How to install an Ajax camera for better AI recognition

Updated

Ajax cameras offer AI-powered object detection for efficient event recording, reduced storage usage, and easy video archive navigation. This helps the surveillance system record only important events, like someone entering a room or a car passing through a parking lot.

Cameras can be configured selectively, and all processing is done in the camera itself, minimizing infrastructure load. Efficiently searching for people, cars, or pets significantly accelerates incident processing at central monitoring stations.

Ajax cameras must be correctly installed and configured for effective operation and AI recognition. Read the article to learn how to improve AI recognition and prevent false alarms.

Installation requirements

Camera positioning and testing before installation are vital to minimize false alarms and achieve the best detection results.

When choosing the installation site, consider the direction of the lens, the viewing angles of the camera, and the presence of obstacles impairing the view. Ensure that furniture, columns, plants, and decorative or glass elements do not obstruct the camera’s view.

The diagram below shows the parameters to consider when installing the camera.

  1. Installation height.
  2. Camera tilt.
  3. Detection area.
  4. Minimum detection distance.
  5. Maximum detection distance.

Ajax cameras should be mounted by a qualified professional installer.

Installation height

Cameras should be mounted at a distance of 3–4.5 m from the horizon. Do not install the camera on moving (e.g., construction cranes) or vibrating objects. The camera’s field of view must be fixed at all times.

Correct installation The camera is too high The camera is too low

Install the camera so that the minimum detection distance exceeds half the camera installation height.

Camera tilt

The camera should be tilted downward enough so that the center of the image is below the horizon. The intended path of the intruder must be perpendicular to the axis of the camera’s lens.

Correct installation Incorrect installation

Cameras should not be tilted more than 45° from the horizontal plane. The optimal tilt is approximately 30°. Some objects may not be recognized if you install the camera in a confined place or with a 90-degree tilt.

Correct installation (30°) Correct installation (45°) Incorrect installation (90°)

Object requirements

The object’s size should be less than 2/3 and more than 1/10 of the camera’s field of view.

Correct installation (the object’s size is normal) Incorrect installation (the object is too small)

Select the maximum recommended detection range according to the table.

Object type Time of the day Recommended detection range (by default) Maximum detection range (max sensitivity)
From, m To, m From, m To, m
Human Day 2 25 1 28
Night 2 21 1 23
Vehicle Day 4 35 3 44
Night 4 26 3 32

The object should be in the camera’s field of view for at least 2 seconds to be detected. So, install the camera where the object cannot appear suddenly.

The object should not be partially obscured or overlapped for the best detection results. This means that the whole object should be visible in the field of view. Object’s characteristic features, such as vehicle wheels or heads and shoulders of people, should not be obstructed by foliage, fences, or other obstacles. Otherwise, detection accuracy will be affected.

Overlapping objects in a crowded scene can decrease recognition accuracy. Thus, some objects will likely not be detected if overlapping.

Humans should be in the most upright position for the best recognition.

Lighting impacts

Avoid light sources, as the camera can be temporarily blinded if bright light shines directly at it. The most common direct light sources are the sun, headlights, and outdoor/indoor illumination. Position the camera so that those sources do not shine directly into the camera’s lens.

Do not install the camera in a place where light conditions change abruptly throughout the day (e.g., in front of a window that has direct sunlight for part of the day). Also, avoid mirrors and other reflective surfaces, such as wet roads, which can cause false alarms.

Indirect light sources, such as built-in IR light, can also affect detection. Do not place the camera too close to the wall. Otherwise, its IR lights will blind it during the night mode.

The IR illuminator must not shine on directly adjacent elements, such as walls or other objects in the near field of the camera. This leads to the overexposure of close-up shots and insufficient target illumination.

Correct installation Incorrect installation

Use wide dynamic range (WDR) technology if you position the camera in a place where lightning conditions change.

WDR enabled WDR disabled

To enable the WDR option, in Ajax apps:

  1. Go to the Devices tab.
  2. Select the camera from the list. If it is connected to the network video recorder, select NVR and tap on Cameras.
  3. Go to Settings by tapping on the gear icon twice.
  4. Go to the Image menu.
  5. Enable the Wide dynamic range (WDR) option.
  6. Adjust Lighting stabilization.
  7. Tap on Close.
  8. Tap on Back to save the new settings.

How to reduce false alarms

False alarms can be caused by:

  1. Objects obstructing the scene.
  2. Cameras positioned near reflective surfaces.
  3. Stationary objects triggering alarms (e.g., parked cars).

You can reduce false alarms by adjusting the activity zone. By configuring this option, you focus the camera on specific areas within its field of view so that the camera detects motion only in those zones. To do this, in Ajax apps:

  1. Go to the Devices tab.
  2. Select the camera from the list. If it is connected to the network video recorder, select NVR and tap on Cameras.
  3. Go to Settings by tapping on the gear icon twice.
  4. Select the Detection menu.
  5. Go to the Motion detection settings menu. Select the required area.
  6. Tap on the icon. Return to the camera settings.

Also, false alarms can be caused by other reasons:

  1. The weather can cause false alarms and influence recognition distance. Heavy rain, fog, or snow reduces such distance significantly.
  2. The presence of bugs and spiders attracted to IR lights can cause false alarms. Clear any spider webs.

Spelling error report

The following text will be sent to our editors: