False Alarm Filter is a feature designed to reduce false alarms triggered by non-target objects or events. AcuSense is a technology developed by Hikvision, a manufacturer of video surveillance equipment.
False alarms can occur when surveillance systems mistakenly detect motion or other events that are not relevant or indicative of a security threat.
For example, moving tree branches, animals, or changes in lighting conditions can sometimes trigger false alarms.
To address this issue, AcuSense devices utilize advanced algorithms and machine learning techniques to analyze video footage and differentiate between potential threats and non-threatening events.
The False Alarm Filter helps to minimize false alarms by applying specific criteria to distinguish between genuine security concerns and benign activities.
Solving the False Alarm Issues
Some false alarms can be filtered by using thermal cameras, vibration optical fiber, radar, and other equipment. The cost of the whole system is too high, and it cannot meet the requirement to find the target quickly afterward.
With the development of artificial intelligence, target classification can be realized directly based on the video screen, and then the function of false alarm filter and target retrieval can be provided, which can provide users with very cost-effective schemes.
System Architecture and Design
This scheme consists of an AcuSense network camera, NVR, and Client, and it can complete the function of target classification and false alarm filter without additional products.
Target Classification
By upgrading the hardware of conventional products, Hikvision has loaded artificial intelligence technology into the chip of the equipment.
The equipment can classify and identify the objects appearing in the picture, such as humans, vehicles, rainwater, light, leaves, animals, and so on.
The function of object classification is the basis of realizing false alarms and key target alarms.
False Alarm Filter and Quick Target Search
Through target classification, the alarm caused by animals, light, rain, leaves, or other objects can be filtered out, and only the alarm triggered by the human body and vehicle can be received, it is to realize the function of false alarm filters.
In addition, in the face of a large amount of monitoring video data, when searching for key information like people or cars, the target retrieval function is used.
The device will record the information of the human body and vehicle automatically, and solve the issue that video playback is time-consuming and easy to miss.
Setting up the Hikvision Camera to Filter Out False Alarms
Parameter configuration
Let’s take the Line Crossing Detection configuration for example.
Step 1. Enable Line Crossing Detection, draw line crossing detection lines and configure the maximum and minimum size of the target, select detection targets, configure line crossing direction and sensitivity, then click Save.
Detection line: click Detection Area, preview interface will appear a line, left click to select the detection line, moving or dragging the line can change location and length.
Note: Up to 4 detection lines can be drawn, and the target classification can be set separately for each detection line.
Maximum and minimum size: click Max. Size or Min. Size, select a point in the preview screen as the starting point, left click and drag to draw a quadrangular maximum size filter box or minimum size filter box. If you need to redraw the filter box, click Max. Size and Min. Size again, then it can be redrawn.
By drawing the maximum size and minimum size, while an object enters the alert area and its size is within the range of the maximum and minimum size, it can be recognized as a target, otherwise, it is not a target. The detection accuracy can be improved.
Detection Target: human body and vehicle can be selected for different types of target detection.
Direction: indicates the direction of the target crossing the detection line, and the direction can be set from area A to area B, from area B to area A, or both.
Sensitivity: indicates the degree of the target crossing the line, the sensitivity value = 100 – S1 / ST * 100, S1 is the area of the target that has crossed the line and ST is the actual area of the target. The higher the sensitivity, the easier it is to trigger the alarm.
Step 2. Set the Arming Schedule (it depends on the demand of the user), and click Save.
Step 3. Set Linkage Method, allows you to select Trigger Recording, Notify Surveillance Center, Upload to FTP/Memory Card/NAS, White Light Flashing, and Audible Warning (White Light Flashing, and Audible Warning are available for Sound-Light Alarm /SL models only).
Step 4. Set White Light Alarm Output (available for Sound-Light Alarm /SL models only), go to [Event]-[Basic Event]-[White Light Alarm Output], it allows to set Flashing Duration and Flashing Frequency.
Flashing Duration can be set from 1 to 60 seconds, Flashing Frequency can be set to High (interval 0.6 seconds), Medium (interval 1 second) or Low (interval 1.5 seconds).
Step 5. Set Audible Alarm Output (available for Sound-Light Alarm /SL models only), go to [Event]-[Basic Event]-[Audible Alarm Output], it allows to set Alarm Sound Type, Alarm Times and Sound Volume.
For Alarm Sound Type, 10 types of built-in alarm sounds are available, don’t support importing local audio files, click the drop-down list box, it will show the details about all Alarm Sound Type.
Alarm Times can be set from 1 to 50 times, Sound Volume can be set from 1 to 100.
Details of 10 types of built-in alarm sounds
- Siren
- Warning, this is a restricted area
- Warning, this is a restricted area, please keep away
- Warning, this is no-parking zone
- Warning, this is no-parking zone, please keep away
- Attention please, the area is under surveillance
- Welcome, please notice that the area is under surveillance 8 Welcome
- Danger! Please keep away
- (Siren) & Danger! Please keep away
Step 1: Complete parameters configuration. (Take the Vehicle as a target for instance)
Step 2: Enable target rule box display.
