Skip to content

Latest commit

 

History

History
48 lines (40 loc) · 1.65 KB

vitamine-e.md

File metadata and controls

48 lines (40 loc) · 1.65 KB

VITAMIN-E: VIsual Tracking And MappINg with Extremely Dense Feature Points

Author: Yokozuka

Year: 2019

Notes:

  • Extremely dense tracking of KP, indiect method
  • Tracking using local extrema instead of KP
  • novel optimisation technique: subspace GN newton to handle large number of kp
  • generate and merge mesh for entire 3D model
  • real time using only CPU

Dense Tracking

  • KLT may result with wrong correspondences over multiple views
  • Track all detected curvature local maxima: $$ \kappa = f_y^2 f_{xx} - 2 f_x f_y f_{xy} + f_x^2f_{yy} $$
  • Coarse feature matching using BRIEF on low resolution image to init dominant flow estimation
  • Affine transformation model using GN and M-estimation for dominant flow estimation ($x_i$ and $y_i$ is a feature correspondence) $$ E = \sum_i^N \rho (||y_i - (A x_i + b)||_2) $$
  • Curvature extrema tracking using the prediction $\bar{x}{t1}$ of the dominant flow and maximizing an evaluation function $$ F(x{t1}) = \kappa(x_{t1}, t1) + \lambda w ( ||x_{t1} - \bar{x}_{t1}||_2) $$

BA

  • Applying GN method to the BA objective function: $$ H \delta x = -g $$
  • With $ H = \begin{bmatrix} H_{cc} H_{cp} \ H_{pc} H_{pp} \end{bmatrix}$ diagonal blocks are sparse while $H_{cp}$ is dense
  • For a high number of KP the H matrix becomes too large
  • so the variable are partially updated using GN subspace (update variables using matrix blocks instead of handling the whole matrix)

Dense reconstruction

  • project 3D points onto the image and apply Delaunay triangulation, then propose NLTGV minimization as in FLAME to get smooth meshes
  • Mesh integration with TSDF (?)

CCL

  • Only tested on EUROC
  • RT on CPU (36 ms / frame)