Ghost Detections in People-Counting Sensors

Meet Butlr

Discover what spatial intelligence can do for you.

Thank you! Your submission has been received!

Oops! Something went wrong while submitting the form.

What are ghost detections?

Ghost detections are spurious sensor readings that an algorithm interprets as a person or movement when no actual person exists. They can appear as single false counts, persistent phantom tracks, or intermittent noise that inflates occupancy metrics.

Why they matter:

They distort analytics and downstream decisions such as HVAC scheduling or retail conversion metrics.
They erode trust in sensor systems and create extra operational overhead to investigate anomalies.
They can lead to unnecessary interventions, such as triggering alerts or adjusting building systems.

Common causes by sensor type

Radar (mmWave) sensors

Multipath reflections: radar pulses reflect off walls, glass, or metallic surfaces and return delayed echoes that can appear as phantom objects.
Side-lobe and secondary reflections: strong reflectors create false angle or distance estimates.
Environmental motion: moving fans, curtains, or HVAC components generate Doppler signatures that can be misinterpreted as people.

Depth cameras and stereo vision

Reflections and specular surfaces: shiny floors and glass can produce erroneous depth readings.
Occlusion and silhouette artifacts: partially visible people or nearby objects produce fragmented depth blobs that confuse tracking.
Lighting variations (for structured light systems): IR interference from sunlight or other sources introduces noise.

Thermal / heat-based sensors

Thermal reflections: hot or reflective surfaces can create misleading heat signatures under some conditions.
Ambient temperature gradients: HVAC vents or sunlit areas may alter local heat patterns, creating ghost hotspots.
Low resolution: coarse thermal imaging can merge nearby people into single blobs or produce spurious small hot spots.

Vision cameras (RGB)

Shadows and lighting changes: moving shadows can be mistaken for motion.
Reflections in glass or polished floors: mirrored images multiply apparent people.
Background motion: screens, signage, and dynamic displays can introduce false motion.

Wireless/BLE/Wi‑Fi sensing

Multipath and signal fluctuation: changes in RF propagation create perceived motion.
Device phantom presence: lingering MAC addresses or probe traffic can be interpreted as people.

Detection and mitigation strategies

Most robust solutions use a combination of hardware choices, installation practices, and software filtering. Use multiple layers of mitigation to address both transient noise and systematic false positives.

Hardware and installation best practices

Mounting and orientation: place sensors to minimize direct view of reflective surfaces (glass, polished metal) and to discourage multipath. Elevation and angle adjustments can remove common reflection paths.
Coverage planning: avoid placing sensors where HVAC vents, rotating equipment, or screens produce repeated environmental motion.
Choose the right sensor for the space: some technologies are inherently less sensitive to specific ghost causes. For instance, anonymous heat-based sensors often avoid the reflective and optical issues that affect cameras and depth sensors.

Signal preprocessing

Background subtraction: maintain a slowly updating model of the static scene and filter out persistent non-human heat or motion sources.
Elevation gating: for sensors that provide elevation or height estimates, discard detections outside an expected human height range.
Temporal filtering: require minimum duration or continuity for a detection to be counted (debounce short spikes).

Tracking and model-based filters

Multi-frame tracking: use object tracking to maintain and validate tracks over time; discard short-lived tracks that do not exhibit human-like motion.
Motion and trajectory heuristics: apply speed and acceleration thresholds consistent with human movement to reject unlikely signatures.
Confidence scoring: combine multiple features (size, persistence, motion pattern) into a confidence score and set thresholds tuned per environment.

Sensor fusion

Combine complementary sensors: fuse thermal + radar or depth + passive infrared to leverage strengths of each modality and reduce modality-specific ghosts.
Cross-validation: require corroboration from two independent sensors before declaring occupancy in sensitive applications.

Regular calibration and monitoring

Periodic recalibration: update background models after layout changes (new furniture, seasonal sunlight changes) to prevent new ghost sources.
Health checks and analytics: monitor for sudden shifts in baseline counts and investigate environmental changes or sensor drift.

