System and method for performing video verification of sensor-generated alerts

Abstract

A method and system are provided for video verification of security system alarms using automated video analytics. When a sensor generates an alarm, the system automatically identifies the associated camera and retrieves relevant video recordings corresponding to the alarm timestamp. A video analytics algorithm analyzes the recording to generate a confidence score for alarm verification. If the confidence score exceeds a threshold, the alarm is reported as verified to a central monitoring station; otherwise, it is reported as unverified. The system can select a specific analytics algorithm based on the alarm type and manage multiple associated cameras with defined priorities. The confidence score can take into account factors such as camera field-of-view overlap and image quality. These analytics algorithms can be trained using operator-provided ground truth data to improve accuracy.

Description

Technical Field

[0001] This disclosure generally relates to security systems, and more specifically to video verification of alarms generated by sensors within a security system. Background Technology

[0002] Security systems comprise a variety of sensors, including motion sensors such as PIR sensors, door contact sensors, window contact sensors, glass break detectors, fire sensors, smoke detectors, and so on. When one or more of these sensors indicate a potential problem, it is desirable to confirm whether the potential problem actually exists or can be attributed to a false, non-safety-related condition. Confirmation can be made using video verification. For example, if a video stream indicates a person climbing over an open window with a window contact sensor indicating an alarm, the alarm can be confirmed as real. As another example, a PIR sensor may indicate motion, and the video stream may show that no one is present at the location where the motion was detected. Therefore, the alarm may not be confirmed and could be a false alarm. Motion detected by a PIR sensor could be, for example, trees swaying in the wind, snowfall, a small animal passing by, or any number of other false, non-safety-related conditions. Video surveillance systems may employ a large number of cameras, and it can be difficult to quickly determine which camera among a large number of cameras is best positioned to confirm (or deny) the alarm generated by the sensors. What is desired is a system and method for automatically identifying the appropriate camera for confirming or denying the alarm generated by the sensors. The desired outcome is a system and method for automatically identifying appropriate video segments captured by the identified cameras and automatically running one or more appropriate video analysis algorithms on the video segments to automatically confirm (or deny) alarms generated by sensors. Summary of the Invention

[0003] This disclosure generally relates to security systems, and more specifically to automatic video confirmation of alarms generated by sensors within a security system. An example may exist in a method for performing video confirmation of alarms generated by sensors of a security system. An exemplary method includes receiving an alarm generated by a sensor of the security system, wherein the alarm has an alarm type and an alarm timestamp. Automatically identifying a camera from a plurality of cameras in the security system corresponding to the sensor that generated the alarm. Retrieving a video segment captured by the identified camera, the video segment containing a time corresponding to the alarm timestamp. Performing a video analysis algorithm on the received video segment to provide video analysis results. Determining a confidence score in the automatic confirmation of the alarm based at least in part on the video analysis results. Comparing the confidence score to a confidence threshold. When the confidence score exceeds the confidence threshold, reporting the alarm as a confirmed alarm to a central monitoring station of the security system. When the confidence score does not exceed the confidence threshold, reporting the alarm as an unconfirmed alarm to the central monitoring station of the security system.

[0004] Another example may exist in a system. An exemplary system includes sensors of a security system configured to generate alarms, multiple cameras of the security system, and a controller operatively coupled to the sensors and the multiple cameras. The controller is configured to receive alarms generated by the sensors, wherein the alarms have an alarm type and an alarm timestamp. The controller is configured to automatically identify the camera among the multiple cameras corresponding to the sensor that generated the alarm. The controller is configured to automatically retrieve video clips captured by the identified camera, the video clips containing a time corresponding to the alarm timestamp. The controller is configured to automatically perform a video analysis algorithm on the received video clips to provide video analysis results. The controller is configured to automatically determine a confidence score in the automatic confirmation of the alarm, at least in part, based on the video analysis results. The controller is configured to automatically compare the confidence score with a confidence threshold. When the confidence score exceeds the confidence threshold, the controller is configured to report the alarm as a confirmed alarm. When the confidence score does not exceed the confidence threshold, the controller is configured to report the alarm as an unconfirmed alarm.

[0005] Another example could exist in a non-transitory computer-readable medium storing instructions. When executed by one or more processors, the instructions cause the processors to receive an alarm generated by a sensor of a security system, the alarm having an alarm type. The processors then automatically identify cameras of the security system pre-associated with the sensors. The processors then automatically select a video analysis algorithm from a plurality of video analysis algorithms corresponding to the alarm type. The processors then automatically apply the selected video analysis algorithm to the video captured by the identified camera to generate video analysis results. Based at least in part on the video analysis results of the selected video analysis algorithm, the processors automatically determine whether the alarm is confirmed. When it is determined that the alarm is confirmed, the processors report the alarm as a confirmed alarm to the central monitoring station of the security system.

[0006] The foregoing description is provided to facilitate understanding of the innovative features unique to this disclosure and is not intended as a complete description. A full understanding of this disclosure can be obtained by considering the entire specification, claims, drawings, and abstract as a whole. Attached Figure Description

[0007] This disclosure can be more fully understood by considering the following description of various examples in conjunction with the accompanying drawings, in which:

[0008] Figure 1 This is a schematic block diagram illustrating an exemplary system;

[0009] Figure 2A , Figure 2B and Figure 2CThis is a flowchart illustrating an exemplary method for performing video verification of alarms generated by sensors;

[0010] Figure 3 It is a flowchart illustrating a series of exemplary steps that can be performed by one or more processors when one or more processors execute instructions stored on a non-transitory computer-readable medium;

[0011] Figure 4 This is a flowchart illustrating an exemplary method; and

[0012] Figure 5 This is a schematic example of a floor plan.

[0013] While this disclosure is subject to various modifications and alternatives, its details have been shown by way of example in the accompanying drawings and will be described in detail. However, it should be understood that this disclosure is not intended to limit it to the specific examples described. Rather, it is intended to cover all modifications, equivalents, and alternatives that fall within the substance and scope of this disclosure. Detailed Implementation

[0014] The following description should be read with reference to the accompanying drawings, in which similar elements in different drawings are numbered in a similar manner. The drawings are not necessarily drawn to scale and depict examples that are not intended to limit the scope of this disclosure. While examples of various elements are illustrated, those skilled in the art will recognize that many of the examples provided have suitable alternatives that can be utilized.

[0015] This document assumes that all numbers are modified by the term “about” unless otherwise explicitly stated. Expressions of numerical ranges using endpoints include all numbers contained within that range (e.g., 1 to 5 includes 1, 1.5, 2, 2.75, 3, 3.80, 4, and 5).

[0016] As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” include plural references unless otherwise expressly stated. As used in this specification and the appended claims, the term “or” is generally used in its meaning to include “and / or” unless otherwise expressly stated.

[0017] It should be noted that references to "implementation schemes," "some implementation schemes," or "other implementation schemes" in the specification indicate that the described implementation schemes may include specific features, structures, or characteristics; however, each implementation scheme need not necessarily include that specific feature, structure, or characteristic. Furthermore, these phrases do not necessarily refer to the same implementation scheme. Additionally, when a specific feature, structure, or characteristic is described in conjunction with an implementation scheme, it is conceivable that, whether explicitly described or not, that feature, structure, or characteristic may be applied to other implementation schemes, unless otherwise expressly stated to the contrary.

[0018] Figure 1 This is a schematic block diagram illustrating an exemplary system 10. Exemplary system 10 may include components of a security system or communicate with such components. The security system may include multiple sensors 12, labeled 12a, 12b, and 12c. Although three sensors 12 are shown in total, the security system may include any number of sensors 12, and system 10 may include or communicate with any number of sensors. Each sensor 12 may be configured to generate an alarm. In some cases, each sensor 12 may independently be a door contact sensor, a window contact sensor, a motion sensor such as a PIR (passive infrared) sensor, a glass break detector, a fire sensor, a smoke sensor, and / or other sensors. In some cases, some sensors 12 may be magnetic contact sensors. The security system may include multiple cameras 14, labeled 14a, 14b, and 14c. Although three cameras 14 are shown in total, the security system may include any number of cameras 14, and system 10 may include or communicate with any number of cameras. System 10 may include a controller 16 that is operatively coupled to sensor 12 and camera 14.

