Reducing false alarms in video surveillance systems

Abstract

A video surveillance system for reducing false alarms. The video monitoring system comprises a camera and an electronic processor. The electronic processor is configured to, when a moving object is detected in a video captured by a camera, perform object detection on the video to determine a score associated with a class of objects, determine a feature associated with the moving object detected in the video using metadata associated with the video, and determine a feature associated with the moving object detected in the video using metadata associated with the video. And analyzing the score associated with the class of objects and the feature associated with the moving object detected in the video using a machine learning algorithm to determine whether the moving object detected in the video is a false alarm or a true alarm. The electronic processor is further configured to generate alarm information when a moving object detected in the video is a true alarm.

Description

Background Art

[0001] Currently, video surveillance systems are configured to trigger an alarm when, for example, a person is detected moving within a restricted area. A restricted area is an observation area monitored by a video surveillance system. When an alarm is triggered, a notification or alarm message is sent to security personnel (e.g., Security Operating Center (SOC) personnel). Summary of the invention

[0002] The video / image processing technology traditionally used in video surveillance systems is able to detect motion in the video. Typically, the video / image processing technology is highly sensitive to motion to ensure that every motion that may pose a threat to the viewing area is detected and an alarm is triggered. However, this high sensitivity results in many false alarms being triggered. For example, using a video / image processing technology that is highly sensitive to motion may cause the video surveillance system to trigger a true alarm when an intruder is crawling across the grass, but it may also cause the video surveillance system to trigger a false alarm when the video surveillance system detects moving trees, raindrops on the camera lens, moving animals, and the like. False alarms may be as high as 90% of the alarms triggered by the video surveillance system.

[0003] Today, there are surveillance systems that utilize deep neural networks (DNNs) to perform object detection in images. However, there are multiple disadvantages to using DNN object detection alone to determine whether an alarm should be generated in a video surveillance system. One disadvantage is that an intruder can easily trick a video surveillance system that relies solely on object detection methods to trigger an alarm. For example, if a video surveillance system is configured to trigger an alarm when a DNN determines that a person is in a restricted area, the video surveillance system may not be able to trigger an alarm when the person covers himself with a piece of cardboard and moves in the restricted area.

[0004] Another disadvantage is that even the most advanced DNNs may not be able to determine that an object belongs to a certain class of objects (e.g., a person or a vehicle) because the training data is not representative of one or more domains or one or more camera positions. DNNs may also not be able to determine that an object belongs to a certain class of objects due to the quality of the images they analyze (e.g., images taken in low light, grayscale images, thermal images, noisy images, images with a low number of pixels, images with low resolution, and the like). These challenges may increase the likelihood of false and misdetections.

[0005] In addition, in a conventional video surveillance system, some areas monitored by the cameras included in the video surveillance system may be so-called "non-sensitive" areas, where objects of interest may already exist (e.g., parking lots where vehicles are parked). In non-sensitive areas, it is desirable to have the video surveillance system generate an alarm only when the detected object of interest is moving. In other words, when a stationary object of interest is detected in a non-sensitive area of ​​the image, the video surveillance system should not generate an alarm. Other areas monitored by the camera may be "sensitive" areas. If an object of interest is present in a sensitive area, the video surveillance system is expected to generate an alarm regardless of whether the object of interest is moving. A DNN trained to recognize static, non-moving objects may be sufficient for generating alarms for sensitive areas, but many false alarms may be generated for non-sensitive areas. In order to avoid generating false alarms for non-sensitive areas, some examples described herein utilize a temporal DNN that is trained to recognize moving objects without detecting static objects, and some examples analyze the video captured by the camera only when motion is detected in the video. In some cases, a video is a video clip comprising one or more frames or images.

[0006] The various aspects, features, and embodiments described herein provide, among other things, systems and methods for reducing false alarms in video surveillance systems. Suppressing true alarms may have the unintended consequence of reducing false alarms in video surveillance systems. However, the various aspects, features, and embodiments described herein also reduce the unintentional suppression of true alarms. In order to reduce false alarms in video surveillance systems, one aspect combines image-based object classification of moving objects using artificial intelligence (e.g., DNN) with features associated with the moving objects. These features are determined using metadata from the video and describe various aspects of the object's motion that are relevant to determining whether the object's motion is related to human activity.

