Camera Installation Best Practices for Ai Analytics

-----------------------------------

Affected Roles: Installer, End-User Administrator

Related DW Products: DW Blackjack Ai Servers & Appliances | DW MEGApix Ai Cameras

Last Edit: October 16, 2026

-----------------------------------

Overview

Artificial intelligence (Ai) analytics — such as object detection, line crossing, intrusion zone, and loitering — require cameras to capture a sufficient level of detail for the analytics engine to process targets reliably. This applies whether the analytics run on a DW Blackjack Ai server or appliance (using standard IP cameras) or on a DW MEGApix Ai camera (where analytics run directly on the camera itself).

The physical installation of a camera — its mounting height, tilt angle, field of view, and lens selection — directly determines how much usable detail the analytics engine receives. Cameras that are mounted too high, angled too steeply, or pointed across too wide an area may fail to generate Ai events consistently, even when the camera's video feed appears normal to an operator.

This article outlines the installation practices that produce the most reliable Ai analytics performance.

Related Resources

• DW MEGApix Ai — DRI Overview
• DW MEGApix Ai — Analytic Rules Setup

Why Camera Placement Affects Ai Analytics

Ai analytics engines detect and classify objects — people, vehicles, and others — by analyzing pixel patterns within the camera's image. For this to work reliably, the target must appear large enough in the frame to provide the analytics engine with distinguishable detail.

When a camera is mounted too high, too far, or its field of view is too wide, targets appear too small within the image. The analytics engine may not generate events even though the video feed looks usable to a human observer. This is a key distinction: a camera image that appears clear and detailed to the eye can still be insufficient for Ai analytics if the targets do not occupy enough of the frame.

DRI Standard and Ai Detection

The DRI (Detection, Recognition, Identification) standard defines the minimum level of image detail - measured in pixels per meter (PPM), required for various levels of target analysis. For analytics, the most relevant threshold is "Detection", rated at 25 PPM, which represents the minimum detail needed for an analytics engine to determine that a human or vehicle is present.

Analytics-based detection ranges differ from standard DRI ranges because the analytics engine detects objects differently, depending on their type. Vehicles produce a larger and more distinct visual profile than people, so the Ai engine can detect them at greater distances with the same lens configuration.

Camera Mounting Height Considerations

Mounting height is one of the most significant factors that can affect Ai analytics performance. The higher a camera is mounted, the steeper its downward angle must be to cover the intended detection area, which compresses the apparent size of targets in the frame and reduces the pixel density the analytics engine can work with.

For reliable Ai analytics performance across both person and vehicle detection, cameras should generally be mounted within the following ranges.

Real-World Example

The following scenario illustrates how mounting height affects Ai analytics performance and how it can be corrected:

1. A facility deployed cameras at two heights: 15 ft and 30 ft, all with fixed wide-angle lenses.
2. Cameras at 15 ft generated Ai analytics events consistently at near 100% accuracy.
3. Cameras at 30 ft produced no Ai events, despite functioning normally as standard security cameras.
4. One 30 ft fixed camera was replaced with a varifocal camera. The lens was zoomed in to narrow the field of view and bring the detection area closer in the frame.
5. After the lens adjustment, the replacement camera began generating Ai events at near 100% accuracy.

The camera's mounting height did not change. Reducing the field of view with the varifocal lens increased the apparent size of targets in the frame, giving the analytics engine sufficient pixel detail to process events reliably.

Camera Angle and Orientation Considerations

Vertical (Tilt) Angle

The vertical tilt angle of the camera — the downward angle from the camera to the detection area — affects how targets appear in the frame. Steep downward angles cause targets to appear foreshortened and smaller, which reduces the detail available to the analytics engine.

For most Ai analytics applications, the camera should be tilted to position the primary detection zone in the center-to-lower portion of the frame. Avoid configurations where the camera is looking nearly straight down, as this significantly degrades Ai analytics performance for person detection.

Horizontal (Pan) Angle

The camera should be aimed so that the primary detection area — where targets are expected to appear — is positioned within the center of the frame rather than at the edges. Analytics engines process the full frame, but targets at extreme edges of a wide-angle image can appear distorted, which may reduce detection accuracy.

Where possible, orient the camera so that targets are moving across or toward the camera rather than away from it. Person and vehicle analytics generally perform best when the target's profile is presented toward the camera.

Camera-to-Target Distance and Field of View

The combination of mounting height, tilt angle, and lens focal length determines the camera's effective coverage area and the size that targets appear in the frame. These three factors must be balanced together:

• A wider field of view covers more area but reduces target size within the frame.
• A narrower field of view (longer focal length or zoomed-in varifocal) covers less area but increases target size, improving Ai detection accuracy.
• A greater distance between camera and target reduces the apparent target size and may fall below the minimum threshold for reliable Ai detection.

When the detection area must be covered at a greater distance or from a greater height, a varifocal or telephoto lens should be used to compensate by narrowing the field of view.

Lens Selection Consideration

Lens selection is directly linked to Ai analytics performance. The focal length of the lens determines the field of view and the magnification of the scene, both of which affect how large targets appear in the camera's image.

Fixed vs Varifocal Lenses

Ai-Based Detection Range by Lens Configuration

The following table reflects the Ai-based detection ranges measured for DW MEGApix Ai cameras with motorized zoom lenses. Detection ranges for Ai analytics differ from standard DRI values because the analytics engine evaluates object characteristics specific to person or vehicle profiles.

2.7 – 13.5 mm Motorized Zoom Lens

5 – 50 mm Motorized Zoom Lens

Scene and Environment Considerations

Lighting

Ai analytics require adequate and consistent scene lighting to produce reliable results. Insufficient or highly variable lighting reduces image quality in ways that may not be obvious on a live view monitor but can significantly degrade analytics performance.

• Ensure that the primary detection zone is adequately illuminated during the hours when Ai analytics are expected to operate.
• Avoid placing detection zones in areas where lighting transitions abruptly between bright and dark regions within the same frame. High contrast between lit and unlit areas can cause the camera's exposure to shift, which may temporarily suppress Ai events.
• If the camera will be used in night mode or low-light conditions, verify that IR illumination or supplemental lighting covers the full detection zone. Ai analytics performance in complete darkness without IR illumination will be poor.
• Wide Dynamic Range (WDR) should be used with caution in scenes with backlit areas. In some configurations, WDR processing can reduce the contrast detail that the analytics engine uses to classify objects.

Scene Clutter and Background Motion

The analytics engine evaluates motion and object profiles within the detection zone. Scene conditions that introduce irrelevant motion or reduce object contrast can increase false events or suppress valid detections.

• Avoid pointing cameras at areas with constant background motion, such as trees in wind, flags, moving water, or busy roadways, unless the analytics detection zone is carefully configured to exclude those areas.
• Ensure that the detection zone is configured within the camera or Blackjack Ai interface to focus on the area of actual interest. A detection zone that covers the entire frame will include all background movement as potential analytics triggers.
• Consider the thermal environment if cameras are used in industrial or outdoor settings. Heat sources, steam, or reflective surfaces can introduce visual artifacts that affect analytics accuracy.

Camera Stability

Cameras used for Ai analytics must be mounted on stable, vibration-free surfaces. Camera movement — even subtle vibration from HVAC equipment, passing vehicles, or wind — can cause the analytics engine to generate false events or suppress valid detections by incorrectly attributing motion to target objects.

• Use rigid mounts. Avoid flexible poles, temporary clips, or surface-mount hardware that may flex under wind or mechanical vibration.
• If mounting on a structure that experiences vibration (such as a rooftop HVAC platform or industrial equipment housing), consider relocating the camera to a more stable surface.

Recommended Camera Settings

In addition to physical installation, certain camera settings affect the quality of the image delivered to the Ai analytics engine. The following settings are recommended as a starting point. Depending on the camera model and scene conditions, adjustments may be necessary.

Exposure and Gain

• Set Auto Exposure to ON to allow the camera to adapt to changing scene brightness.
• Set the maximum Shutter Limit to approximately 1/960 – 1/1000. A minimum shutter speed that is too slow will introduce motion blur on moving targets, reducing the detail available for analytics processing.
• Set the maximum Gain Limit to 24 dB as a starting point to balance noise and image clarity. In scenes where overexposure occurs, reducing the gain limit to 9 dB may improve results.

Day/Night Mode

• Set the Day & Night mode to Auto. The camera must be able to transition into Night mode when ambient light drops below the appropriate threshold. Locking the camera into Day mode in low-light environments will degrade image quality and analytics performance.

Wide Dynamic Range (WDR)

• Use WDR settings with caution. In scenes with significant backlit areas or high contrast, WDR can improve overall image appearance but may reduce the local contrast detail that the Ai analytics engine uses to classify objects. If analytics performance is inconsistent in a WDR-enabled scene, consider reducing the WDR level or disabling it in the analytics camera.

Focus

• Set focus to Auto. If the image appears soft in the detection area, use Manual focus to sharp-focus the target zone, then return to Auto. An out-of-focus image — even one that appears acceptable on a monitor — can substantially degrade Ai analytics accuracy.

Region of Interest (ROI) and Detection Zone Configuration

Using the camera's Web UI or the DW Blackjack Ai Engine Client, configure the analytics detection zone to cover only the area of actual interest rather than the entire camera frame. A well-defined detection zone:

• Reduces the likelihood of false events from background motion outside the target area.
• Focuses analytics processing resources on the zone where detection accuracy matters most.
• Improves the signal-to-noise ratio of analytics events, making valid alerts easier to act on.

Installation Checklist

Use the following checklist when planning or reviewing a camera installation for Ai analytics use.

______________________________________________________________________________

For More Information or Technical Support

DW Technical Support: https://www.digital-watchdog.com/contact-tech-support/

DW Sales: sales@digital-watchdog.com | www.digital-watchdog.com

Copyright © All rights reserved. Specifications and pricing subject to change without notice.