# --- Loitering Logic --- (This code continues inside the while loop)

    for box, track_id in zip(boxes, track_ids):
        x, y, w, h = box
        center_point = (int(x), int(y))

        # Check if the center of the person is inside the alert zone
        is_inside_zone = cv2.pointPolygonTest(ALERT_ZONE_POLYGON, center_point, False) >= 0

        if is_inside_zone:
            # If person is inside, start or check their timer
            if track_id not in loitering_timers:
                # First time this person is detected in the zone
                loitering_timers[track_id] = time.time()
            else:
                # Person is still in the zone, check duration
                elapsed_time = time.time() - loitering_timers[track_id]
                
                if elapsed_time > LOITERING_THRESHOLD_SECONDS:
                    # Loitering detected! Trigger alert.
                    alert_triggered_ids.add(track_id)
        else:
            # If person leaves the zone, reset their timer
            if track_id in loitering_timers:
                del loitering_timers[track_id]
            if track_id in alert_triggered_ids:
                alert_triggered_ids.remove(track_id) # Optional: reset alert when they leave

    # Add visual alert text for tracked individuals who triggered the alert
    for track_id in alert_triggered_ids:
        # This part requires re-iterating through boxes to find the right one to draw on.
        # A more optimized way would be to store box coordinates with the track_id.
        for box, tid in zip(results[0].boxes.xyxy.cpu(), track_ids):
            if tid == track_id:
                x1, y1, _, _ = map(int, box)
                cv2.putText(annotated_frame, "ALERT: LOITERING!", (x1, y1 - 10), 
                            cv2.FONT_HERSHEY_SIMPLEX, 0.9, (0, 0, 255), 2)
                break

    # Display the annotated frame
    cv2.imshow("Suspicious Activity Tracker", annotated_frame)

    if cv2.waitKey(1) & 0xFF == ord('q'):
        break

cap.release()
cv2.destroyAllWindows()

# Hashtags: #SecurityAI #AlertSystem #HomeSafety

---

#Step 5: Results and Discussion

When you run the script, you will see your video with:
• A yellow polygon outlining your defined security zone.
• Bounding boxes around each person with a unique tracking ID.
• If a person stays inside the zone for longer than LOITERING_THRESHOLD_SECONDS, a red "ALERT: LOITERING!" message will appear above their bounding box.

Discussion and Ethical Considerations:
Defining "Suspicious": This system does not understand "suspicious behavior." It only follows a simple rule: person + location + time. The interpretation of this as "suspicious" is a human one.
False Positives: The system will trigger an alert for anyone who meets the criteria, including a homeowner, a delivery driver waiting for a signature, or a neighbor stopping to chat. The LOITERING_THRESHOLD_SECONDS must be carefully tuned to balance security with convenience.
Tracking Stability: In low-light conditions, the tracker might lose a person and assign them a new ID when they reappear, resetting the timer. Using a higher-quality camera or a YOLO model fine-tuned on night-time data (e.g., from thermal or infrared cameras) would significantly improve performance.
Bias and Fairness: AI models can be less accurate in identifying individuals from underrepresented groups in their training data. In a security context, this could lead to a higher rate of missed detections or false alarms for certain people, which is a serious ethical issue. This tool should be used to assist human judgment, not replace it.

#ProjectComplete #AIforGood #ResponsibleAI

━━━━━━━━━━━━━━━
By: @DataScienceN ✨

Discussion and Potential Improvements:
Real Barcode Scanner: This application works directly with a USB barcode scanner. A scanner acts as a keyboard, so when it scans a code, it types the numbers and sends an "Enter" keystroke, which perfectly triggers our returnPressed signal.
Data Integrity: We added a basic check for stock (quantity > 0). A more robust system would check if the quantity in the cart exceeds the quantity in stock before allowing the sale to complete.
Features for a Real Pharmacy: A production-level system would need many more features: prescription management, patient records, batch tracking for recalls, advanced reporting (e.g., top-selling drugs, low-stock alerts), user accounts with different permission levels, and receipt printing.
Database: SQLite is perfect for a single-user, standalone application. For a pharmacy with multiple terminals, a client-server database like PostgreSQL or MySQL would be necessary.

This project provides a solid foundation, demonstrating how to integrate hardware (like a barcode scanner) with a database-backed desktop application to solve a real-world business problem.

#ProjectComplete #SoftwareEngineering #PythonGUI #HealthTech

━━━━━━━━━━━━━━━
By: @DataScienceN ✨