\(\mathbf X_w=[X,Y,Z,1]^{\mathsf T}\)
STATION 12 · WORLD 03 CAMERA GEOMETRY
内外参与投影矩阵
观察世界坐标如何经过外参变换、归一化投影和内参映射,最终得到像素坐标。
公式推导 · K[R|t]
K · [R|t] · REPROJECTION
PROJECTION MATRIX PLAYGROUND
坐标变换流水线
世界坐标 / World Coordinate
8 points
\(\mathbf P=\mathbf K[\mathbf R|\mathbf t]\)
\(\mathbf R=\mathbf I,\ \mathbf t=\mathbf 0\)
\(\mathbf X_c=\mathbf R\mathbf X_w+\mathbf t\)
\(x_n=\frac{X_c}{Z_c},\ y_n=\frac{Y_c}{Z_c}\)
\(\mathbf K\) maps normalized points to pixels
\(s[u,v,1]^{\mathsf T}\)
像素坐标
Pixel Coordinate · ghost/current/principal
- 1世界坐标World Coordinate · \(\mathbf X_w\)
- 2外参变换Extrinsic Transform · R,t
- 3相机坐标Camera Coordinate · \(\mathbf X_c\)
- 4内参投影Intrinsic Projection · K
- 5像素坐标Pixel Coordinate · u,v