🎯 Object Tracker - Pose Estimation with PnP and SIFT

This project estimates the 6DoF (position + orientation) pose of a planar object (e.g., gate) using a webcam or RealSense camera. It uses SIFT feature matching, homography, and Perspective-n-Point (PnP) pose estimation.

---

🧩 Modules Overview

📷 tracker.py - Real-Time Object Pose Estimation

• Allows user to select 4 corners of a planar object (e.g., gate).
• Extracts SIFT keypoints from selected region.
• Matches keypoints in real-time against webcam input.
• Uses Homography + PnP to compute:
  • Rotation (Yaw, Pitch, Roll)
  • Translation (Tx, Ty, Tz)
• Displays:
  • Projected object outline on current frame.
  • Orientation and position (in cm) on screen.
  • Matched features between frames.

📌 Uses:

• cv2.solvePnP with IPPE method for stable results.
• Calibrated camera matrix (for RealSense or webcam).
• Draws matches and overlay info live.

---

🛠️ function.py - Utility Functions

Contains helper functions:

• rotational_to_euler(R): Converts 3x3 rotation matrix to Euler angles.
• draw_quadrilateral(): Overlays selected corners as a quadrilateral.
• get_four_points(): GUI-based point selector for user to click 4 corners.

---

🧪 camera_calibration/zhang.py - Camera Calibration (Zhang's Method)

• Detects checkerboard corners in calibration images.
• Uses OpenCV's calibrateCamera() to compute:
  • Intrinsic matrix (Camera Matrix)
  • Distortion coefficients
  • Rotation & translation vectors for each image.
• Outputs all matrices and vectors.

📝 Note:

• You can use the camera_calibration/capture_img_for_zhang.py to capture checkerboard images.

---

📂 Folder Structure

📂 object_tracker/
├── tracker.py
├── function.py
├── README.md
├── camera_calibration/
│   ├── zhang.py
│   └── capture_img_for_zhang.py

✅ Requirements

• Python 3.x
• OpenCV (cv2)
• NumPy
• PyRealSense2 (optional for RealSense camera)

Install dependencies:

pip install opencv-python numpy pyrealsense2