Step 3: Check the alarm on the network camera web preview interface. When someone crosses the line, the rule box on the web interface turns red, but at this time, the iVMS 4200 client does not show the human alarm.
Note: The rule box displayed in the web preview interface of the network camera does not make target classification judgment, the alarm information is subject to the search result received from the alarm receiver and NVR.
A comparison of the web interface with the alarm receiver (such as iVMS-4200 client), can be used to distinguish whether the false alarm filter has been performed.
The Settings via Hikvision iVMS-4200 Client
PreconditionL Check “Notify Surveillance Center” in the event linkage method.
Step 1: Add camera on Device Management of iVMS-4200.
Step 2: View the triggered alarm at the iVMS-4200 Alarm Event interface. Alarm detail information will be labeled Human or Vehicle, as shown in the figure below.
Note: The Target Type shown on iVMS-4200 is the same as the Detection Type set in the configuration interface, alarm information triggered by the unchecked category will be filtered and will not be displayed on iVMS-4200.
- Check Human on Detection Type, then the iVMS-4200 client will only display the alarm information of Target Type as Human.
- Check Vehicle on Detection Type, then iVMS-4200 client will only display the alarm information of Target Type as Vehicle
The Settings via the NVR’s Interface
Step 1: Add a network camera to the camera management interface.
Step 2: Configure Smart Event on the system configuration interface. Configuration method is similar to the Web interface, drawing crossing the detection line and the min-max size.
Check the detection targets, choose crossing line direction and sensitivity, and click save.
Notice: If you check Enable Smart Analysis, the event will be analyzed by NVR. NVR only acquires the camera stream for intelligent event analysis, and NVR does not acquire the smart event configuration of the camera.
Smart Analysis of AcuSense NVR maximum supports 4 channel 4MP cameras and sub-stream resolution of these 4 channel cameras have to be under WD1 (960*576).
If NVR does not check Enable Smart Analysis, the event will be analyzed by the camera, and NVR acquires smart event configuration and alarm information from the camera.
Configure Arming Schedule.
Set Linkage Action.
GUI interface view alarm picture
NVR local GUI interface cannot classify Humans and Vehicles at this version. The current mechanism: after checking Smart Detection, the GUI interface will display all alarm pictures, without classification between Human and Vehicle.
File management retrieves human or vehicle events
Step 1: Go to File Management, select the type of file (Human or Vehicle), Camera, and Time period to retrieve.
Step 2: Click Search to get the relevant event file, double-click the event file, and you can replay the event video or view the capture.
Note:
- Select the target image in the red box area in the upper left corner, the listed picture is a close-up view of the target;
- Select the source image in the red box area in the upper left corner, the listed picture is the original picture;
- Select the video in the red box in the upper right corner, the listed files are all video files; D. Select the picture in the red box in the upper right corner, and the listed files are all picture files; E. Select all, the image and video will be displayed together.
Installation Requirements for Intelligent Analytics
The recommended altitude of Installation is 2.5~ 5 meters, the equipment bow Angle is 10° or so, and specific adjustments are according to the environment.
According to the number of millimeters of the lens, the maximum monitoring distance is different. The table of the maximum monitoring distance for the specific number of millimeters is as follows:
Lens(mm) and Recommended max monitoring distance:
- 2.8mm: 10m, 20m
- 4mm: 15m, 25m
- 6mm: 22m, 35m
- 8mm: 30m
- 12mm: 40m, 50m
The monitoring area cannot be covered by nearby objects. Do not shoot backlight at the installation position, which will affect the image effect. The following picture is an installation sketch of the perimeter environment:
The equipment installation needs to pay attention to certain blind areas. The calculation method of blind area distance is as follows:
D: Camera monitoring blind area
H: Mounting height
α: Vertical field Angle
β: Angle of depression
Recommendations
Try to avoid getting too close. It recommends that the target be more than 3 meters away from the camera. For example, avoid the scene with lots of trees nearby.
If there is a mirror in the environment, the mirror image or shadow can easily lead to a false alarm trigger.
Adjust the camera angle during installation to avoid interference from high-brightness lights or headlights.
Domes are not recommended for outdoor scenes, IR reflection can seriously affect the accuracy of the alarm.
The scene of heavy traffic will bring a lot of perimeter alarms, such as stations, airports, theaters, etc., so it is suggested to avoid this kind of scene.
Avoid situations where personnel targets are too large. AcuSense NVR can analyze the target size between 1/16 and 1/2 of the image’s vertical size. For example, the camera’s resolution is 1080p, and the vertical size of the target should be between 64 to 540 pixels.
