05_vo

visual odometry

sfm: structure from motion, more general than VO and tackles the problem of 3D reconstruction of both the structure and camera poses from unordered image sets.

Visual slam: loop closing.

• camera calibration
• feature detection
• motion estimation
  • appearance based
  • feature based(prefer)
• local optimization

2D-2D motion estimation

• epipolar geometry: 对极几何
• essential matrix:E
• fundamental matrix: F
• epipole
• epipolar line: $I,I'$
• point correspondence

essential matrix: $x\rightarrow I'$

camera model: $p=Zs=KTP_w$

$$p_1=KP\quad p_2=K(RP+T) \\ x_1=K^{-1}p_1 \quad x_2=K^{-1}p_2 \\ x_2=Rx_1+t \\ x_2^T\hat{t}x_2=x_2^T\hat{t}Rx_1=0 \\ x_2^TEx_1=0 \quad E=\hat{t}R \\ p_2^TFp_1=0 \quad F=K^{-T}EK^{-1}=K^{-T}\hat{t}RK^{-1}$$

derive $R,T$ from $E$ with SVD.obj_descriptor

reference

1. ref1
2. ref2
3. rpg_svo