Good Tech Article

I first saw machine vision as a kid when I worked in the local glass factory. They mass produced glass soda bottles and had a machine vision system for bottle inspection and compressed air would blast the reject bottles off the line real quickly. It was impressive for 1980. I am sure now its amazing.
It is much the same: a way to detect "bad" items on the transport line, and a way to get them off the line.
Detection is: optical (example: detect color of a soda can - remove it if wrong color), optical with picture analysis (aka "machine visioon", detect wrong shape, missing content, etc), opto-mechanical (example: light-sensors that detect full / empty containers), weight (detects empty containers) and so on. Sound could also be used, but is rarely seen in the wild, mostly due to automated assembly lines producing lots of noise.
Removal is: mechanical (arm, bolt, lever) or force (compressed air...).
It's not particularly new, in general. That particular Union Pacific project is probably new, and probably a new development of the concepts, but the railways have been using general machine vision in various forms for quite a long time now. A reasonably famous implementation is Network Rail's NMT / "Flying Banana" (it's extremely yellow, and the same class of train was used as the previous generation of the "Flying Scotsman") in the UK; a 4000 hp, 125 mph track measurement/monitoring train. It is jammed full of cameras, lasers, racks of computers, and other tech; and engineering people. It is constantly patrolling most of the UK network looking for track defects, and monitoring track condition.