[0007] One example provides a video surveillance system for reducing false alarms. The video surveillance system includes a camera and an electronic processor. The electronic processor is configured to: when a moving object is detected in a video captured by the camera, perform object detection on the video to determine a score associated with a class of objects, wherein the score represents the likelihood that the moving object detected in the video is associated with a class of objects; use metadata associated with the video to determine features associated with the moving object detected in the video; and use a machine learning algorithm to analyze the score associated with the class of objects and the features associated with the moving object detected in the video to determine whether the moving object detected in the video is a false alarm or a true alarm. The electronic processor is also configured to generate alarm information when the moving object detected in the video is a true alarm.

[0008] Another example provides a method for reducing false alarms in a video surveillance system. The method includes: when a moving object is detected in a video captured by a camera, performing object detection on the video to determine a score associated with a class of objects, wherein the score represents the likelihood that the moving object detected in the video is associated with a class of objects; using metadata associated with the video to determine features associated with the moving object detected in the video; and using a machine learning algorithm, analyzing the score associated with the class of objects and the features associated with the moving object detected in the video to determine whether the moving object detected in the video is a false alarm or a true alarm. The method also includes generating alarm information when the moving object detected in the video is a true alarm.

[0009] Other aspects, features, and embodiments will become apparent by consideration of the detailed description and accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] Figure 1 is a block diagram of a video surveillance system for reducing false alarms according to one example.

[0011] Figure 2 is a flow chart of a method for reducing false alarms in a video surveillance system according to one example.

[0012] Figure 3 An example image of a surveillance area is shown with metadata associated with moving objects overlaid thereon.

[0013] Figure 4 Example pseudocode for a random forest algorithm is shown.

[0014] Figure 5 is a process flow diagram illustrating techniques for combining AI-based and metadata-based object detection and classification.

[0015] Figure 6 is an example graph illustrating the performance of certain systems and methods described herein relative to the performance of other alternative methods. DETAILED DESCRIPTION

[0016] Before explaining any aspect, feature or embodiment in detail, it should be understood that the present disclosure is not intended to limit its application to the details of construction and arrangement of components set forth in the following description or shown in the following drawings. Various aspects, features and embodiments can have other configurations and can be practiced or carried out in various ways.

[0017] A plurality of hardware- and software-based devices and a plurality of different structural components may be used to implement the various embodiments. In addition, the embodiments may include hardware, software, and electronic components or modules, which, for purposes of discussion, may be shown and described as if they were mostly implemented only in hardware. However, a person of ordinary skill in the art will recognize, based on a reading of this detailed description, that in at least one embodiment, the electronic-based aspects of the present invention may be implemented in software (e.g., stored on a non-transitory computer-readable medium) that is executable by one or more processors. For example, the "control unit" and "controller" described in the specification may include one or more electronic processors, one or more memory modules including a non-transitory computer-readable medium, one or more communication interfaces, one or more application-specific integrated circuits (ASICs), and various connections (e.g., a system bus) connecting the various components.

[0018] Figure 1 An example of a video surveillance system 100 for reducing false alarms is shown. The video surveillance system 100 includes a camera 105, a second camera 110, an input device 115, a display device 120, a memory 125, and an electronic processor 130. The display device 120 may include, for example, a touch screen, a liquid crystal display (LCD), a light emitting diode (LED), an LED display, an organic LED (OLED) display, an electroluminescent display (ELD), and the like. The input device 115 may include, for example, a keypad, a mouse, a touch screen (e.g., as part of the display device 120), a microphone, a camera, or the like (not shown). The electronic processor 130, the memory 125, the camera 105, the second camera 110, the input device 115, and the display device 120 communicate wirelessly via one or more communication lines or buses or a combination thereof. It should be understood that the video surveillance system 100 may include in various configurations in addition to Figure 1 In some cases, the video surveillance system 100 may include additional components other than those shown in FIG. Figure 1 For example, video surveillance system 100 may include only one camera (eg, camera 105), or the functionality of input device 115 and display device 120 may be combined into a single device.

[0019] The memory 125 includes object recognition software 135, behavioral signature determination software 140, and machine learning fusion software 145. In some cases, the object recognition software 135, behavioral signature determination software 140, and machine learning fusion software 145 include computer executable instructions that, when executed by the electronic processor 130, cause the electronic processor 130 to perform the functions described herein. It should be understood that the functions described herein as being performed when multiple software components are executed by the electronic processor 130 may be performed by a single software component executed by the electronic processor 130. For example, the behavioral signature determination software 140 and the machine learning fusion software 145 may be combined into a single software component. It should be understood that the memory 125 may include in various configurations in addition to the computer executable instructions. Figure 1 The invention may include additional components beyond those shown in the drawings and may perform additional functions beyond those described herein.

[0020] Figure 2 A flow chart of an example method 200 for reducing false alarms in a video surveillance system (e.g., video surveillance system 100) is shown. The method 200 begins at block 205 when the electronic processor 130 detects a moving object in a video captured by the camera 105. When the moving object is detected in the video, at block 210, the electronic processor 130 performs object detection on the video to determine a score associated with a class of objects, wherein the score represents a likelihood that the moving object detected in the video is associated with the class of objects. For example, a first class of objects may be a person, a second class of objects may be a vehicle, and a third class of objects may be an aerial drone. In one example, the video may include motion that causes the electronic processor 130 to detect the moving object. In another example, the video may include a predetermined length of continuous footage (e.g., 5 seconds) before the time when the electronic processor 130 detects the moving object and a predetermined length of continuous footage (e.g., 5 seconds) after the time when the electronic processor 130 detects the moving object. In yet another example, the video may include a single frame (e.g., a frame captured by the camera 105 at the moment when the electronic processor 130 detects the moving object). In some cases, the electronic processor 130 uses a deep neural network (DNN) (e.g., a convolutional neural network, a recurrent neural network, and the like) to perform object detection on the video. For example, based on the video, the DNN can determine a first score of 0.15 associated with a first class of objects (e.g., a vehicle), which indicates a 15% probability that the moving object is a vehicle, and a second score of 0.75 associated with a second class of objects (e.g., a person), which indicates a 75% probability that the moving object is a person. In some cases, in addition to performing object detection on the video to determine a score associated with a class of objects, the electronic processor 130 can also use the DNN to determine a location associated with the moving object over time and a bounding box associated with the moving object over time.

[0021] At block 215, electronic processor 130 uses metadata associated with the video to determine features associated with a moving object detected in the video. In some cases, the metadata includes a timestamp location of the moving object, a bounding box around the moving object, a trajectory of the moving object, and the like. In some cases, using metadata associated with the video, electronic processor 130 determines a plurality of features associated with a moving object detected in the video. The features determined by the electronic processor 130 include, for example, at least one selected from the group consisting of: (a) displacement of a moving object over time; (b) change in bounding box height associated with the moving object over time; (c) average change in direction of the moving object over time; (d) standard deviation of the change in direction of the moving object over time from a starting position of the moving object; (e) average distance traveled by the moving object over time; (f) standard deviation of the distance traveled by the moving object over time; (g) difference in bounding box width between frames; (h) difference in bounding box height between frames; (i) mean absolute percentage error associated with fitting lines to direction values ​​associated with the moving object over time; (j) mean absolute percentage error associated with fitting lines to position values ​​associated with the moving object over time; and (k) mean absolute percentage error associated with fitting lines to distance values ​​associated with the moving object over time. In other cases, the features determined by the electronic processor 130 may include fewer, additional, or different features than those described above. Using the features determined by the electronic processor 130, knowledge about the motion of the moving object can be determined, including: (a) whether the motion of the moving object is smooth and continuous; (b) whether the size of the moving object is consistent when the moving object is not moving toward or away from the camera 105; (c) the linear transformation of the bounding box width and height between frames when the moving object is moving toward or away from the camera 105; (d) whether the moving object is moving in a consistent direction; and (e) whether the motion of the moving object is purposeful. These features indicate whether the motion is associated with human activity and, therefore, with a true alarm (e.g., a vehicle or a person). For example, if the moving object remains the same size (bounding box width and height remain the same) when it appears to move toward the camera 105, the moving object may be a false alarm (e.g., a bug crawling on the lens or raindrops dripping from the lens).

[0022] Figure 3An example image of a surveillance area is shown overlaid with metadata associated with a moving object, which the electronic processor 130 can use to determine characteristics associated with the moving object. The metadata can be in the Open Network Video Interface Forum (ONVIF) Profile-M standard format, a proprietary format, or the like. For example, in Figure 3 In , the metadata includes the position of the moving object over multiple frames of the video. Figure 3 Each (green) point in represents the position of a moving object in the frame. Figure 3 Each rectangle included in represents an approximate portion of the surveillance area occupied by the object in the frame. In some cases, the position of the moving object is the center point of the bounding box. For example, a first point 300 represents a first position of the moving object at a first time, and a first bounding box 305 represents a first approximate portion of the surveillance area occupied by the moving object at the first time. A second point 310 represents a second position of the moving object at a second time, and a second bounding box 315 represents a second approximate portion of the surveillance area occupied by the moving object at the second time. The trajectory or direction of the moving object can be determined by fitting a line to the points representing the position of the moving object over time.

