A well designed lidar pulse is basically a lot of photons shaped roughly the size of a soccerball.
As that soccerball mass of photons starts hitting stuff, some of the photons get reflected back to the lidar instrument, which is typically in a airplane. If the object reflects enough photons, the distance to the object is recorded and called a "return". The photons that were not reflected continue on the path, and so there can be more than one return from a single pulse.
There is a whole protocol the vender will use to determine the validity of a lidar return (for example a return can be generated by a bird flying in the air).
There is another protocol for determining that the return most likely made it to the ground, and if so can be classified as a "ground" return.
While the images you see here look solid, the truth is likely only about 10% of the terrain is in the path of a lidar pulse, but still it is more than enough to identify these kinds of structures.
Here's the mind blowing thing about new lidar instruments (new meaning 6 years ago or so): They can give a precision of about 1 cm of the distance from where the return was generated to airplane. That means the clock resolution (how fast the clock runs), has to work at a granularity of the time it takes light to travel 2 cm. I'll leave it as an exercise for the reader to calculate that time interval.
Absolutely none of this would be practical without a global GPS system, which gives a very precise estimate of the position of the airplane.