Case study: reducing ghost targets in an open-office deployment

A mid-size office experienced persistent false counts near a glazed partition and an HVAC diffuser. Steps taken:

Site survey: identified multipath-prone geometry with glass and a low ceiling over the diffuser.
Hardware adjustment: sensors were remounted higher and angled to reduce direct reflection into the sensor field of view.
Software tuning: applied elevation gating and required a 2-second continuity threshold for new detections.
Background update: enabled adaptive background modeling to learn the diffuser's thermal signature.
Results: false positives dropped by over 85% and event logs showed improved correlation with badge entries and manual head counts.

This approach—combining installation changes, software filters, and adaptive models—produced reliable reduction without sacrificing sensitivity to genuine occupancy.

Why sensor choice matters for privacy and reliability

Sensor selection affects both accuracy and privacy. Camera-based systems can deliver high granularity but raise privacy concerns and are susceptible to optical artifacts. Radar offers good motion sensitivity but is prone to multipath ghosts. Heat-based sensing provides anonymous occupancy detection by using thermal signatures rather than identifiable images.

Heat-based, anonymous sensors tend to:

Reduce privacy risk because they do not capture facial or identifiable visual imagery.
Be less sensitive to visual lighting and reflections that plague cameras.
Produce lower false positives in certain reflective or visually noisy environments, though they require careful handling of thermal gradients.

No single technology is perfect for every environment. The best result often comes from matching sensor strengths to the specific installation challenge and combining modalities when necessary.

Deployment checklist to minimize ghost detections

Conduct a site survey to identify reflective surfaces, HVAC outlets, and moving equipment.
Choose sensor mounting height and angle to avoid direct reflections and limit sightlines to noisy areas.
Apply elevation gating and minimum-duration thresholds in software configuration.
Enable background subtraction and adaptive models that can learn static environmental signatures.
Use multi-sensor fusion in high-criticality areas where false positives have high operational cost.
Schedule regular recalibration after layout or seasonal changes.
Monitor baseline metrics for sudden shifts and investigate quickly.

FAQs

How can I tell a ghost detection from a real person?

Look for inconsistencies: ghost detections are often short-lived, lack coherent motion, repeat at fixed locations, or correlate with environmental events such as HVAC cycles or lights. Cross-checking with other sensors or manual observation helps confirm.

Can firmware or software fixes eliminate ghosts completely?

They can dramatically reduce them but may not eliminate all ghosts, especially those caused by complex multipath or physical reflections. Combining fixes—installation, preprocessing, tracking, and fusion—provides the best results.

Which sensor types are most susceptible to ghost detections?

Depth cameras and optical systems are sensitive to reflections and lighting; radar is particularly prone to multipath. Thermal systems and wireless sensors have their own unique ghost causes but often offer advantages in privacy and certain environments.

Should I use multiple sensor types?

For critical spaces where accuracy is essential, combining sensor types often yields the most reliable outcome. Fusion strategies allow one modality to compensate for another's weaknesses.

Final recommendations

Start deployments with a thorough site assessment and tune both hardware and software to the environment. Favor layered mitigation: good mounting and coverage planning, robust preprocessing, tracking heuristics, and sensor fusion when needed. For privacy-sensitive or reflection-prone spaces, consider anonymous heat-based sensing as a compelling option that balances reliability and anonymity.

If ghost detections are affecting your analytics or operations, schedule a diagnostic assessment with a sensor specialist. A focused investigation typically finds a small set of actionable changes that deliver large improvements in accuracy and trust.

Conclusion

Ghost detections are a common challenge across people-counting technologies but can be effectively managed with a layered approach that combines careful installation, signal preprocessing, robust tracking, sensor fusion, and ongoing monitoring. Tailor sensor choice and mitigation strategies to the specific environment to maximize accuracy while preserving privacy.

Ghost Detections in People‑Counting Sensors: Causes, Mitigation, and Best Practices