[0019] Controller 16 is configured to receive alarms with alarm type and alarm timestamp from one or more sensors 12, and automatically identify one or more cameras 14 corresponding to the sensor 12 that generated the alarm. Controller 16 is configured to automatically retrieve video segments captured by the identified camera 14, the video segments containing the time corresponding to the alarm timestamp, and automatically perform video analysis algorithms on the received video segments to provide video analysis results. In some cases, controller 16 may be configured to select a video analysis algorithm from multiple available video analysis algorithms based at least in part on the alarm type of the alarm generated by the sensor 12. In some cases, controller 16 may be configured to train the video analysis algorithm over time based on a baseline truth provided by the operator of the security system by manually indicating whether the alarm is actually an acknowledged alarm or an unacknowledged alarm.

[0020] Controller 16 is configured to determine a confidence score in the automatic confirmation of an alarm, at least in part, based on video analytics results. In some cases, the confidence score in the automatic confirmation of an alarm may depend on the overlap between the field of view of the identified camera 14 and the alarm conditions corresponding to the alarm. In some cases, the confidence score in the automatic confirmation of an alarm may depend on the image quality of a video segment containing an alarm timestamp captured by the identified camera 14. These are merely examples. Controller 16 is configured to automatically compare the confidence score with a confidence threshold. When the confidence score exceeds the confidence threshold, controller 16 is configured to report the alarm as a confirmed alarm to central monitoring station 17. When the confidence score determined by the video analytics algorithm does not exceed the confidence threshold, controller 16 is configured to report the alarm as an unconfirmed alarm to central monitoring station 17. In some cases, central monitoring station 17 may also be operatively coupled to controller 16, sensor 12, and / or camera 14. In some cases, this can significantly reduce the need for manual intervention to analyze the video stream for alarm confirmation. This can significantly improve operator productivity and significantly reduce response time for handling alarms in safety systems.

[0021] Figure 2A , Figure 2B and Figure 2C This is a flowchart illustrating an exemplary method 18 for performing video confirmation of an alarm generated by sensors of a security system. Exemplary method 18 includes receiving an alarm generated by sensors of the security system, wherein the alarm has an alarm type and an alarm timestamp, as shown in box 20. Automatically identifying a camera from a plurality of cameras in the security system corresponding to the sensor that generated the alarm, as shown in box 22. Retrieving a video segment captured by the identified camera, the video segment containing a time corresponding to the alarm timestamp, as shown in box 24. Performing a video analysis algorithm on the received video segment. Determining a confidence score in the automatic confirmation of the alarm based at least in part on the video analysis results, as shown in box 26. Comparing the confidence score to a confidence threshold, as shown in box 28. When the confidence score exceeds the confidence threshold, reporting the alarm as a confirmed alarm to the central monitoring station of the security system, as shown in box 30. When the confidence score does not exceed the confidence threshold, reporting the alarm as an unconfirmed alarm to the central monitoring station of the security system, as shown in box 32. In some cases, method 18 may also include selecting a video analysis algorithm from a plurality of available video analysis algorithms, at least in part based on the alarm type generated by the sensors of the security system, as shown in box 34.

[0022] continue Figure 2BMethod 18 may include storing a sensor-camera association table that associates cameras with sensors of the security system that generate alarms, as shown in box 36. In some cases, method 18 may include automatically identifying cameras of the security system corresponding to sensors of the security system that generate alarms by referring to the sensor-camera association table. In some cases, the sensor-camera association table may also identify video analysis algorithms from multiple video analysis algorithms based at least in part on the alarm type of the alarms generated by the sensors of the security system. In some cases, the sensor-camera association table may associate two or more cameras of the security system with sensors and a defined priority for each of the two or more cameras.

[0023] In some cases, method 18 may include monitoring one or more monitoring parameters, as indicated in box 40. These one or more monitoring parameters may include the lighting conditions in the field of view of each of the two or more cameras, as shown in box 40a. These one or more monitoring parameters may include the online status of each of the two or more cameras, as shown in box 40b. These one or more monitoring parameters may include the confidence level of each video analysis algorithm of each of the two or more cameras in confirming a corresponding alarm generated by a sensor, compared to an alarm manually confirmed by an operator at a central monitoring station, as shown in box 40c. Method 18 may include updating a sensor-camera association table over time, at least in part based on one or more of the monitoring parameters, as shown in box 42. In some cases, the defined priorities of one or more of the two or more cameras may be updated in the sensor-camera association table, at least in part based on one or more of the monitoring parameters, as shown in box 44.

