3D_Colmap.md

COLMAP Pipeline

Note: We just give an example pipeline for COLMAP usage. You should change some hacked path as your own file path in your workspace.

• Folder Structure

├─created
│  └─sparse
│   +── cameras.txt
│   +── images.txt
│   +── points3D.txt
├─dense
├─images
│   +── images001.jpg
│   +── images002.jpg
│   +── ...
└─triangulated
    └─sparse

• Feature Extraction

colmap feature_extractor --database_path database.db --image_path images

• Update Intrinsic

python3 database.py --database_path /home/albert/UAV_Planning/Recon/planner_0831_2_spatial/database.db

• Feature Matching

colmap exhaustive_matcher (or other matcher in gui) --database_path database.db

• Mapper

colmap mapper --database_path database.db --image_path images --output_path triangulated/sparse

• Triangulation (Not necessary)

colmap point_triangulator --database_path database.db --image_path images --input_path created/sparse --output_path triangulated/sparse

• Dense Reconstruction

If the sparse from point triangulation

colmap image_undistorter --image_path images --input_path triangulated/sparse --output_path dense
colmap patch_match_stereo --workspace_path dense
colmap stereo_fusion --workspace_path dense --output_path dense/fused.ply

If the sparse from camera poses directly

colmap image_undistorter --image_path images --input_path created/sparse --output_path dense
python3 dense_cfg.py
colmap patch_match_stereo --workspace_path dense --PatchMatchStereo.depth_min 0.0 --PatchMatchStereo.depth_max 10.0
colmap stereo_fusion --workspace_path dense --output_path dense/fused.ply