Skip to content

Instantly share code, notes, and snippets.

@seaside2mm
Last active December 21, 2022 06:06
Show Gist options
  • Save seaside2mm/c91d97840f561a0deebe84b5a152b510 to your computer and use it in GitHub Desktop.
Save seaside2mm/c91d97840f561a0deebe84b5a152b510 to your computer and use it in GitHub Desktop.
[多视图几何] #cv

定义:射影映射是$IP^2$到它自身的一种满足下列条件的可逆映射h: 三点$x_1,x_2,x_3$共线当且仅当$h(x_1),h(x_2),h(x_3)$也共线

  • 中心投影

2D Projective Geometry

ref: https://www.cvg.ethz.ch/teaching/3dvision/2021/slides/class02eth21.pdf

2D primitives (points, lines, conics)

点线的齐次表示(Homogenous coordinates)以及一些性质

  • 叉积

  • Point-Line Duality

理想点与无穷远直线

  • Transformations of lines and conics

A hierarchy of transformations

Projective linear group - Affine group (last row(0,0,1)) - Euclidean group (upper left 2x2 orthogonal)
- Oriented Euclidean group (upper left 2x2 det 1)

  • 2D Projective Transformations

Transformation of Points and Lines

Example

  • 2.12 removing projective distortion

3D Projective Geometry

2D单应

直接线性变换算法

Cost Function

代数距离

几何距离

  • Symmetric transfer error
  • Reprojection error

RANSAC

2D单应

  • 直接线性变换算法

  • Cost Function

http://zhaoxuhui.top/blog/2019/11/07/robotics-perception-note-3-1.html

特征应该具备的性质

具备检测的可重复性(Detection Repeatability)或者说在不同影像中的检测不变性(Detection Invariance)。 为了能匹配正确的点对,我们也希望特征具有描述不变性(Descriptor Invariance)

SIFT SIFT尺度空间的建立过程:将原始影像与不同参数的高斯核做卷积,即可得到不同模糊程度的影像。 影像金字塔的建立过程。每当高斯核的σ变为原来的2倍时,对影像进行降采样。

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment