Systems and methods for performing video confirmation of sensor-generated alarms
Methods and systems are provided for performing video confirmation of security system alarms using automated video analytics. When a sensor generates an alarm, the system automatically identifies an associated video camera and retrieves relevant video footage corresponding to the alarm timestamp. A video analytics algorithm analyzes the footage to generate a confidence score for alarm confirmation. If the confidence score exceeds a threshold, the alarm is reported as confirmed to a central monitoring station; otherwise it is reported as unconfirmed. The system can select specific analytics algorithms based on alarm type and manage multiple associated cameras with defined priorities. The confidence scoring can consider factors like camera field of view overlap and image quality. The analytics algorithms can be trained using operator-provided ground truth data to improve accuracy.
1 . A method for performing video confirmation of an alarm that is generated by a sensor of a security system, the method comprising:
storing a sensor-camera association table that associates the video camera with the sensor of the security system that generated the alarm;
receiving the alarm generated by the sensor of the security system, wherein the alarm has an alarm type and an alarm timestamp;
automatically identifying a video camera of a plurality of video cameras of the security system that corresponds to the sensor of the security system that generated the alarm by referencing the sensor-camera association table;
retrieving a video clip captured by the identified video camera that encompasses a time that corresponds to the alarm timestamp;
selecting a video analytics algorithm from a plurality of available video analytics algorithms based at least in part on the alarm type of the alarm generated by the sensor of the security system, wherein the sensor-camera association table identifies the video analytics algorithm from the plurality of video analytics algorithms based at least in part on the alarm type of the alarm qenerated by the sensor of the security system;
performing the video analytics algorithm on the received video clip to provide a video analytics result;
determining a confidence score in an automatic confirmation of the alarm based at least in part on the video analytics result;
comparing the confidence score to a confidence threshold;
when the confidence score exceeds the confidence threshold, reporting the alarm to a central monitoring station of the security system as a confirmed alarm; and
when the confidence score does not exceed the confidence threshold, reporting the alarm to the central monitoring station of the security system as an unconfirmed alarm.
2 . The method of claim 1 , wherein the sensor-camera association table associates two or more video cameras of the security system with the sensor, and a defined priority for each of the two or more video cameras.
3 . The method of claim 2 , comprising:
monitoring one or more monitored parameters including one or more of:
lighting conditions in a field of view of each of the two or more video cameras;
on-line status of each of the two or more video cameras;
a confidence level of each of the video analytics algorithms of each of the two or more video cameras in confirming respective alarms generated by the sensor as compared to alarms that are manually confirmed by an operator at the central monitoring station; and
updating the sensor-camera association table based at least in part on one or more of the monitored parameters.
4 . The method of claim 3 , comprising:
updating the defined priority for one or more of the two or more video cameras in the sensor-camera association table based at least in part on one or more of the monitored parameters.
5 . The method of claim 2 , wherein:
automatically identifying two or more video cameras of the security system that corresponds to the sensor of the security system including the video camera;
retrieving a video clip of each of the two or more identified video cameras that corresponds to the sensor, wherein each video clip encompasses a time that corresponds to the alarm timestamp;
performing a video analytics algorithm on the received video clip of each of the two or more identified video cameras clip to provide a corresponding video analytics result; and
determining an individual confidence score in the automatic confirmation of the alarm for each of the two or more identified video cameras based at least in part on the corresponding video analytics result.
6 . The method of claim 5 , comprising determining the confidence score in the confirmation of the alarm based on the individual confidence score.
7 . The method of claim 1 , wherein the confidence score in the automatic confirmation of the alarm is dependent on one or more of:
an overlap of a field of view of the identified video camera with an alarm condition that corresponds to the alarm; and
an image quality of the video clip captured by the identified video camera that encompasses the alarm timestamp.
8 . The method of claim 1 , wherein the sensor is one of a motion sensor, a glass break sensor and a magnetic contact sensor.
9 . The method of claim 1 , wherein determining the confidence score in the automatic confirmation of the alarm based at least in part on the video analytics result uses an Artificial Intelligence (AI) model, the method comprising:
training the AI model based on a ground truth provided by an operator of the central monitoring station of the security system by manually indicating whether the alarm is in fact a confirmed alarm or an unconfirmed alarm.
10 . A system comprising:
a sensor of a security system that is configured to generate an alarm;
a plurality of video cameras of the security system;
a controller operatively coupled to the sensor and the plurality of video cameras, the controller is configured to:
store a sensor-camera association table that associates the video camera with the sensor of the security system that qenerated the alarm;
receive the alarm generated by the sensor, wherein the alarm has an alarm type and an alarm timestamp;
automatically identify a video camera of the plurality of video cameras that corresponds to the sensor that generated the alarm by referencing the sensor-camera association table;
automatically retrieve a video clip captured by the identified video camera that encompasses a time that corresponds to the alarm timestamp;
automatically select a video analytics algorithm from a plurality of available video analytics algorithms based at least in part on the alarm type of the alarm generated by the sensor of the security system, wherein the sensor-camera association table identifies the video analytics algorithm from a plurality of video analytics algorithms based at least in part on the alarm type of the alarm qenerated by the sensor of the security system;
automatically perform the video analytics algorithm on the received video clip to provide a video analytics result;
automatically determine a confidence score in an automatic confirmation of the alarm based at least in part on the video analytics result;
automatically compare the confidence score to a confidence threshold;
when the confidence score exceeds the confidence threshold, report the alarm as a confirmed alarm; and
when the confidence score does not exceed the confidence threshold, report the alarm as an unconfirmed alarm.
11 . The system of claim 10 , wherein the confidence score in the automatic confirmation of the alarm is dependent on one or more of:
an overlap of a field of view of the identified video camera with an alarm condition that corresponds to the alarm; and
an image quality of the video clip captured by the identified video camera that encompasses the alarm timestamp.
12 . The system of claim 10 , wherein the sensor is one of a motion sensor, a glass break sensor and a magnetic contact sensor.
13 . The system of claim 10 , wherein the controller is configured to train the video analytics algorithm based on a ground truth provided by an operator of the security system by manually indicating whether the alarm is in fact a confirmed alarm or an unconfirmed alarm.
14 . A non-transitory computer readable medium storing instructions that when executed by one or more processors cause the one or more processors to:
store a sensor-camera association table that associates the video camera with the sensor of the security system that qenerated the alarm;
receive an alarm generated by a sensor of a security system, the alarm having an alarm type;
automatically identify a video camera of the security system that is pre-associated with the sensor by referencing the sensor-camera association table;
automatically select a video analytics algorithm from a plurality of video analytics algorithm that corresponds to the alarm type of the alarm, wherein the sensor-camera association table identifies the video analytics algorithm from a plurality of video analytics algorithms based at least in part on the alarm type of the alarm qenerated by the sensor of the security system;
automatically apply the selected video analytics algorithm to a video captured by the identified video camera to generate a video analytics result;
based at least in part on the video analytics result of the selected video analytics algorithm, automatically determine whether the alarm is confirmed or not; and
report the alarm to a central monitoring station of the security system as a confirmed alarm when the alarm is determined to be confirmed.
15 . The non-transitory computer readable medium of claim 14 , wherein the instructions cause the one or more processors to:
determine a confidence score in whether the alarm is confirmed based at least in part on the video analytics result of the selected video analytics algorithm; and
determine that the alarm is confirmed when the confidence score exceeds a confidence threshold.
16 . The non-transitory computer readable medium of claim 14 , wherein automatically determining whether the alarm is confirmed is based at least in part on an Artificial Intelligence (AI) Model, and where the instructions cause the one or more processors to:
train the AI model based on a ground truth provided by an operator of the central monitoring station of the security system.