1 Why do we need it?

1.1 Modeling: new object via primitive composition

cube & cube man

旗帜，石柱等都是同一个几何，平移到不同的位置进行显示；

1.2 Viewing: (3D to 2D) projection

1.3 Animation composed of various transformations

Pixar - Luxo Jr.YouTube · Cooper Hewitt2 minutes, 9 secondsOct 8, 2015

1.4 Sensor Calibration

KITTI Odometry Dataset

http://www.cvlibs.net/datasets/kitti/

calibration   = pd.read_csv('data_odometry_gray/dataset/sequences/00/calib.txt', delimiter=' ', header=None, index_col=0)

calibration

Before understanding what all these numbers in the calibration files are, let us take a look at the schematic of the sensor configuration provided by the KITTI team.

1. four cameras: a pair of grayscale stereo cameras & another pair of RGB stereo.
   1. P0,P1,P2,P3 are the camera projection matrices: 3x4 matrix;
   2. The distance between a left and right stereo camera is called baseline, which is 0.54 m in this configuration.
2. Tr: Velo-to-Cam
3. IMU-to-Velo: kitti没提供
4. Cam-toCamRect & CamRect-to-Image: kitti没提供