AI Store Suspicious Behavior Detection with Computer Vision

I recently built a Computer Vision project focused on analyzing customer behavior inside a store and identifying potentially suspicious activity.

The idea was to move beyond simple object detection and create a system that can analyze a sequence of actions.

How It Works

The system tracks people inside a store and monitors their position relative to predefined zones such as:

• Store entrance
• * Product shelves
• * Checkout area
• * Checkout lane
• * Store exit
• * Staff area

Each person receives a tracking ID, allowing the system to follow their movement throughout the video.

The behavior is then analyzed based on the sequence of events.

For example:

Normal behavior:

Tracking → Product interaction → Checkout → Exit → LEFT STORE

Potentially suspicious behavior:

Tracking → Product interaction → Product Taken → Moving toward Exit without Checkout → SUSPICIOUS

Behavior States