3D Reconstruction from Stereo Cameras

This project implements a 3D reconstruction system using stereo vision techniques. It computes a 3D point cloud from stereo images using depth estimation and rectification methods.

---

Features

• Stereo Image Rectification: Aligns the left and right images to correct distortions and align image planes.
• Depth Map Calculation: Uses Semi-Global Block Matching (SGBM) to calculate the disparity map.
• 3D Point Cloud Generation: Reprojects the disparity map into 3D space.
• Point Cloud Export: Saves the 3D point cloud in PLY format for visualization in tools like MeshLab.

---

Requirements

Libraries

• OpenCV 4.x or higher

Install OpenCV using:

sudo apt-get install libopencv-dev

Build Tools

• A C++17-compatible compiler (e.g., GCC, Clang, MSVC).

---

Project Structure

• path/to/left_image.jpg: Left image for stereo vision.
• path/to/right_image.jpg: Right image for stereo vision.
• point_cloud.ply: Output 3D point cloud file.

---

How to Run

1. Replace Calibration Parameters

Replace the intrinsic (K) and distortion (D) parameters, along with the rotation (R) and translation (T), with values obtained from your stereo camera calibration.

2. Compile the Code

Save the code as stereo_3d_reconstruction.cpp and compile it using a compatible C++ compiler:

g++ stereo_3d_reconstruction.cpp -o stereo_3d_reconstruction `pkg-config --cflags --libs opencv4`

3. Run the Program

Provide the paths to the left and right stereo images:

./stereo_3d_reconstruction

4. View the Point Cloud

Open the point_cloud.ply file in a visualization tool like MeshLab or CloudCompare to inspect the reconstructed 3D scene.

---

Example Calibration Parameters

Replace these example values in the code:

• Intrinsic Matrix (K):

  Mat K = (Mat_<double>(3, 3) << 700, 0, 320, 0, 700, 240, 0, 0, 1);
  

• Distortion Coefficients (D):

  Mat D = (Mat_<double>(1, 5) << 0.1, -0.25, 0.001, 0.001, 0.1);
  

• Rotation Matrix (R) and Translation Vector (T):

  Mat R = Mat::eye(3, 3, CV_64F);
  Mat T = (Mat_<double>(3, 1) << 0.1, 0, 0);
  

---

Output

• Point Cloud: The output point_cloud.ply contains the 3D points reconstructed from the stereo images.

---

Notes

1. Ensure the stereo images are properly captured and aligned.
2. Tune the SGBM parameters for optimal depth map quality.
3. Use high-resolution images for better reconstruction accuracy.

---

Future Enhancements

• Implement additional stereo algorithms for disparity calculation.
• Use GPU acceleration for faster processing.
• Automate calibration and parameter tuning.

---