CameraInfo

# Projection Matrix

First Heard about this from Alvin Li. Itβs quite simple.

The projection matrix converts a 3D coordinate to a 2D coordinate, combining the Intrinsic Matrix $K$ with the Extrinsic Matrix $T$, i.e.

$P=K[Rβ£t]$

You can add a column at the end composed of 0s to make it a 3x4 matrix.

### In ROS

In ROS, they call this the projection matrix, even though really its just the calibration matrix with a shift for the baseline.

So the left 3x3 matrix is simply the Calibration Matrix, and there is $T_{x}$ and $T_{y}$ which are 0 if it is the left camera.

You need to understand

$P=βf_{x}00β0f_{y}0βc_{x}c_{y}1βT_{x}T_{y}0ββ$

http://docs.ros.org/en/melodic/api/sensor_msgs/html/msg/CameraInfo.html