There are different ways to achieve this and it's hard to say what you have seen. The "classical" and "easy" approach is to use vectors and apply rotation matrices. But for a flying device you often rotate around an arbitrary axis and not around x, y or z. That is where quarternions come in handy. They can represent a rotation around an arbitrary vector. But they are harder to understand so most teaching of those concepts is done the "classical" way. Quaternions are also numerically stable. Computers can only represent an approximation of a number (for example you can't store Pi, only the beginning). Using matrices it's happens more easily that an error introduced through these approximations changes your result a lot.

They are often used to describe rotations in the 3D world. So when you have a camera drone (multicopter) there is a good change that they are used to process the sensor inputs (gyroscope/accelerometer) and calculate the corrections needed for a stable flight.

You could even go two dimensions further and use quaternions...