Maxence Boels

Artificial Intelligence Researcher

Pan-Tilt Tracking Camera

#ROBOTICS #COMPUTER-VISION #AI

Project Overview

A computer vision system for tracking people using a USB camera and controlling a pan-tilt servo mechanism with YOLOv8-based object detection.

Key Features

  • YOLO Person Detection: Uses YOLOv8 for robust and fast person detection
  • Real-time Tracking: Smooth tracking with Kalman filtering and position smoothing
  • Pan-Tilt Control: Arduino-based servo control with inverted pan servo support
  • Tracking Modes: Surveillance mode (keeps person in scene) and Turret mode (precise targeting)
  • Motion Compensation: Prevents feedback loops for stable tracking

Project Progress

3D Printed Turret Parts

Hardware assembly in progress and electronics integration

Initial testing of the pan & tilt turret mechanism with servos

Pan-tilt camera turret with improved stability and control

Screen recording of the turret's camera feed during operation

Technical Details

Hardware Components

  • Camera: USB 2.0 UVC camera (U20CAM-1080P-1)
  • Servos: 2x MG996R servos for pan and tilt movement
  • Servo Controller: PCA9685 16-channel servo controller board
  • Arduino: Arduino compatible microcontroller (tested with Arduino Uno/Nano)
  • Computer: Linux system (tested on Ubuntu and NVIDIA Jetson Nano)

Tracking Modes

  • Surveillance Mode: Keeps the person in the scene with stable movements using longer history buffer for smooth tracking
  • Turret Mode: Precision targeting at the center of detected bounding box with more immediate response

Advanced Features

  • Kalman Filtering: Predicts target positions, maintains tracking during occlusions, reduces jitter
  • Motion Compensation: Prevents feedback loops, distinguishes between real and apparent motion
  • Performance Logging: Comprehensive evaluation tools to analyze tracking results