[0023] At block 220, the electronic processor 130 uses a machine learning algorithm to analyze the scores associated with the class of the object detected in the video and the features associated with the moving object to determine whether the moving object detected in the video is a false alarm or a true alarm. In some cases, the electronic processor 130 may also use a machine learning algorithm to analyze the position associated with the moving object over time as determined using the DNN, the bounding box associated with the moving object over time as determined using the DNN, or both. In some cases, the electronic processor 130 may generate an overlap score and additionally analyze the overlap score with a machine learning algorithm. The overlap score may be determined based on the similarity between the position associated with the moving object over time as determined using the DNN, the bounding box associated with the moving object over time as determined using the DNN, or both, and the metadata-based features determined in block 215. In some cases, the machine learning algorithm is a random forest classifier. Random forest is a model ensemble method for classification, regression analysis, and other types of problems that builds multiple decision trees on randomly sampled training data sets. The random forest classifier performs well even in the presence of outliers and noise. Random forests also provide an estimated error associated with the classification, an estimate of the strength or accuracy of each tree, correlations between trees, and an estimate of the importance associated with each of a number of variables.

[0024] Figure 4An example pseudocode 400 for a random forest algorithm is shown. The first section 405 of the pseudocode 400 includes pseudocode associated with creating a random forest using training data. The "p variables" described in the pseudocode 400 may include the class of the object to which the mobile object detected in the video may belong and the features associated with the mobile object. Splitting variables may refer to nodes that create new levels in a decision tree in the forest. For example, a variable associated with the class "person" of the object may be represented by a first node. The first node may be split so that when the input is to the random forest and the score associated with the class "person" of the object is greater than 0.6, the electronic processor 130 selects the second node, and when the score associated with the class "person" of the object is less than or equal to 0.6, the electronic processor 130 selects the third node. In another example, a variable (feature) associated with whether the size of the mobile object is consistent when the mobile object is not moving toward or away from the camera 105 may be represented by a first node, and the first node may be split so that when the size of the mobile object is consistent, the electronic processor 130 selects the second node, and when the size of the mobile object is inconsistent, the electronic processor 130 selects the third node. The second section 410 of pseudocode 400 involves classifying a moving object in a video ("new data point x"). In classification problems, such as the one presented herein, majority voting is used to predict the class to which a moving object in a video belongs (in this case, a false alarm or a true alarm). For example, if 60% of the trees included in the random forest determine that a moving object in a video is a true alarm, and 40% of the trees included in the random forest determine that a moving object in a video is a false alarm, then the electronic processor 130 will determine that the moving object in the video is a true alarm.

[0025] At box 225, when the moving object detected in the video is a true alarm, the electronic processor 130 generates alarm information. In some cases, the electronic processor 130 sends the generated alarm information to an output device (e.g., a display device 120, a speaker, an LED light, a combination of the foregoing, or the like). In some cases, the electronic processor 130 sends the video along with the generated alarm information to the display device 120. The display device 120 displays the video along with the generated alarm information to allow security personnel to view the video. Based on their analysis of the video, the security personnel can provide feedback to the electronic processor 130 via the input device 115 regarding the generated alarm information. For example, if the electronic processor 130 generates alarm information based on a video showing tumbleweeds blowing in the wind, the security personnel can provide feedback that the video contains a false alarm. Based on the received feedback, the electronic processor 130 adjusts or retrains the DNN executed at box 210, the machine learning algorithm executed at box 220, or both. The retraining process is performed by Figure 5In some cases, based on the feedback received, the software developer can adjust which metadata-based features the electronic processor 130 uses the machine learning algorithm to analyze.

[0026] In some cases, to minimize the number of true alarms for which the electronic processor 130 fails to generate an alarm message, the electronic processor 130 operates in a "high alarm message mode" for a predetermined amount of time after a camera in the video surveillance system 100 captures a moving object that the electronic processor 130 determines to be a true alarm. For example, in some cases, the electronic processor 130 generates a second alarm message when a second moving object or moving objects are detected in a second video captured by a second camera 110 a predetermined amount of time after the camera 105 captures the video and the moving object detected in the video is a true alarm. In other words, when the electronic processor 130 determines that a moving object captured by the camera 105 in the video surveillance system 100 is a true alarm and generates alarm information, if another camera in the video surveillance system 100 (e.g., the second camera 110) captures a moving object (e.g., the moving object captured by the camera 105 or a different moving object) within a predetermined amount of time (e.g., 30 minutes after the moving object is detected in the video captured by the camera 105), the electronic processor 130 generates alarm information without performing the method 200 to determine whether the moving object captured by the second camera 110 is a false alarm. In another example, when a second moving object or moving objects are detected in a second video captured by the camera 105 a predetermined amount of time after the moving object is detected in the video and the moving object detected in the video is a true alarm, the electronic processor 130 generates a second alarm.

