CSP Video Analytics & AI Monitoring 3 — Questions and Answers
Question 1: In the context of CSP surveillance, what does 'scene calibration' enable in video analytics?
- Adjusting monitor brightness
- Converting pixel measurements to real-world dimensions (Correct answer)
- Setting recording schedules
- Enabling night vision mode
Correct 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.
Question 2: Which behavioral analytics application would BEST help detect a retail shoplifting event in progress?
- License plate recognition
- Crowd density mapping
- Object removal detection (Correct answer)
- Facial age estimation
Correct answer: Object removal detection
Object removal detection identifies when tagged or placed merchandise disappears from a defined area, alerting staff to potential theft.
Question 3: What is 'deep learning-based anomaly detection' in surveillance primarily designed to do?
- Enhance video color saturation
- Identify unusual patterns deviating from learned normal behavior (Correct answer)
- Reduce camera maintenance intervals
- Synchronize multiple NVR clocks
Correct 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.
Question 4: A surveillance operator notices AI analytics flagging shadows as intruders frequently. What is the BEST corrective action?
- Increase recording resolution
- Retrain the model with shadow examples labeled as non-threats (Correct answer)
- Replace all cameras with thermal units
- Disable analytics during daytime hours
Correct 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.
Question 5: Which metric best evaluates overall video analytics performance when both false positives and false negatives must be minimized?
- Frame rate
- F1 Score (Correct answer)
- Bit rate
- Signal-to-noise ratio
Correct 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.
Question 6: In AI surveillance, what is the risk of deploying a model trained exclusively on daytime footage for 24/7 monitoring?
- Increased storage costs during the day
- Poor detection accuracy in low-light or nighttime conditions (Correct answer)
- Higher frame rates at night
- Reduced bandwidth during daylight hours
Correct 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.
Question 7: Which standard or framework specifically addresses ethical AI use and bias mitigation in public surveillance systems in the US?
- NIST AI Risk Management Framework (AI RMF) (Correct answer)
- ISO 9001
- ONVIF Profile S
- FIPS 140-2
Correct 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.
CSP Video Analytics & AI Monitoring 3
In the context of CSP surveillance, what does 'scene calibration' enable in video analytics?