential mugging interaction event associated with two or more people concurrently present in a surveilled area comprising:

a video camera with an associated FOV (field-of-view) directed towards the surveilled area; and

a video processing system operatively associated with the video camera, the video processing system configured to:

a) acquire a video stream from the video camera, the video stream including a temporal sequence of video frames including the surveilled area within the FOV associated with the video camera;

b) detect and track two or more people within the common temporal sequence of video frames included in the video stream, and generate a trajectory of each person tracked within the common temporal sequence of video frames;

c) process the trajectories of the tracked people to extract one or more trajectory interaction features (TIFs) associated with the trajectories of the two or more people tracked within the common temporal sequence of video frames,

wherein the TIFs include one or more of a position, a velocity, a travel orientation/direction, an acceleration, and a relative distance associated with the two or more people within the common temporal sequence of video frames; and

d) apply predefined heuristics to the extracted TIFs to detect a potential mugging interaction event has occurred between at least two people of the two or more people tracked within the common temporal sequence of video frames,

wherein the predefined heuristics applied in step d) include the calculation of an evidence vector state, the evidence vector state calculated as a function of a proximity threshold, a change of travel orientation/direction threshold, an acceleration threshold, and a velocity threshold associated with the two or more people tracked within the common temporal sequence of video frames.

7. The video system for automatically detecting the occurrence of a potential mugging interaction event associated with two or more people according to claim 6 , further comprising the video processing system configured to:

e) collect evidence of the detected interaction event in step d), the evidence including one or more of the common temporal sequence of video frames, an indicator associated with the common temporal sequence of video frames, one or more of extracted TIFs associated with the trajectories of the two or more people within the common temporal sequence of video frames, a time duration and start/end time associated with the detected potential mugging interaction event, a calculated probability of the occurrence of the potential mugging interaction event, and a classification of at least one of the two or more people within the common temporal sequence of video frames.

8. The video system for automatically detecting the occurrence of a potential mugging interaction event associated with two or more people according to claim 7 , further comprising the video processing system configured to:

f) communicate an alert to an operatively associated central system, the alert indicating one or more of the common temporal sequence of video frames, an indicator associated with the common temporal sequence of video frames, one or more of extracted TIFs associated with the trajectories of the two or more people within the common temporal sequence of video frames, the time duration and start/end time associated with the detected potential mugging interaction event, the calculated probability of the occurrence of the potential mugging interaction event, and the classification of at least one of the two or more people within the common temporal sequence of video frames.

9. A video system for automatically detecting the occurrence of a potential mugging interaction event associated with two or more people concurrently present in a surveilled area comprising:

a video camera with an associated FOV (field-of-view) directed towards the surveilled area; and

a video processing system operatively associated with the video camera, the video processing system configured to:

a) acquire a video stream from the video camera, the video stream including a temporal sequence of video frames including the surveilled area within the FOV associated with the video camera;

b) detect and track two or more people within the common temporal sequence of video frames included in the video stream, and generate a trajectory of each person tracked within the common temporal sequence of video frames;

c) process the trajectories of the tracked people to extract one or more trajectory interaction features (TIFs) associated with the trajectories of the two or more people tracked within the common temporal sequence of video frames, and

d) apply predefined heuristics to the extracted TIFs to detect a potential mugging interaction event has occurred between at least two people of the two or more people tracked within the common temporal sequence of video frames,

wherein the predefined heuristics detect the potential mugging interaction event has occurred if the two or more people tracked are within a predefined proximity threshold relative to each other, at least one of the two or more people in proximity change a travel direction by a predefined change of travel orientation/direction, and at least one of the two or more people increases travel speed by a predefined travel speed threshold subsequent to being within the predefined proximity threshold relative.

10. A video system for automatically detecting the occurrence of a potential mugging interaction event associated with two or more people concurrently present in a surveilled area comprising:

a video camera with an associated FOV (field-of-view) directed towards the surveilled area; and

a video processing system operatively associated with the video camera, the video processing system configured to:

a) acquire a video stream from the video camera, the video stream including a temporal sequence of video frames including the surveilled area within the FOV associated with the video camera;

b) detect and track two or more people within the common temporal sequence of video frames included in the video stream, and generate a trajectory of each person tracked within the common temporal sequence of video frames;

c) process the trajectories of the tracked people to extract one or more trajectory interaction features (TIFs) associated with the trajectories of the two or more people tracked within the common temporal sequence of video frames, and

d) apply predefined heuristics to the extracted TIFs to detect a potential mugging interaction event has occurred between at least two people of the two or more people tracked within the common temporal sequence of video frames,

wherein the predefined heuristics detect the potential mugging interaction event has occurred if the two or more people tracked are within a predefined proximity threshold relative to each other and a speed profile pattern associated with at least one of the people tracked indicates an increase of travel speed above a predetermined speed threshold after a relatively short duration speed slow down within the predefined proximity threshold.

