That is a moderately long story. I needed something that I understood top-to-bottom for my PhD. At that time (2008), paparazzi was the best-in-class, and I was fortunate to help design and build the first booz2 quadrotors. When I returned from France, I needed a slower moving target to base my PhD work on, so I forked the paparazzi code and continued to work on it.
In 2009/10 I worked on the early pixhawk system at ETH. The original onboard code was an amalgamation of wasp and paparazzi, the comms protocol design was from wasp. Once I returned from ETH I could no-longer access pixhawk hardware so persisted in using wasp. In the recent years pixhawk has advanced tremendously in capability; the comms protocol has become more standardised (and is now known as mavlink).
Pixhawk is a great project with a focus on visual control. It is well supported at ETH, both financially and enthusiastically. This is more support than I could ever muster at University of Canterbury, and one of the golden rules of open-source software is to go to the active and supported project!