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 900 fundamental-matrix inliers and 1799 matched 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: