Nothing really new here.
I was seeing this general sort of selective herbicide technology in the sixties. It wasn't just herbicides that affected weeds but not the crop, but for certain sorts of crop it was already to cultivate to kill weeds between the crop rows, and it was trivial to use the same control system for a herbicide spray. It was called band-spraying.
I remember seeing ways of pulsing the spray on and off to hit non-crop plants at an early LAMMA show. I think that was when the show was at Lincoln, and that would put it in the last century, but I can't find the date when it moved to the Newark showground.
One of the problems that might have been solved is how the whole system behaves. You need some way of producing a cloud of liquid particles, usually a mix of water (as carrier) and pesticide, that are the right size particles to stick to the plant, not bounce off, when they hit, and large enough not to be too easily affected by wind, or evaporate too rapidly. There may be adjuvants, such as wetting agents. And somehow with the rapidly varying flow, you need to keep the pressure right, if you're using a hydraulic nozzle.
There are other ways of generating the droplets. with more control over droplet size, but I am not sure they can be used to aim the droplets.
This is starting to look like somebody having a clever idea about image recognition, which is good, who know sweet F.A. about actual farming.
We were in at the beginning of the Green Revolution, and I was coming home from school and doing things as routine that my teachers had been telling me were impossible.
At least my teachers knew about crop rotations. They did get confused about Jethro Tull.