[0024] continue Figure 2C Method 18 may include automatically identifying two or more cameras in a security system, the two or more cameras corresponding to a camera-included sensor of the security system, as shown in box 46. In some cases, video clips from each of the two or more identified cameras corresponding to the sensor may be retrieved, wherein each video clip contains a time corresponding to an alarm timestamp, as shown in box 48. A video analysis algorithm may be performed on the received video clips of each of the two or more identified camera clips to provide corresponding video analysis results. Based at least in part on the corresponding video analysis results, an individual confidence score in the automatic confirmation of alarms for each of the two or more identified cameras is determined, as shown in box 50. In some cases, method 18 may include determining a confidence score in alarm confirmation based on the individual confidence scores, as shown in box 52.

[0025] The confidence score in automatic alarm confirmation may depend on the overlap between the field of view of the identified camera and the alarm conditions corresponding to the alarm. The confidence score in automatic alarm confirmation may also depend on the image quality of the video clips captured by the identified camera, including the alarm timestamp. These are just examples. In some cases, the sensor may be one of a motion sensor, a glass breakage sensor, and a magnetic contact sensor. In some cases, the confidence score in automatic alarm confirmation is determined, at least in part, based on video analytics results using an artificial intelligence (AI) model, and the AI ​​model is trained over time based on a baseline ground truth provided by an operator at the central monitoring station of the security system by manually indicating whether the alarm is actually a confirmed or unconfirmed alarm, as shown in box 54.

[0026] Figure 3 This is a flowchart illustrating a series of exemplary steps 56 that can be performed by one or more processors when executing instructions stored on a non-transitory computer-readable medium. For example, the one or more processors may be part of a controller 16. The one or more processors receive an alarm generated by a sensor of a security system, the alarm having an alarm type, as shown in box 58. The one or more processors automatically identify a camera of the security system pre-associated with the sensor, as shown in box 60. The one or more processors automatically select a video analysis algorithm from a plurality of video analysis algorithms corresponding to the alarm type of the alarm, as shown in box 62. The one or more processors automatically apply the selected video analysis algorithm to the video captured by the identified camera to generate a video analysis result, as shown in box 64. Based at least in part on the video analysis result of the selected video analysis algorithm, the one or more processors automatically determine whether the alarm is acknowledged, as shown in box 66. When it is determined that the alarm is acknowledged, the one or more processors report the alarm as an acknowledged alarm to the central monitoring station of the security system, as shown in box 68.

[0027] In some cases, the one or more processors may determine an alarm confirmation score, at least in part, based on the output of a selected video analytics algorithm, as shown in box 70. An alarm confirmation may be determined when the confidence score exceeds a confidence threshold, as shown in box 72. In some cases, the automatic determination of alarm confirmation may be based at least in part on an artificial intelligence (AI) model. These instructions may enable the one or more processors to train the AI ​​model over time based on baseline truth values ​​provided by operators at the central monitoring station of the security system, as shown in box 74.

[0028] Figure 4This is a flowchart illustrating exemplary method 76. Method 76 begins with a sensor detecting an alarm, as shown in box 78. The sensor number and alarm details are obtained, as shown in box 80. Alarm details may include one or more of the following: alarm type, area location, sensor type, area description, region description, date / time, and / or any other suitable alarm details. It is determined whether a camera priority list exists, as shown in decision box 82. This may include obtaining camera priority information and camera status (online, offline) from database 84, as shown in box 86. If a camera priority list does not exist, control proceeds to box 88, where manual alarm verification may be suggested. If a camera priority list exists, control proceeds to box 90, where one or more cameras are automatically selected from the camera priority list for alarm verification.

[0029] At decision box 92, the sensor type of the sensor that triggered the alarm is determined. In this example, the options are a PIR sensor, a fire sensor, and a glass breakage detector. If the PIR sensor triggered the alarm, control moves to box 94, and video analysis for human movement is performed on the video clip including the alarm event. If the fire sensor triggered the alarm, control moves to box 96, and video analysis for detecting a fire is performed on the video clip including the alarm event. If the glass breakage detector triggered the alarm, control moves to box 98, and video analysis for detecting glass breakage is performed on the video / audio clip including the alarm event. Control then moves to summation box 100, and then to box 102, where a confidence score is calculated in the confirmation of the alarm. In some cases, the confidence score ranges from 0% to 100% based on the level of detection and prediction output for each camera / video analysis. A confidence score greater than 70% is associated with strong confirmation, a confidence score between 30% and 70% indicates partial confirmation, and a confidence score less than 30% indicates no confirmation. These are just examples. If multiple cameras are involved, a combined confidence score can be calculated and used. At decision box 104, determine if the confidence score (or combined confidence score) is greater than 70% (a predetermined confidence score threshold). If not, control proceeds to box 106, and manual verification is recommended. If yes, control proceeds to box 108, and a confirmation alarm is reported as a confirmed alarm.

