By /

Back to: Advanced Physical Security Integration (APSI)

0

Lesson 8.4: Behavioral Analytics & Business Intelligence

Module: 8 – AI & Advanced Analytics

Prerequisites: Lesson 8.1 (AI Stack)

Estimated Time: 45–60 Minutes

1. Learning Objectives

By the end of this lesson, you will be able to:

  • Distinguish between Security Analytics (Intrusion/Loitering) and Business Intelligence (Heatmaps/Conversion Rates).
  • Configure “Time-in-Zone” logic to filter out innocent passersby from actual threats.
  • Explain how metadata allows for “Forensic Search” (finding a missing child in seconds vs. hours).
  • Apply Object Counting to retail scenarios (Occupancy Management).

2. Security Behaviors: The “Time” Factor

Simple motion detection is instant. Behavioral analytics adds the dimension of Time.

A. Loitering (The “Casing” Alarm)

  • Scenario: A person stands near the ATM for 5 minutes without using it.
  • Logic:
    1. Define a Region of Interest (ROI) around the ATM.
    2. Set a Timer Threshold (e.g., > 60 seconds).
    3. The Trigger: If Object Class = “Human” AND Time in ROI > 60s -> ALARM.
  • Value: This predicts a crime before it happens.

B. Left Object Detection (Abandonment)

  • Scenario: A backpack left alone in an airport terminal.
  • Logic:
    1. AI maps the background scene.
    2. New object enters the scene.
    3. Object Class = “Bag/Luggage.”
    4. “Owner” (Human) moves > 10 meters away.
    5. Time > 2 minutes -> ALARM.

3. Business Intelligence (Turning Cost into Profit)

This is how you sell a $50,000 security system to a Marketing Director. You aren’t selling cameras; you are selling Data.

A. Heatmapping

  • What it is: A thermal-style overlay on the floor plan showing where people walked.
  • Red: High Traffic (Hot).
  • Blue: Low Traffic (Cold).
  • Retail Use: “Why is nobody buying the items on Aisle 4?”
    • Heatmap reveals: No one ever walks down Aisle 4 because the display on Aisle 3 blocks the path.
  • Privacy: Totally anonymous. It tracks blobs, not faces.

B. People Counting (Conversion Rate)

  • The “Virtual Line”: Draw a line across the store entrance.
  • Directional Logic: Count “In” vs. Count “Out.”
  • The Metric:
    • POS Data: You sold 100 items.
    • Camera Data: 1,000 people entered the store.
    • Conversion Rate: 10%. (Now the manager knows 900 people left unhappy).

C. Queue Management

  • Scenario: The checkout line is too long. Customers are leaving.
  • Logic: Count number of Humans in “Queue Zone.”
  • Action: If Count > 5 Humans -> Send alert to Manager: “Open Register 2.”

4. Forensic Search: The Metadata Revolution

In the old days, if a client said, “Find the guy in the red shirt from yesterday,” you had to fast-forward through 24 hours of video (Scrubbing). It took hours.

New Way: Metadata Search

Because the AI is constantly tagging objects (Lesson 8.1), you can search video like Google.

  • Query: Class: Human + Color: Red + Time: Tuesday 09:00 - 17:00
  • Result: The system displays 12 thumbnails instantly. You click the one that matches.
  • Time Saved: 4 hours -> 4 seconds.
  • Note: This requires massive database storage for the metadata, separate from the video storage.

5. Safety Analytics (Slip, Fall, & PPE)

AI is now a Health & Safety officer.

  • Slip & Fall:
    • Logic: Detects a rapid change in the “Aspect Ratio” of a human box.
    • Standing: Tall vertical box.
    • Fallen: Wide horizontal box.
    • Velocity: The change happened in < 0.5 seconds.
    • Trigger: “Medical Alert.”
  • PPE Detection (Hard Hats / Vests):
    • Training: The AI is trained on neon yellow/orange pixels on the upper torso.
    • Trigger: If “Human” detected in “Construction Zone” WITHOUT “Vest Attribute” -> Warning Log.