Procedure of foul point cloud reconstruction tools

Bundler + CMVS-PMVS:

1. Check image character using jhead.
   
2. Put all images into a folder.
   a. Possible issue: image too large. Use ImageMagick to change the size. http://blog.sina.com.cn/s/blog_68f88be10102vohm.html
   mogrify -path newdir -resize 200% *.png 这句命令的意思是将当前目录中所有的 png 文件都放大一倍，并存放到 newdir 目录中
   b. …/…/RunBundler.sh under image folder
   …/…/bin/Bundle2PMVS prepare/list.txtbundle/bundle.out
   

OpenMVG + OpenMvs:

1. Cd ~/openMVG_Build/software/SfM
2. Edit SfM_Gxxx.py and SfM_Seqxxx.py file.
   If: camera model in database. https://github.com/openMVG/CameraSensorSizeDatabase
   Else: add focal length -f 1.2*max(w, h)
   Else: add K_matrix
3. python xxx.py path/to/image path/to/output_dir
4. In output_dir, can find reconstruction_sequential or reconstruction_global directory. And then get sfm_data.bin
5. Create a new folder for dense reconstruction, cp sfm_data.bin /path/to/dense reconstruction
6. Convert to .mvs file
   openMVG_main_openMVG2openMVS -i sfm_data.bin -o scene.mvs(file.name)
   Check the results simply by Viewer scene.mvs
7. Create dense reconstruction.
   DensifyPointCloud sence.mvs
   Or ReconstructMesh scene_dense.mvs
   Or RefineMesh scene_mesh.mvs
   Or TextureMesh scene_dense_mesh.mvs
   Check the result by using Viewer

VisualSfM:

Check the following link
https://blog.csdn.net/lhanchao/article/details/52238470

OpenSfM:

https://blog.csdn.net/weixin_43896089/article/details/98173871

参看资源：
https://blog.csdn.net/lhanchao/article/details/52238470
https://blog.csdn.net/X_kh_2001/article/details/83690094
https://blog.csdn.net/moneyhoney123/article/details/78454659