Why a Security Camera Misses People
Missed person events are often caused by scene geometry or filtering rules rather than a broken AI model. Test the exact walking path before increasing every sensitivity setting.
Start with the clip, not the slider. Determine whether the person entered the active zone at a useful size and contrast before changing confidence or sensitivity.
Check these in order
- Detection zone: confirm the person’s full path actually crosses an enabled area.
- Target size: minimum object filters can exclude distant people; overly restrictive maximum filters can also reject close targets.
- Lighting: backlight, darkness, glare and overexposure reduce usable shape detail.
- Occlusion: parked vehicles, columns, foliage and doorframes can hide enough of the person to delay detection.
- Confidence/sensitivity: lower confidence can catch harder targets but can also raise false positives.
- Notifications: verify the camera detected the event before assuming the alert pipeline failed.
Retest systematically
Walk the same route at different distances and lighting conditions. Change one parameter at a time and keep a simple pass/fail record. Maximum sensitivity is not automatically best if it floods the system with irrelevant motion.
TP-Link’s current VIGI smart-event documentation confirms that object size filters, sensitivity, human/vehicle classification and confidence thresholds all change what qualifies as a valid target: VIGI smart-event configuration.
Related: false-alert troubleshooting · identification-distance calculator · person detection without subscription.