Camera requirements for the Smoke detection module


Unlike the standard smoke detection systems, the smoke detector faces many problems regarding the scene and the background of the video image. Thus, it is not guaranteed that a smoke will be detected in all the circumstances. The smoke detector should be considered as a system that increases the smoke detection possibilities. Thereby, there can still be false alarms and situations when the detector does not react to real cases of smoke in the video camera FOV.

At the same time, the smoke detector is based on a neural network, which main principle is learnability.

If the smoke detector does not react to real cases of smoke in the video camera FOV, we suggest you record the video and send it to AxxonSoft. We will train the neural network based on the provided video, the detector pack will be updated, and future detectors operation under the similar conditions will be optimal.

The Smoke detection module can be configured on the settings panel of the Smoke detection object created on the basis of the Camera object on the Hardware tab of the System settings dialog window.

The Smoke detection module is configured as follows:

  1. Go to the Smoke detection object's control panel.

  2. In the Minimum alarm time, sec field (1) enter the period in seconds during which the repeated smoke alarms will be ignored. 

    For example, if 5 seconds is set for the Time in seconds between processed frames parameter, and there is a smoke on the image during a longer period of time, then in order to avoid receiving messages of smoke from the detector every 5 seconds, set a higher value for the Minimum alarm time, sec parameter (for example, 10 seconds). In this case, only one message about the smoke will be displayed, and all the repeated recognitions will not be displayed in the Event viewer until the time interval between the recognitions of the smoke becomes longer than 10 seconds.


  3. Set the Track alarm end checkbox (2) if it is necessary to generate an end-of-alarm event only after the expiration of the time, which is calculated as: "time of the last received alarm" + "time specified in the Minimum alarm time, sec parameter".
  4. Click the  button (3) and in the standard Windows box that opens select the neural network file with the smoke detection model.
  5. In the Detection sensitivity: 0..100 field (4) enter the sensitivity of detection – integer value from 0 to 100. This parameter has an impact on sufficient probability of smoke recognition for alarm. For example, if sensitivity is 0, than frame on which smoke is recognized with certainty more than 95% will be alarming. If sensitivity is 100, than frame with 50% certainty of smoke recognition will be alarming.
  6. In the Time in seconds between processed frames field (5) enter the time period in seconds separating frames used by algorithm for smoke analysis. This parameter has an impact on processor capacity: the less the parameter, the more capacity. Also it has an impact on operation speed of algorithm: the less distance between frames. the faster required number of frames will be collected for decision and alarm, if it's required.

  7. In the Number of frames for analysis and output field (6) enter minimum number of frames for detection analysis before alarm generating. The more the value, the more certain the result of detection operation. At the same time, if value is too large, the short-time smoke can be missed.

    Multiplication of Time in seconds between processed frames * Number of frames for analysis and output parameters characterizes time period (in seconds) at the end of which after smoke alarm will be triggered.


  8. In the Device (7) drop-down list select the device where the smoke detection will operate:

    1. Auto — automatic device selection.

    2. CPU — central processing unit (not recommended).

    3. NVIDIA GPU (depends on the number of installed NVIDIA video cards).

  9. Specify the detection surveillance area on the video image:
    1. Click the Setup (8) button. The Detection settings window will open. 
           
    2. Click the Stop video (1) button to capture the video image.
    3. Click the Area of interest (2) button.
    4. Specify area on which smoke recognition will be detected (3).
    5. Click the OK (4) button.
  10. Click the Apply (9) button.

Configuring the Smoke detection module is complete.