[0027] Figure 5 5 is a process flow diagram 500 illustrating a technique for combining AI-based and metadata-based object detection and classification. Block 505 includes some advantages (listed under "+") and disadvantages (listed under "-") of using artificial intelligence (AI) (e.g., DNN) to perform object detection to determine whether a moving object in a video is a true alarm or a false alarm. For example, a DNN may be able to classify whether a moving object is a vehicle or a person, but may not perform well when the video is dark, has low contrast, or is pixelated. Figure 5As shown, the input received by the DNN at box 505 is one or more images (e.g., one or more frames included in a video). Box 510 includes some advantages (listed under "+") and disadvantages (listed under "-") of using metadata-based features to determine whether a moving object in a video is a true alarm or a false alarm. For example, metadata-based features are generally accurate when used to determine whether a moving object in a dark, low-contrast, or pixelated video is a true alarm or a false alarm. However, when the moving object is an animal or a shadow, the metadata-based features are generally less accurate in determining whether a moving object in a video is a true alarm or a false alarm. The input received at box 510 includes metadata from a camera included in the video surveillance system 100. The camera from which the metadata is received is the same camera as the camera that captured the image received at box 505.

[0028] The one or more scores generated at block 505 and the features determined at block 510 are input to block 515. At block 515, the scores and features are analyzed with a machine learning algorithm (via the electronic processor 130 executing the machine learning fusion software 145) to determine whether the moving object is a true alarm or a false alarm. As described above, when the electronic processor 130 determines that the moving object is a true alarm, the electronic processor 130 generates alarm information and sends the alarm information to the display device 120. The display device 120 may display a visual representation of the alarm information, a video that caused the alarm information to be generated, or both (via a software application user interface (UI)). Figure 5 520 in FIG. 1 ). In some cases, security personnel may review the alarm and video and decide whether the electronic processor 130 accurately determined whether the moving object included in the video was a false alarm or a true alarm. Based on their determination, the security personnel may provide feedback. For example, when the security personnel determines that the moving object is a false alarm, the security personnel may select a graphical user interface (GUI) button displayed on the display device 120 via, for example, a touch screen of the display device 120 (e.g., the input device 115). As described above, feedback from the security personnel may be used to retrain the machine learning algorithm (by Figure 5 525 in FIG. Figure 5 Block 530 represents the high alert information mode of the video surveillance system 100 described above.

[0029] Figure 66 is a graph 600 showing the performance of the systems and methods described herein relative to the performance of other alternative methods. The x-axis 605 of the graph 600 represents the false rejection rate, or the percentage of true alarms that are incorrectly determined to be false alarms. The y-axis 610 of the graph 600 represents the true rejection rate, or the percentage of false alarms that are correctly determined to be false alarms. The systems and methods described herein are represented by line 615. Figure 6 As shown, the systems and methods described herein are able to achieve a false rejection rate of less than 0.5% while maintaining a true rejection rate of greater than 50%.

[0030] Thus, the aspects, features and embodiments described herein provide, inter alia, a video surveillance system and a method for reducing false alarms in a video surveillance system. Various features and advantages of the embodiments are set forth in the appended claims.

Claims

1. A video surveillance system for reducing false alarms, the video surveillance system comprising: Camera; as well as An electronic processor configured to: When a moving object is detected in the video captured by the camera: performing object detection on the video to determine a score associated with a class of objects, wherein the score represents a likelihood that the moving object detected in the video is associated with the class of objects; determining, using metadata associated with the video, features associated with the moving object detected in the video; Using a machine learning algorithm, analyzing the scores associated with the class of objects and features associated with the moving object detected in the video to determine whether the moving object detected in the video is a false alarm or a true alarm; and When the moving object detected in the video is a true alarm, alarm information is generated.

2. The video surveillance system according to claim 1, wherein: The electronic processor is configured to perform object detection using a deep neural network.

3. The video surveillance system according to claim 1, wherein: The machine learning algorithm is a random forest classifier.

4. The video surveillance system according to claim 1, wherein: The feature associated with the moving object detected in the video is at least one item selected from the group consisting of: the displacement of the moving object over time; the change in the height of a bounding box associated with the moving object over time; the average change in direction of the moving object over time from a starting position of the moving object; the standard deviation of the change in direction of the moving object over time; the average distance traveled by the moving object over time; the standard deviation of the distance traveled by the moving object over time; the difference in bounding box width between frames; The difference in bounding box height between frames; a mean absolute percentage error associated with fitting a line to direction values ​​associated with the moving object over time; a mean absolute percentage error associated with fitting a line to position values ​​associated with the moving object over time; and the mean absolute percentage error associated with fitting a line to the distance values ​​associated with the moving object over time.

5. The video surveillance system according to claim 1, wherein: The features associated with the moving object detected in the video are relevant to determining whether the moving object is associated with human activity.

6. The video surveillance system according to claim 1, further comprising a display device, and wherein: The electronic processor is also configured to transmit the alarm information and the video to the display device.

7. The video surveillance system according to claim 1, further comprising an input device, and wherein: The electronic processor is further configured to: receiving feedback regarding the alarm information based on the video via the input device; and Based on the feedback, the machine learning algorithm is adjusted.

8. The video surveillance system according to claim 1, further comprising a second camera, and wherein: The electronic processor is further configured to: Second alarm information is generated when a second moving object or the moving object is detected in a second video captured by the second camera within a predetermined amount of time after the moving object is detected in the video and the moving object detected in the video is a true alarm.

9. The video surveillance system according to claim 1, wherein: The electronic processor is further configured to: Second alarm information is generated when a second moving object or the moving object is detected in a second video captured by the camera within a predetermined amount of time after the moving object is detected in the video and the moving object detected in the video is a true alarm.

10. The video surveillance system according to claim 1, wherein: The metadata includes at least one selected from the group consisting of a time-stamped position of the moving object, a bounding box around the moving object, and a trajectory of the moving object.

11. A method for reducing false alarms in a video surveillance system, the method comprising: When a moving object is detected in the video captured by the camera: performing object detection on the video to determine a score associated with a class of objects, wherein the score represents a likelihood that the moving object detected in the video is associated with the class of objects; determining, using metadata associated with the video, features associated with the moving object detected in the video; Using a machine learning algorithm, analyzing the scores associated with the class of objects and features associated with the moving object detected in the video to determine whether the moving object detected in the video is a false alarm or a true alarm; and When the moving object detected in the video is a true alarm, alarm information is generated.

12. The method according to claim 11, wherein: Performing object detection on the video to determine a score associated with a class of objects includes performing object detection using a deep neural network.

13. The method according to claim 11, wherein: The machine learning algorithm is a random forest classifier.

14. The method according to claim 11, wherein: The feature associated with the moving object detected in the video comprises at least one item selected from the group consisting of: displacement of the moving object over time; a change in the height of a bounding box associated with the moving object over time; an average change in direction of the moving object over time from a starting position of the moving object; a standard deviation of the change in direction of the moving object over time; the average distance traveled by the moving object over time; the standard deviation of the distance traveled by the moving object over time; the difference in bounding box width between frames; The difference in bounding box height between frames; a mean absolute percentage error associated with fitting a line to direction values ​​associated with the moving object over time; a mean absolute percentage error associated with fitting a line to position values ​​associated with the moving object over time; and the mean absolute percentage error associated with fitting a line to the distance values ​​associated with the moving object over time.

15. The method according to claim 11, wherein: The features associated with the moving object detected in the video are relevant to determining whether the moving object is associated with human activity.

16. The method of claim 11, further comprising sending the alarm information and the video to a display device.

17. The method according to claim 11, further comprising: receiving feedback regarding the alarm information based on the video via an input device; as well as Based on the feedback, the machine learning algorithm is adjusted.

18. The method according to claim 11, further comprising: Second alarm information is generated when a second moving object or the moving object is detected in a second video captured by a second camera within a predetermined amount of time after the moving object is detected in the video and the moving object detected in the video is a true alarm.

19. The method according to claim 11, further comprising: Second alarm information is generated when a second moving object or the moving object is detected in a second video captured by the camera within a predetermined amount of time after the moving object is detected in the video and the moving object detected in the video is a true alarm.

20. The method according to claim 11, wherein: The metadata includes at least one selected from the group consisting of a time-stamped position of the moving object, a bounding box around the moving object, and a trajectory of the moving object.

Images