Three-dimensional reconstruction of computer vision (3) (single view measurement)

2D transformation

Isometric transformation
  • rotation translation
  • Preserve shape and area
  • Usually describes the motion of rigid objects
similarity transformation
  • Add scaling features based on isometric transformation
Projective transformation
  • The intersection ratio of collinear and four-collinear points remains unchanged.
Affine transformation
  • Area ratios, parallel relationships, etc. remain unchanged
  • Affine transformation is a special projective transformation

Shadow cancellation point and shadow cancellation line

2D infinity point
  • The intersection of two straight lines can be obtained by the cross product of the two straight lines, expressed as (x 1, x 2, z) (x_1, x_2, z)(x1,x2,z)。若 z = 0 z=0 z=0 , then the point is an infinite point (Euclidean coordinates are expressed as(x 1 z, x 2 z) (\frac{x_1}{z},\frac{x_2}{z})(zx1,zx2))。
  • A point at infinity becomes a point at infinite distance after projective transformation.
  • An infinity point remains an infinity point after affine transformation.
2D infinity line
  • The set of points at infinity lies on a line, which becomes the infinity line (can be expressed as linf = [ 0 0 1 ] l_{inf}=[0 \space 0 \space 1]linf=[0 0 1])。
  • A point at infinity becomes a point at infinite distance after projective transformation.
  • An infinity point remains an infinity point after affine transformation.
line transformation

It is known that lx = 0 lx=0lx=0 , solvel ′ H x = 0 l'Hx=0lHx=0.
The derivation process is: known equation: l T x = 0 Add inverse matrix: l TH − 1 H x = 0 Ungroup: ( H − 1 l ) T ( H x ) = 0 Obtain: l ′ = H
− T l = 0 \begin{equation} \begin{split} Known equation: l^{T}x=0 \\ Add inverse matrix: l^{T}H^{-1}Hx=0 \\ Split Group: ({H^{-1}l})^T(Hx)=0 \\ Obtain: l'=H^{-T}l=0 \\ \end{split} \end{equation}Known equation: lTx=0Add the inverse matrix: lTH1Hx=0Disassembly: ( H1l)T(Hx)=0Available: l=HTl=0
The infinity line is expressed as [ 0 0 1 ] \begin{bmatrix} 0\\ 0\\ 1 \end{bmatrix} 001

  • Infinity line perspective (projection) transformation H = [ A tvb ] H=\begin{bmatrix} A & t\\ v &b \end{bmatrix}H=[Avtb] is no longer an infinity line.
  • Infinity line affine transformation H = [ A t 0 b ] H=\begin{bmatrix} A & t\\ 0 &b \end{bmatrix}H=[A0tb] is the infinity line.
Points and areas in space
  • 面: a x + b y + c z + d = 0 ax+by+cz+d=0 ax+by+cz+d=0
  • 点: x ∞ = [ a b c 0 ] x_{\infty}=\begin{bmatrix} a\\ b\\ c\\ 0 \end{bmatrix} x= abc0
vanishing point
  • The projection point p ∞ of the infinity point in the three-dimensional space on the two-dimensional pixel plane = [ abc ] p_{\infty}=\begin{bmatrix} a\\ b\\ c \end{bmatrix}p= abc
  • Shadow extinction point = camera internal parameters * straight line direction.
shadow elimination line
  • A collection of vanishing points.
  • Identifying shadow cancellation lines helps reconstruct three-dimensional scenes.

Relationship with plane normal vector:
plane normal vector = camera internal parameter transpose matrix * shadow cancellation line

infinity plane
  • Parallel planes at infinity are compared to a common line - the infinity line.
  • The set of 2 or more infinite straight lines is defined as an infinite plane.

Single view reconstruction

step

  • Calibrate camera internal parameters
  • Restore information of 3D scene
  • Reconstruction
    Disadvantages: manual selection of shadow cancellation points and shadow cancellation lines; requires scene priori; the actual proportion of the scene cannot be restored.

Guess you like

Origin blog.csdn.net/qq_44116998/article/details/132353904