[0030] Figure 5This is a schematic plan view showing a plurality of sensors, including a first motion sensor 110 with a field of view (FOV) 110a, a second motion sensor 112 with an FOV 112a, and a door contact sensor 114. A first camera 116 has a fixed FOV 116a, and a second camera 118 has an adjustable FOV 118a. In some cases, the second camera 118 has pan-tilt-zoom (PTZ) capability. Based on the floor plan, sensor and camera capabilities, and installation locations, the following camera priority list is initially prepared for each camera. The camera priority list can then be dynamically adjusted based on the confirmation capability, confidence score, configuration changes, and current state of each of the multiple cameras.

[0031]

[0032] The table below provides the types of sensors and corresponding video analytics algorithms that can be used to better confirm alarms. This can improve alarm detection in less time and result in fewer false alarms.

[0033]

[0034] A confidence score ranging from 0% to 100% can be calculated based on the detection level, video analytics results, and the number of cameras confirming the alarm. In some cases, a formula can be used to determine the confidence score. The detection level can be based at least in part on, for example, image quality, field of view (FOV) covering the alarm event, scene lighting, camera specifications (night vision, PTZ), etc. The formula can include weights associated with each of several parameters. In the example below, 30 / 20 is the weight, and parameters such as the detection / prediction or object detection level are Boolean values ​​(1 or 0) based on the analysis results of a threshold. This 30 / 20 is an example weight for available parameters with a site-specific configuration. These weights and parameters can be updated for other sites based on, for example, a specific site configuration, site sensors, site cameras, and site cameras mapped to each sensor. These are just examples.

[0035] For motion detection / entry-exit gate detection events

[0036] Confidence score = (30) (Level of video analytics for detection and prediction) +

[0037] 30 (Objects detected as moving as humans in VA) +

[0038] 20 (More than one camera confirmation alert) +

[0039] 20 (Image quality / field of view and lighting)

[0040] For fire detection events

[0041] Confidence score = (30) (Level of video analytics for detection and prediction) +

[0042] 30 (Smoke / Flame Detection in VA) +

[0043] 20 (More than one camera confirmation alert) +

[0044] 20 (Image quality / field of view and lighting)

[0045] For earthquake detection events

[0046] Confidence score = (30) (Level of video analytics for detection and prediction) +

[0047] 30 (Weapons and / or human loitering detected in VA) +

[0048] 20 (More than one camera confirmation alert) +

[0049] 20 (Image quality / field of view and lighting)

[0050] Regarding glass breakage detection incidents

[0051] Confidence score = (30) (Level of video analytics for detection and prediction) +

[0052] 30 (Detecting scratches, dents, and cracks on glass in VA) +

[0053] 20 (More than one camera confirmation alert) +

[0054] 20 (Image quality / field of view and lighting)

[0055] For tampering detection events

[0056] Confidence score = (30) (Level of video analytics for detection and prediction) +

[0057] 30 (Detection in VA that the cover / sensor has been torn off or obstructed) +

[0058] 20 (More than one camera confirmation alert) +

[0059] 20 (Image quality / field of view and lighting)

[0060] Although several exemplary embodiments of this disclosure have been described thus, those skilled in the art will readily understand that other embodiments can be made and used within the scope of the appended claims. However, it should be understood that this disclosure is merely illustrative in many respects. Changes may be made to details, particularly those relating to shape, size, arrangement of parts, and exclusion and sequence of steps, without departing from the scope of this disclosure. The scope of this disclosure is, of course, defined by the language expressed in the appended claims.

