ML4360学习笔记--2.1 Primitives and Transformations(一)

Primitives

geometry primitives释义(来自GPT):

In computer vision and computer graphics, the term "geometry primitives" refers to fundamental geometric shapes or elements used to construct more complex objects or scenes. These primitives serve as the building blocks for creating and representing objects in a digital environment.

Geometry primitives typically include basic shapes such as points, lines, curves, polygons, and 3D primitives like spheres, cubes, and cylinders. They provide a foundation for modeling and representing objects in a coordinate space.

2D Points

表示方法

2D points有两种等价的表示方法 =>

  • inhomogeneous coordinates(非齐次坐标)
    • 表示为一个二维向量:\bold x = \begin{pmatrix} x \\ y \\ \end{pmatrix} \in \mathbb{R}^2
  • homogeneous coordinates(齐次坐标)
    • 表示为一个三维向量:\bold{\widetilde{x}}= \begin{pmatrix} \widetilde{x} \\ \widetilde{y} \\ \widetilde{w} \\ \end{pmatrix} \in\mathbb{P}^2
    • 2D Projective Space\mathbb{P}^2=\mathbb{R}^3\setminus\{0,0,0\}
    • equivalent homogeneous vectors:homogeneous vectors that differ only in scale are considered equivalent and define an equivalent class(成比例/平行的齐次向量是等价的,构成一个等价类)

inhomogeneous与homogeneous坐标之间的转换

  • inhomo => homo
    • 将非齐次坐标转化为齐次坐标,只需扩充最后一个维度,并将其值设为1
    • 得到一个特殊的齐次向量——augmented vector/增广向量
      • \bold{\bar{x}}= \begin{pmatrix}x \\ y \\ 1 \end{pmatrix}= \begin{pmatrix}\bold{x} \\ 1 \end{pmatrix}
    • 每个inhomogeneous vector对应其augmented vector的整个等价类
      • \bold{\bar{x}}= \begin{pmatrix}x \\ y \\ 1 \end{pmatrix}= \begin{pmatrix}\widetilde{x} \\ \widetilde{y} \\ \widetilde{w} \end{pmatrix}, \widetilde{x}=\widetilde{w}x,\widetilde{y}=\widetilde{w}y
  • homo => inhomo
    • 一般齐次向量转化为增广向量
      • \bold{\bar{x}}=\frac{1}{\widetilde{w}}\bold{\widetilde{x}}
    • 取前两个维度,即为对应的非齐次向量
      • \bold{\bar{x}}=\begin{pmatrix}\bold{x} \\ 1 \end{pmatrix},\bold{x}=\begin{pmatrix}x \\ y \end{pmatrix}

Ideal points/Points at infinity

  • 定义:一类特殊点,齐次空间中最后一个维度为0的点(二维齐次空间中\bold{\widetilde{x}}=\left ( \widetilde{x},\widetilde{y},0 \right )^T,\widetilde{x}\widetilde{y}\ne0)称为ideal points/points at infinity
  • ideal points/points at infinity在同维度的非齐次坐标系中不存在对应的inhomogeneous vector

几何意义

ML4360学习笔记--2.1 Primitives and Transformations(一)_第1张图片

  • \bold{x}\bold{\bar{x}}在齐次空间中\widetilde{w}=1平面上的投影
  • 为什么定义等价类:将齐次空间原点看作相机中心,一系列平行平面\widetilde{w}=w看作投影平面,augmented vector\bold{\bar{x}}=\left ( \widetilde{x},\widetilde{y},1 \right )^T所在直线看作一条相机光线,则该直线与这些平面的交点代表三维物体上的同一个点,是等价的,而这些交点的坐标即为一组等价的齐次坐标(一组平行向量,differ only in scale)
  • 为什么n维投影空间去掉n+1维欧式空间的原点:从上述几何意义来看,n+1维欧氏空间的原点可以与投影空间中任意一点等价,因此没有意义
  • 如何理解points at infinity:这些点满足\widetilde{w}=0,可以看作等价于无穷远处的点\bold{\widetilde{x}}=\left ( \infty,\infty,\widetilde{w}^* \right )^T,\widetilde{w}^*\ne0

3D Points

表示方法

  • inhomogeneous coordinates
    • 表示为一个三维向量:\bold{x}= \begin{pmatrix} x \\ y \\ z \end{pmatrix} \in \mathbb{R}^3
  • homogeneous coordinates
    • 表示为一个四维向量:\bold{\widetilde{x}}= \begin{pmatrix}\widetilde{x} \\ \widetilde{y} \\ \widetilde{z} \\ \widetilde{w} \end{pmatrix} \in\mathbb{P}^3
    • 3D Projective Space:\mathbb{P}^3=\mathbb{R}^4-\{0,0,0,0\}

你可能感兴趣的:(ML4360,学习,笔记)