It’s not really that hard to figure out, unfortunately. It gets trickier for moving targets… but it’s still not that hard. These days, the easiest way is just to use a neural network trained in a simulator. You don’t even have to understand the math or even fully grep the sensor/image processing. The methods used for conventional targeting systems are actually pretty well documented.

Hobbiests are already making self landing (spacex-style) model rockets. That is a much harder problem.