tips-make3D

  写真から3Dへ

沖縄の首里城や熊本城の復元に使われたらしい

• https://www.cs.cornell.edu/projects/bigsfm/
• https://lm-7.hatenadiary.org/entry/20100103/1262482038

• https://aicam.jp/tech/sfmmvs

目標

• python(pip) だけで構成する
• 屋内版はDepthMapが作れること
• 屋外版は複数の装置(マルチコンピュータでの運用)ができる様にする

構成

• https://daily-tech.hatenablog.com/entry/2019/08/04/032130
• https://daily-tech.hatenablog.com/entry/2018/01/01/133755

原理

• https://qiita.com/sobeit@github/items/d419ea39b06e43e6200f
• https://qiita.com/ringo156/items/9d4f81cd9aa70aa626ff

アプリ

• http://ccwu.me/vsfm/
• https://github.com/alicevision/Meshroom
• https://styly.cc/ja/tips/photogrammetry-rinmizouchi-meshroom/
• https://psarlin.com/superglue/
• https://colmap.github.io/
• https://github.com/colmap/pycolmap
• https://github.com/scomup/q3dviewer

カメラキャリブレーション

• http://labs.eecs.tottori-u.ac.jp/sd/Member/oyamada/OpenCV/html/py_tutorials/py_calib3d/py_table_of_contents_calib3d/py_table_of_contents_calib3d.html#py-table-of-content-calib
• https://qiita.com/ReoNagai/items/5da95dea149c66ddbbdd
• https://jp.mathworks.com/help/symbolic/developing-an-algorithm-for-undistorting-an-image.html
• https://www.qoosky.io/techs/67a3d876c4

• https://daily-tech.hatenablog.com/entry/2019/06/05/153145
• https://daily-tech.hatenablog.com/entry/2019/06/07/022607
• https://daily-tech.hatenablog.com/entry/2019/06/10/042630
• 三角測量
• https://sites.google.com/view/senmitsu/%E6%B8%AC%E9%87%8F%E7%B3%BB/

黄金分割探索
3ten no kankaku ga ougonhi (1+sqrt(5.0))/2.0

• http://ysrken.seesaa.net/article/lens-parameter-optimization-for-distortion.html

number plate
Truck Big  440x220 (275)
Normal Car 330x165 (210)

• https://info.alumania.net/info-car/%E3%83%8A%E3%83%B3%E3%83%90%E3%83%BC%E3%83%97%E3%83%AC%E3%83%BC%E3%83%88%E3%81%AE%E5%A4%A7%E3%81%8D%E3%81%95%E9%81%95%E3%81%84%E3%81%A8%E3%82%B5%E3%82%A4%E3%82%BA%E5%AF%B8%E6%B3%95/14611/

 信号機　大きさ
1250x450 (lenz 300)

• https://always3d.com/knowledge-signal-size/
• https://note.com/roadsign/n/ndf89d949eab8

 横断歩道
ゼブラ模様の白線と道路（黒い部分）の幅は、日本の道路規定（道路標示設置基準）で決まっています。
 規定例: 白線の幅は 45cm〜60cm、間隔も 45cm〜60cm。
 拘束条件: 変換後の画像で、白線の幅 W1​ と、次の白線までの間隔 W2​ が W1​:W2​=1:1 にならなければならない。

SfM/MVS

• https://github.com/alicevision/meshroom
• https://gijyutusyanokyuujitu.hatenablog.jp/entry/2018/04/29/205308
• https://aicam.jp/tech/sfmmvs
• https://github.com/cdcseacave/openMVS
• https://www.opensfm.org/docs/building.html
• https://github.com/mapillary/OpenSfM

• https://daily.belltail.jp/?p=1387
• https://daily.belltail.jp/?p=1387
• https://github.com/snavely/bundler_sfm

TEST(SfM)( 2枚の画像から3次元にする)

• https://github.com/alyssaq/3Dreconstruction

pip3 install numpy
pip3 install matplotlib
pip3 install opencv-python
pip3 install opencv-contrib-python

• https://cvml-expertguide.net/terms/cv/epipolar-geometry/
• https://www.qoosky.io/techs/24007913d6

OpenSfM

> git clone --recursive https://github.com/mapillary/OpenSfM
> cd ./OpenSfM
> python3.9 install -r ./requirements.txt
> python3.9 setup.py build 

openMVS

• https://github.com/cdcseacave/openMVS

openMVG

openSfM の他にもと調べてたら出てきた。。。

• https://github.com/openMVG/openMVG

build の際に
   #include <Eigen/Dense>
   が見つからない旨のエラーは
　　#include <eigen3/Eigen/Dense>
   にするか。。Makeをなおす。。

• https://github.com/openMVG/openMVG
• https://github.com/cdcseacave/openMVS

TEST multi View stereo

• https://github.com/trungnguyencs/MultiView-Stereo
• https://github.com/cdcseacave/openMVS
• https://github.com/trungnguyencs/MultiView-Stereo

Test カメラ位置推定

• http://blog.livedoor.jp/tmako123-programming/archives/52508207.html
• https://github.com/snavely/bundler_sfm

• http://ccwu.me/vsfm/
• https://code.google.com/archive/p/osm-bundler/wikis
• http://grail.cs.washington.edu/projects/mcba/

TESTデータ

• https://www.robots.ox.ac.uk/~vgg/data/mview/

OpenSfM

• https://mlengineer.hatenablog.com/entry/2019/09/16/150820#OpenSfM%E3%82%92%E4%BD%BF%E3%81%86
• https://www.opensfm.org/docs/index.html
• https://github.com/mapillary/OpenSfM
  • https://www.kkaneko.jp/tools/ubuntu/suitesparselinux.html

OpenMVG/OpenMVS

• https://aicam.jp/tech/sfmmvs
• http://www.regard3d.org/index.php/documentation/compiling
• https://openmvg.readthedocs.io/en/latest/software/SfM/SfM/
• http://cdcseacave.github.io/openMVS/
• https://github.com/openMVG/openMVG

OpenMVS

• https://qiita.com/fujin/items/d816a7e9b8c2577a7e37
• https://aicam.jp/tech/sfmmvs
• https://openmvg.readthedocs.io/en/latest/

• https://qiita.com/fujin/items/d816a7e9b8c2577a7e37

OpenGV

• https://laurentkneip.github.io/opengv/page_installation.html

OpenMesh

• https://www.graphics.rwth-aachen.de/software/openmesh/

深度マップ

• https://github.com/isl-org/MiDaS
• https://www.qoosky.io/techs/24007913d6
• https://dev.classmethod.jp/articles/stereo-depth-estimation-with-opencv-2nd/
• https://resanaplaza.com/2023/09/15/%E3%80%90%E5%AE%9F%E8%B7%B5%E3%80%91python-%E3%81%A8-midas-%E3%81%A7%E6%B7%B1%E5%BA%A6%E6%8E%A8%E5%AE%9A%EF%BC%81depthmap%E3%82%92%E4%BD%9C%E3%82%8D%E3%81%86%EF%BC%81/

  3Dスキャナ

• https://en.openscan.eu/openscan
• https://github.com/OpenScanEu/OpenScan
• https://github.com/LibreScanner/horus
• https://github.com/bqlabs/ciclop

イメージ・データ

• https://www.robots.ox.ac.uk/~vgg/data/mview/

RANSRAC

• tips-ransac
• https://github.com/scipy/scipy-cookbook/blob/master/ipython/attachments/RANSAC/ransac.py
• https://homepages.inf.ed.ac.uk/rbf/CVonline/LOCAL_COPIES/FISHER/RANSAC/#fischler
• https://hkawabata.github.io/technical-note/note/ML/ransac.html
• 直線の推定
• https://qiita.com/kazetof/items/b3439d9258cc85ddf66b
• https://scipy.github.io/old-wiki/pages/Cookbook/RANSAC
• NG-RANSAC

pyntcloud (点群の処理ライブラリ )

• https://github.com/daavoo/pyntcloud

平面フィッティング

• https://qiita.com/sage-git/items/f64620d18eeff8a11308

ヤコビアン

• http://math-juken.com/sikaku/jacobian/
• https://www.it-swarm-ja.com/ja/python/python%E3%81%A7%E3%83%A4%E3%82%B3%E3%83%93%E8%A1%8C%E5%88%97%E3%82%92%E8%A8%88%E7%AE%97%E3%81%97%E3%81%BE%E3%81%99/837274750/
• http://math-juken.com/sikaku/jacobian/

pip3 install autograd

• https://github.com/HIPS/autograd

Bundle

• https://qiita.com/scomup/items/f8632151712828e9625d
• https://github.com/scomup/MathematicalRobotics

  最終目標 ( SLAM )

• https://jp.mathworks.com/discovery/slam.html
• https://gijyutusyanokyuujitu.hatenablog.jp/entry/2018/05/17/044851
• https://github.com/raulmur/ORB_SLAM2

https://qiita.com/sage-git/items/f64620d18eeff8a11308

open3d

tips-Open3D

• https://blog.negativemind.com/2018/10/17/open3d/
• https://ichi.pro/python-de-tengun-kara-3-d-messhu-o-seiseisuru-tame-no-5-suteppu-gaido-60176041760954

SuperGlue

特徴点の一致

• https://psarlin.com/superglue/
• https://github.com/magicleap/SuperGluePretrainedNetwork

CloudCompare

• https://cloudcompare.org/

点群の結合

• https://qiita.com/xasdas/items/7c563888993272cec72b
• https://qiita.com/xasdas/items/87ba9ac8eb5ca440719b
• https://qiita.com/xasdas/items/552fceb0d46fc53c107c

瞳孔間距離 ( 視差の距離？？ )

• https://www.tanaka-megane.co.jp/contents/megane-select-pd/

男性 64mm
女性 62mm
* 瞳のサイズ
11.7mm +- 0.5mm