Just to summarise. Thanks to everyone for their input.
The requirement was to be able to switch the frog polarity electonically, without any input from the point linkage. Applications would include 1) trailling through spring loaded turnouts, and 2) crossovers.
Specifically for 1) The point blades need to electrically bonded to their adjacent stock rail (as in the DCC friendly method) The polarity switching must be independent of the point blade linkage to avoid the "flick back" after each wheel passes through. So this tends to eliminate any solution where the polarity is swiched by the turnout motor, or using reed switches to detect the linkage movement. And anyway, when trailing through, the loco has already crossed the frog before getting to the turnout blades.
Specifically for 2) There is no linkage. So any solution using any kind of mechanical connection, point motors, switches etc is out.
It would seem that there are three possible solutions. In no particular order.
a) Use a DCC autoreversing module. b) Current sensing (Twin-T) module c) Optical detectors
a) I have a PSRev on order for another project, so I'll use that to test if this idea works. All the DCC autoreversing module diagrams I have seen connect both "output" wires to the track, and I am only going to connect one (to the frog). US$40
b) Need to research this one a bit more. Circuitron have a BD-2 block detector (current sensing). But I think that the "block" would have to overlap the frog in order to trigger in both directions. US$20
c) Might be the easiest to connect up. Would need 4 sensors (to detect traffic in both directions). Again Circuitron have AR-1CC. US$47
b) and c) would probably need to drive a relay as they are not designed to source/sink full loco power as they are usually intended for panel lights or signals.
Of course, there is still solution d). Insulated frog! Free.