Claims

1. A method for performing video verification of an alarm generated by a sensor in a security system, the method comprising: Receive the alarm generated by the sensor of the security system, wherein the alarm has an alarm type and an alarm timestamp; Automatically identify the camera among multiple cameras in the security system that corresponds to the sensor in the security system that generated the alarm; Retrieve video clips captured by the identified cameras, the video clips containing the time corresponding to the alarm timestamp; The received video segments are processed using video analysis algorithms to provide video analysis results. The confidence score in the automatic confirmation of the alarm is determined at least in part based on the video analysis results; The confidence score is compared with the confidence threshold; When the confidence score exceeds the confidence threshold, the alarm will be reported as a confirmed alarm to the central monitoring station of the security system. as well as When the confidence score does not exceed the confidence threshold, the alarm is reported as an unconfirmed alarm to the central monitoring station of the security system.

2. The method of claim 1, further comprising: The video analysis algorithm is selected from a plurality of available video analysis algorithms, based at least in part on the alarm type of the alarm generated by the sensors of the security system.

3. The method according to claim 1, wherein the method comprises: The system stores a sensor-camera association table, which associates the camera with the sensor that generated the alarm in the security system. as well as Automatically identifying the camera of the security system corresponding to the sensor that generates the alarm in the security system includes referencing the sensor-camera association table; Optionally, where: The sensor-camera association table identifies the video analysis algorithm from multiple video analysis algorithms based at least in part on the alarm type of the alarm generated by the sensors of the security system; or The sensor-camera association table associates two or more cameras of the security system with the sensors and the defined priorities of each of the two or more cameras.

4. The method according to claim 3, wherein the method comprises: Monitor one or more monitoring parameters, which include one or more of the following: Illumination conditions in the field of view of each of the two or more cameras; The online status of each of the two or more cameras; The confidence level of each video analysis algorithm in the video analysis algorithms of each of the two or more cameras in confirming the corresponding alarm generated by the sensor, compared to an alarm manually confirmed by an operator at the central monitoring station. as well as The sensor-camera association table is updated at least in part based on one or more of the monitoring parameters.

5. The method according to claim 4, wherein the method comprises: The priority of the definition of one or more cameras among the two or more cameras is updated in the sensor-camera association table, at least in part, based on one or more of the monitoring parameters.

6. The method according to claim 3, wherein: Automatically identify two or more cameras in the security system, wherein the two or more cameras correspond to the sensors of the security system including the cameras; Retrieve video clips from each of the two or more identified cameras corresponding to the sensor, wherein each video clip contains a time corresponding to the alarm timestamp; A video analysis algorithm is performed on the received video segments of each of the two or more identified camera segments to provide corresponding video analysis results; as well as Based at least in part on the corresponding video analysis results, determine an individual confidence score in the automatic confirmation of the alarm for each of the two or more identified cameras; Optionally, the method includes determining the confidence score in the confirmation of the alarm based on the individual confidence score.

7. The method according to any one of claims 1 to 6, wherein the confidence score in the automatic confirmation of the alarm depends on one or more of the following: The overlap between the identified camera's field of view and the alarm conditions corresponding to the alarm; and Image quality of the video clip captured by the identified camera, including the alarm timestamp.

8. The method according to any one of claims 1 to 6, wherein the confidence score in the automatic confirmation of the alarm is determined at least in part based on the video analysis results using an artificial intelligence (AI) model, the method comprising: The AI ​​model is trained based on a baseline ground truth provided by the operator of the central monitoring station of the security system, who manually indicates whether the alarm is actually an acknowledged alarm or an unacknowledged alarm.

9. A system comprising: A sensor in a security system, the sensor being configured to generate an alarm; The security system has multiple cameras; A controller, operatively coupled to the sensors and the plurality of cameras, is configured to: Receive the alarm generated by the sensor, the alarm having an alarm type and an alarm timestamp; Automatically identify the camera among the plurality of cameras that corresponds to the sensor that generated the alarm; Automatically retrieve video clips captured by the identified cameras, the video clips containing the time corresponding to the alarm timestamp; The received video segments are automatically processed using video analysis algorithms to provide video analysis results. The confidence score in the automatic confirmation of the alarm is automatically determined, at least in part, based on the video analysis results. The confidence score is automatically compared with the confidence threshold. When the confidence score exceeds the confidence threshold, the alarm will be reported as a confirmed alarm; as well as When the confidence score does not exceed the confidence threshold, the alarm will be reported as an unconfirmed alarm.

10. The system of claim 9, wherein the confidence score in the automatic confirmation of the alarm depends on one or more of the following: The overlap between the identified camera's field of view and the alarm conditions corresponding to the alarm; and Image quality of the video clip captured by the identified camera, including the alarm timestamp.

Images