11. A computer-implemented method for automatically detecting an interaction event involving two or more people concurrently present in a surveilled area using a video camera directed towards the surveilled, the method comprising:

a) acquiring a video stream from the video camera, the video stream including a temporal sequence of video frames including the surveilled area within a FOV (field-of-view) associated with the video camera;

b) detecting and tracking two or more people within a common temporal sequence of video frames included in the video stream, and generating a trajectory of each person tracked within the common temporal sequence of video frames;

c) processing the trajectories of the tracked people to extract trajectory interaction features (TIFs) associated with the trajectories of the two or more people tracked within the common temporal sequence of video frames, the TIFs including a travel direction/orientation TIF associated with each of the two or more people, an acceleration TIF associated with each of, the two or more people and a relative distance TIF associated with a relative distance between the two or more people; and

d) applying predefined heuristics to the extracted TIFs to detect an interaction event associated with the predefined heuristics has occurred between at least two people of the two or more people tracked within the common temporal sequence of video frames,

wherein the predefined heuristics include the calculation of an evidence vector state, the evidence vector state calculated as a function of a proximity threshold, a change of travel orientation/direction threshold, an acceleration threshold, and a velocity threshold associated with the two or more people tracked within the common temporal sequence of video frames.

12. The computer-implemented method for automatically detecting an interaction event according to claim 11 , wherein the interaction event is a potential mugging involving at least two of the two or more people.

13. The computer-implemented method for automatically detecting the occurrence of an interaction event associated with two or more people according to claim 12 , further comprising:

e) collecting evidence of the detected potential mugging interaction event in step d), the evidence including one or more of the common temporal sequence of video frames, an indicator associated with the common temporal sequence of video frames, one or more of extracted TIFs associated with the trajectories of the two or more people within the common temporal sequence of video frames, a time duration and start/end time associated with the detected potential mugging interaction event, a calculated probability of the occurrence of the potential mugging interaction event, and a classification of at least one of the two or more people within the common temporal sequence of video frames.

14. The computer-implemented method for automatically detecting the occurrence of an interaction event associated with two or more people according to claim 13 , further comprising:

f) communicating an alert to an operatively associated central system, the alert indicating one or more of the common temporal sequence of video frames, an indicator associated with the common temporal sequence of video frames, one or more of extracted TIFs associated with the trajectories of the two or more people within the common temporal sequence of video frames, the time duration and start/end time associated with the detected potential mugging interaction event, the calculated probability of the occurrence of the potential mugging interaction event, and the classification of at least one of the two or more people within the common temporal sequence of video frames.

15. A computer-implemented method for automatically detecting an interaction event involving two or more people concurrently present in a surveilled area using a video camera directed towards the surveilled, wherein the interaction event is a potential mugging involving at least two of the two or more people, the method comprising:

a) acquiring a video stream from the video camera, the video stream including a temporal sequence of video frames including the surveilled area within a FOV (field-of-view) associated with the video camera;

b) detecting and tracking two or more people within a common temporal sequence of video frames included in the video stream, and generating a trajectory of each person tracked within the common temporal sequence of video frames;

c) processing the trajectories of the tracked people to extract trajectory interaction features (TIFs) associated with the trajectories of the two or more people tracked within the common temporal sequence of video frames, the TIFs including a travel direction/orientation TIF associated with each of the two or more people, an acceleration TIF associated with each of the two or more people and a relative distance TIF associated with a relative distance between the two or more people; and

d) applying predefined heuristics to the extracted TIFs to detect an interaction event associated with the predefined heuristics has occurred between at least two people of the two or more people tracked within the common temporal sequence of video frames,

wherein the predefined heuristics detect the potential mugging interaction event has occurred if the two or more people tracked are within a predefined proximity threshold relative to each other, at least one of the two or more people in proximity change a travel direction by a predefined change of travel orientation/direction, and at least one of the two or more people increases travel speed by a predefined travel speed threshold subsequent to being within the predefined proximity threshold relative.

16. A computer-implemented method for automatically detecting an interaction event involving two or more people concurrently present in a surveilled area using a video camera directed towards the surveilled, wherein the interaction event is a potential mugging involving at least two of the two or more people, the method comprising:

a) acquiring a video stream from the video camera, the video stream including a temporal sequence of video frames including the surveilled area within a FOV (field-of-view) associated with the video camera;

b) detecting and tracking two or more people within a common temporal sequence of video frames included in the video stream, and generating a trajectory of each person tracked within the common temporal sequence of video frames;

c) processing the trajectories of the tracked people to extract trajectory interaction features (TIFs) associated with the trajectories of the two or more people tracked within the common temporal sequence of video frames, the TIFs including a travel direction/orientation TIF associated with each of the two or more people, an acceleration TIF associated with each of the two or more people and a relative distance TIF associated with a relative distance between the two or more people; and

d) applying predefined heuristics to the extracted TIFs to detect an interaction event associated with the predefined heuristics has occurred between at least two people of the two or more people tracked within the common temporal sequence of video frames,

wherein the predefined heuristics detect the potential mugging interaction event has occurred if the two or more people tracked are within a predefined proximity threshold relative to each other and a speed profile pattern associated with at least one of the people tracked indicates an increase of travel speed above a predetermined speed threshold after a relatively short duration speed slow down within the predefined proximity threshold.