Complete Guide to CCTV Guard AI Anomaly Detection
Last updated: April 9, 2026
This guide comprehensively explains how the CCTV Guard AI anomaly detection system works, types of anomalies detected, how to optimize detection, and steps to respond to alerts.
1. What Is an Anomaly?
In the CCTV Guard AI system, an anomaly is a condition or event observed from CCTV camera footage that deviates from normal activity and potentially requires attention or response from security personnel.
The system automatically analyzes each camera frame using artificial intelligence (AI Vision), evaluating anomalies based on location context, time, and normal activity patterns.
2. Types of Anomalies Detected
a. Intrusion / Unknown Person
- Severity: CRITICAL
- Description: Someone detected entering a restricted area or an area that should be empty at a specific time
- Examples: Person entering warehouse outside business hours, unknown person in parking area at night
- Recommended response: Verify immediately via Live View, activate alarm if needed, contact security
b. Suspicious Movement
- Severity: HIGH
- Description: Abnormal movement patterns such as loitering, peeking, or running
- Examples: Someone pacing back and forth in front of entrance, rapid movement in sensitive area
- Recommended response: Monitor via Live View, observe situation development
c. Abnormal Crowd
- Severity: MEDIUM - HIGH
- Description: Number of detected people exceeds normal limits for a specific area and time
- Examples: Crowd at loading dock, people gathering at emergency exit
- Recommended response: Evaluate situation, consider dispersal action if needed
d. Abandoned Object
- Severity: MEDIUM
- Description: A stationary object detected in a location that should be clean/empty
- Examples: Bag/package left in lobby, box in evacuation route
- Recommended response: Visually inspect the object, coordinate with field personnel
e. Camera Blocked / Tampered
- Severity: CRITICAL
- Description: Camera detects that its view is blocked or repositioned
- Examples: Lens covered, camera rotated, sudden dark view
- Recommended response: Immediately dispatch personnel to check the physical camera
f. Unusual Activity
- Severity: LOW - MEDIUM
- Description: Activity inconsistent with normal patterns based on time and location
- Examples: Vehicle entering at unusual hours, door open for too long
- Recommended response: Log and monitor, increase vigilance
3. Severity Levels
| Level | Color | Description | Response |
|---|---|---|---|
| CRITICAL | 🔴 Red | Immediate security threat | Immediate response (<2 minutes) |
| HIGH | 🟠Orange | Significant potential threat | Quick response (<10 minutes) |
| MEDIUM | 🟡 Yellow | Needs attention | Evaluate (<30 minutes) |
| LOW | 🔵 Blue | Information / note | Review during business hours |
4. How AI Detection Works
- Capture: System takes snapshots from each camera at configured intervals
- Analysis: AI Vision (DeepSeek) analyzes images to detect objects, people, and activities
- Context Evaluation: AI considers location context (warehouse, office, etc.), time, and previous detection history
- Scoring: Each detection receives an anomaly score and severity level
- Alert: If score exceeds threshold, notifications are sent through configured channels
- Evidence: Snapshots and video clips (if available) are stored as evidence
5. Optimizing Detection
a. Camera Placement
- Ensure camera faces the monitored area with a clear viewing angle
- Avoid direct light on the camera lens (backlight)
- Ensure adequate area lighting, especially at night
- Position cameras at 2.5-4 meters height for optimal coverage
b. Detection Zone Configuration
- Define critical zones within the camera frame (entrances, cashier areas, warehouses)
- Exclude areas with frequent normal movement (public roads in background)
- Set different sensitivity for different zones
c. Monitoring Schedule
- Set different detection profiles for business hours vs. after hours
- Increase sensitivity at night and on weekends
- Reduce false positives by adjusting schedules to match normal activity
6. Understanding and Responding to Alerts
a. Through Mobile App
- Open the push notification or go to the Anomaly tab in the app
- View details: snapshot photo, AI description, severity, time, camera
- Use Live View for real-time verification
- Mark as "Resolved" or "False Positive" to improve AI accuracy
b. Through WhatsApp
- Automatic alerts sent with snapshot photo and brief description
- Click link to open full details in the app
c. Through Email
- Detailed email with snapshot attachment and event summary
- Suitable for review and documentation
7. Tips for Reducing False Positives
- Ensure camera is stable and not shaking
- Configure detection zones to exclude areas with constant movement (trees, curtains)
- Mark false positives in the app so AI learns from mistakes
- Adjust detection profiles based on your location type
- Update camera firmware regularly for best image quality
8. Need Help?
Our support team is ready to help you optimize the detection system:
- PT Indo Smart Media
- Address: Jl Crusher No. 26 Kota Bandung
- Phone / WhatsApp: 08119629391
- Email: rizal.ihsan@indosmartmedia.com