Skip to main content

AI Analytics Overview

Overview​

AI Analytics is one of GCXONE's most powerful features, using intelligent algorithms to analyze video footage and distinguish real security threats from false alarms. This technology can reduce false alarms by approximately 80%, significantly reducing the burden on security operators while ensuring critical alarms are processed within strict service level agreements.

What you'll learn:

  • How AI Analytics works in GCXONE
  • Detection capabilities (human, vehicle, object recognition)
  • False alarm filtering and reduction
  • SLA-driven alarm processing
  • Priority alarm classification
  • Integration with alarm workflows

Key Capabilities​

AI-Powered False Alarm Filtering​

The Problem: Traditional motion detection and alarm systems generate numerous false alarms from:

  • Animals (dogs, cats, birds)
  • Environmental factors (wind, shadows, moving vegetation)
  • Lighting changes
  • Camera movement or vibration
  • Reflections and glare

The Solution: GCXONE's AI Analytics uses advanced computer vision algorithms to analyze video frames and identify actual security threats versus false triggers. The system can reduce false alarms by up to 80%, allowing operators to focus on genuine security incidents.

Detection Capabilities​

Human Detection​

  • Person Recognition: Accurately identifies human figures in video footage
  • Bounding Boxes: Draws visual indicators around detected persons
  • Confidence Scoring: Provides confidence levels for detections
  • Movement Tracking: Tracks person movement across camera fields of view

Vehicle Detection​

  • Vehicle Classification: Identifies various vehicle types (cars, trucks, motorcycles, buses)
  • Multiple Vehicle Types: Recognizes bicycles, cars, motorbikes, buses, trains, trucks
  • Parked vs. Moving: Distinguishes between parked and moving vehicles
  • Size and Type Analysis: Analyzes vehicle characteristics for classification

Object Detection​

  • General Object Recognition: Identifies various objects that may trigger alarms
  • Custom Object Classes: Supports detection of specific object types
  • Behavioral Analysis: Analyzes object behavior patterns

Centralized Alarm Processing​

The AI Analytics system operates as a centralized processing engine:

  1. Alarm Reception: System receives alarms from various connected devices
  2. Video Analysis: AI algorithms analyze associated video footage
  3. Classification: Alarms classified as "real" (human/vehicle detected) or "false" (no threat)
  4. Distribution: Processed alarms distributed to operators based on customer preferences
  5. Human Review: Unclear cases forwarded for operator verification

Strict Service Level Agreements (SLA)​

60-90 Second Processing Window: GCXONE is engineered for early detection and timely intervention, aiming to process alarms within a strict 60-90 second window. This ensures rapid response to genuine security threats.

Automatic Escalation: If the AI cannot process an alarm within the SLA timeframe, it automatically forwards it as a "real" alarm for human review to ensure safety - no genuine threat is missed.

Priority Classification: The system automatically classifies alarms based on:

  • Detection confidence levels
  • Threat type (human vs. vehicle vs. unknown)
  • Alarm source and context
  • Historical patterns

Alarm Quad Views​

For simultaneous monitoring during incidents, GCXONE provides Alarm Quad Views that display:

  • Pre-Alarm Image: Video frame before the alarm event
  • Current Image: Video frame at the moment of alarm
  • Post-Alarm Image: Video frame after the alarm event
  • AI Analysis Result: Bounding boxes and detection annotations

This provides operators with complete context for rapid decision-making.

How It Works​

AI Processing Pipeline​

  1. Alarm Trigger: Device generates alarm and sends to GCXONE
  2. Video Frame Capture: System captures video frames associated with alarm
  3. AI Analysis: Computer vision algorithms analyze frames for:
    • Object presence (human, vehicle, animal)
    • Object characteristics (size, shape, movement)
    • Scene context (environment, lighting, conditions)
  4. Classification: Alarm classified as real or false based on analysis
  5. Confidence Scoring: System assigns confidence level to classification
  6. Distribution: Processed alarm routed to appropriate operator or workflow
  7. Verification: Unclear cases or high-priority alarms sent for human review

Supported Alarm Types​

AI Analytics processes various alarm types:

  • Motion Detection Alarms: Analyzes motion triggers for real vs. false
  • Intrusion Detection: Verifies human/vehicle presence in restricted areas
  • Line Crossing: Confirms object type crossing defined lines
  • Loitering Detection: Identifies person vs. other objects loitering
  • Tamper Detection: Distinguishes real tampering from environmental changes
  • Camera Events: Analyzes camera-related events for actual issues

Device Integration​

AI Analytics works with alarms from:

  • IP Cameras: Direct camera motion detection alarms
  • VMS Systems: Alarms from integrated VMS platforms (Milestone, Avigilon, etc.)
  • IoT Sensors: Motion sensor alarms with video verification
  • Specialized Devices: Various security devices with video capabilities

Benefits​

Operational Efficiency​

  • 80% False Alarm Reduction: Dramatically reduces operator workload
  • Focus on Real Threats: Operators concentrate on genuine security incidents
  • Faster Response Times: Real alarms processed and routed quickly
  • Resource Optimization: System resources used efficiently

Cost Savings​

  • Reduced Operator Fatigue: Less time spent on false alarms
  • Lower Operational Costs: Fewer operators needed for same alarm volume
  • Improved ROI: Better utilization of security personnel
  • Reduced Equipment Wear: Less unnecessary system activity

Enhanced Security​

  • No Missed Threats: SLA ensures all alarms reviewed within timeframe
  • Consistent Analysis: AI provides consistent, unbiased analysis
  • 24/7 Operation: Continuous monitoring without human limitations
  • Scalable Processing: Handles high alarm volumes efficiently

Use Cases​

Commercial Security​

  • Retail Stores: Filter motion alarms from customers vs. after-hours intrusions
  • Office Buildings: Distinguish employees from unauthorized access
  • Warehouses: Identify human vs. vehicle movement
  • Parking Lots: Detect vehicles vs. animals or environmental triggers

Critical Infrastructure​

  • Perimeter Protection: Verify human presence at boundaries
  • Access Control Points: Confirm vehicle types at entry points
  • High-Security Areas: Strict filtering of all motion events

Remote Monitoring​

  • Unmanned Sites: Essential for sites without on-site security
  • After-Hours Monitoring: Critical for detecting real threats during closed hours
  • Multi-Site Operations: Consistent filtering across multiple locations

Limitations and Considerations​

Processing Time​

  • 60-90 Second SLA: Alarms processed within timeframe, but may take full window
  • High Volume Impact: Very high alarm volumes may affect processing speed
  • Network Dependency: Requires video frames to be available for analysis

Accuracy Factors​

  • Video Quality: Lower quality video may reduce detection accuracy
  • Lighting Conditions: Extreme lighting (very dark or very bright) affects accuracy
  • Camera Angle: Optimal detection requires clear view of detection area
  • Weather Conditions: Severe weather may impact analysis

Configuration Requirements​

  • Proper Camera Placement: Cameras must be positioned for optimal detection
  • Detection Zones: Proper configuration of detection areas improves accuracy
  • Threshold Settings: Balancing sensitivity vs. false alarm rate requires tuning

Best Practices​

Camera Configuration​

  • Optimal Positioning: Position cameras for clear view of detection areas
  • Adequate Lighting: Ensure sufficient lighting for video analysis
  • Proper Focus: Maintain camera focus for clear video quality
  • Stable Mounting: Secure cameras to prevent vibration-induced false alarms

Detection Zone Setup​

  • Define Clear Zones: Set detection zones to focus on relevant areas
  • Exclude Problem Areas: Exclude areas prone to false triggers (trees, shadows)
  • Size Appropriately: Detection zones should be appropriately sized
  • Test and Adjust: Regularly test and fine-tune detection zones

Threshold Tuning​

  • Start Conservative: Begin with higher confidence thresholds
  • Monitor Results: Track false alarm rates and adjust as needed
  • Balance Sensitivity: Find balance between detection and false alarms
  • Review Regularly: Periodically review and adjust thresholds

Need Help?​

If you're experiencing issues with AI Analytics, check our Troubleshooting Guide or contact support.