Video Analytics & AI Monitoring Flashcards

In the context of CSP surveillance, what does 'scene calibration' enable in video analytics?

Answer: Converting pixel measurements to real-world dimensions

Scene calibration maps pixels to real-world units (e.g., meters), allowing analytics to accurately measure object size, speed, and distance.

Which behavioral analytics application would BEST help detect a retail shoplifting event in progress?

Answer: Object removal detection

Object removal detection identifies when tagged or placed merchandise disappears from a defined area, alerting staff to potential theft.

What is 'deep learning-based anomaly detection' in surveillance primarily designed to do?

Answer: Identify unusual patterns deviating from learned normal behavior

Deep learning anomaly detection trains on normal activity patterns and flags statistically unusual events that may indicate security threats.

A surveillance operator notices AI analytics flagging shadows as intruders frequently. What is the BEST corrective action?

Answer: Retrain the model with shadow examples labeled as non-threats

Retraining the model with properly labeled shadow examples as negative (non-threat) samples reduces false positives caused by shadows.

Which metric best evaluates overall video analytics performance when both false positives and false negatives must be minimized?

Answer: F1 Score

The F1 Score is the harmonic mean of precision and recall, providing a balanced measure when both types of errors are operationally important.

In AI surveillance, what is the risk of deploying a model trained exclusively on daytime footage for 24/7 monitoring?

Answer: Poor detection accuracy in low-light or nighttime conditions

A model trained only on daytime images lacks exposure to low-light conditions, leading to degraded detection performance at night.

Which standard or framework specifically addresses ethical AI use and bias mitigation in public surveillance systems in the US?

Answer: NIST AI Risk Management Framework (AI RMF)

The NIST AI Risk Management Framework provides guidance on identifying and mitigating risks including bias and fairness issues in AI-based systems.