3D Point Cloud Generation from Mobile Camera Images
End-to-end photogrammetry workflow from calibration to multi-view reconstruction.
This project implements a full multi-view photogrammetry pipeline from mobile-phone images: camera calibration, keypoint detection and matching, projective reconstruction, and Euclidean refinement.
Key outcomes:
- Implemented Zhang-based mobile camera calibration and robust multi-view matching.
- Best detector/descriptor pairing: KAZE + SURF with
900fundamental-matrix inliers and1799matched points. - Generated consistent 3D point cloud reconstruction with improved reprojection error after bundle adjustment.
Tech stack:
- MATLAB
- Multi-view geometry
- Feature matching
- Bundle adjustment
Links: