There isn't actually any technology available that can be effectively put towards solving this issue.
Just stop to think about what it is this technology would entail before you keep posting.
OK. Here is the system outline. I'm stepping in as one of TJB's minions.
1) You have a stump camera that is wide angle and watches the crease from a low vantage point - probably installed low in the stump. The current Stump cam, might actually be sufficient with an adjustment to capture the crease area in field of view
A software algorithm determines the boundaries of the crease visually. This is simply mapping areas by colour and can dyamically change as the crease loses altitude from bowlers feet pounding it lower and lower.
Likewise, the software determines when each foot fall lands on the ground. This can be done by mapping distortion in the soul of the foot, or even 3d mapping. These are shape recognition algorithms that exist and happen in an instant.
2) Another camera is watching the bowler from side on. For simplicity and economy, this might as well be the cameras that watch for run outs. It determines when the ball left the hand.
3) For each ball, from stump cam, software determines when the bowler's foot lands and whether the heal lifted - it will observe the entirety of the crease boundary area beneath the foot if it lifted or landed over the line.
-condition a - foot never reaches line - end of program
-condition b - foot lands on line - compare with ball release video
-condition c - foot lands completely over line - compare with ball release video to determine if ball released before foot lands (ie. back foot release)
4) For each ball software determines when the ball left the hand from the side on camera
5) Compare 3 & 4
6) Beep or Don't Beep.
It all comes down to how much resolution the camera has. Current software can find and mark out lines of thickness in pictures - that is easy. Current software can determine when those lines stop being fully visible (foot on the line).
This system relies on a clearly painted crease. Being visual it can account for the ground shape changing - as ground gets pounded in. It can adjust for batsman scratching out portions of the crease by filling in missing gaps in the line.
The lower the camera is placed, the more accurate it can be in determining if the heel has lifted. However, currently the standard is to judge this from about 5 to 6 feet high (umps eyes). Anything below this should be acceptable.
The software part is not that difficult. That is what software engineers do. Sure, it will take a few days observing a bowler to iron out all the false hits with a number of recognition states.
A) Works regardless if a bowler is bowling over or around the wicket, and regardless of where the batsman stands and how he moves - solved
B) Can identify the moment a front foot lands, and doesn't confuse that with the bowler's backfoot or follow through. - solved
C) Isn't a permanent fixture in the ground (because not all grounds would pay to install/maintain it, and if it cannot be used in every single game of a series/tournament, it's unlikely to be used at all) - solved
D) Doesn't interrupt the game (get in the way of bowlers, batsmen, fielders) - solved
E) Will not break or get damaged over the process of a game, or during bad weather - solved
F) Is accurate and affordable - solved
G) Doesn't require an additional human being present at the ground to operate (you may as well just have a guy staring at the crease whole game via a pair of binoculars calling no-balls instead) - solved