Well, one possibility is using something known as Fabry-Perot filter. It allows an extremely narrow frequency to pass, due to multiple reflections and interferences inside the material. Put the light source material within this filter, and you get a laser. That’s essentially the main difference between a led and a semiconductor laser. The filter makes only a narrow band of the emission be “stuck” there, creating a feedback effect that eventually tends to infinity, and a good chunk of that power passes through the filter reflectors, which are intentionally not perfect.

Other than that, I don’t think there is a filter that could be as narrow as the line emission from vapor lamps. Maybe using metamaterials, but a laser would be so much cheaper and easier. A vapor laser would certainly get the job done, but they are large and hard to maintain.

A laser, maybe, but definitely not LEDs. Vapor/gas lamps produce the narrowest frequency bands possible, because it comes from very well defined atomic transitions (Hz range). LEDs produce frequency bands with widths in the GHz/THz range, while semiconductor lasers can maybe reach KHz if they are really good. So, unfortunately, for this type of applications, vapor lamps would probably still be needed.

Source: I work with lasers and spectroscopy.

Edit: very good idea about using non-visible light!