Nym answered well on all your points but how do you implement a system that detects things on tracks when in some places the rats and mice are clamouring everywhere over the tracks and the signalling?
Since I'm not going to entertain someone trying to escalate my mood as a result of my pre-emptive measure to stop people trying to turn round and personally attack "Oh, you don't know what you're on about." I'll answer Clip and jonmorris0844 who have asked valid questions rather than surrounding them within a rant.
Lasers cannot be used, the platforms are not straight and lasers have a tenancy to go in a straight line if not contained within something, that would kind of defeat the object, unless you want to bounce an array of lasers over the platform edges up and down it, but that presents the other issue that is if you want to get a clear signal from one end to the other, you'd need between 20 and 50mW of lasers to do it, and this is not safe for use around the general public, at least if they want to retain the use of their eyesight after looking at a deflected beam.
Infra Red motion detectors would not work in LUL sub surface or deep level stations due to the high wind rates and temperature differences within these airflows, this would require a dangerously low sensitivity to change to be set on any IR-M or IR-I sensors, resulting in it being unsafe.
Ultrasonic would be affected by the traction motor controllers of a service on the opposite platform. (That operate with a switching frequency between 10kHz and 40kHz, but tend to have super-harmonics all over the place up to 150kHz, and if you want to do it on the Jubilee Line with their GTO Thyristor motor controllers that operate between 100Hz and 15kHz, good luck...
To answer clip, you can detect and there is software that can be used to detect intrusion into a visualised 3D zone that can be programmed into a computer, the software exists, I've seen it in use, an evolution on it was sold as "Hawk Eye" for Tennis. But this software is not in it's maturity, and required a very high processing capability to operate, and still has a reaction time of approximately 0.5 to 5 seconds, now, in an environment such as I designated the use of this system appropriate where sudden changes are not likely and persons are usually well behaved (by means of fear or compliance) a 2 second reaction time was not inappropriate, as the system was also equipped with ultrasonic sensors for proximity that triggered an all stop.
The human reaction time to a major event is usually around 0.2 to 0.05 seconds, depending on the individual, so the lowest standard human is still twice as fast as this computer.
But this brings me neatly round to the main reason why I would not implement one of these systems at Deep Level and Sub Surface stations; cost. I mentioned previously that it requires a great level of processing power, and the equipment used by the company designing the initial systems is not available to consumers or businesses in the UK. So it would require a bank of 1 server per camera, (I'd estimate approximately 18 high quality 100Hz cameras required for a single platform, again, costly), bvut would also require a server 'blade' for every single input to the decision making process to paralell process all variables to make a decision regarding the approach speed, or stop command to an incoming unit, including data fed from the unit it'self. You'd then need two sets of cores to make their own decision, each set consisting of five with four over five ruling internaly and different algorythms, simply for the stop/go command, but beyond that to make the system more reliable and prevent someone sticking their hand out to cause an emergency stop (that it would) an even more complex system. By this point we've filled a 64U server cabinet to handle all of this data, still with a poorer response time than a well trained human.
So is it worth implementing this system (that can be done now), due to the cost. No.
Especially when a good alternative exists anyway to prevent passengers jumping infront of trains in single stock environments, Platform Edge Doors, these cost less than the system described above. But then you come into the world of CBA (there are better things to spend £180,000 per station on) and the major assumption I have made throughout all of this...
The trains can talk to the Platforms.
Back to cost again, there aren't really any 'hot spot' station with high suicide rates, so it would need to be fitted to all or none. Pushing this up to a billions project.
Onto...
I am not sure how you'd best monitor for obstructions and expect you'd need a range of sensors, possibly also a camera (and then night vision) to perhaps have software analyse the 'threat'.
You might have simple radar of some sort, then perhaps thermal imaging to see if the object is perhaps a living creature, then you'd want to perhaps have an idea of what the object is (is it a large concrete block, or a branch from a nearby tree) and so on.
Of course you'd probably be able to build such sensors right now (not sure how big it would all be, especially if a series of separate devices connected together as against one single sensor with all of these features) but I doubt it would be cost-effective, and would also need months or years of 'real world testing'.
Far better to seek sensors to alert a driver, and keep the driver as the proper eyes with the ability to make an analysis of a danger far quicker than computer in most situations.
I've already discussed software with the associated costs so I don't need to do it again.
Radar is an interesting idea, but I can see a fair few problems with implementing a radiosensitive system on LU with the likes of 1992 and 1972/1973 stock kicking around, (also the new stock but to a lesser degree) the amount of electromagnetic interference kicking around, even from the passengers would be enough to send any such system haywire, and if it where sensitive enough, result in severe damage to any such systems within months of installation.
Thermal Imaging would require the same decision making software as the visual light sensors, but would be subject to masses more interference from the hot air flows. (As previously mentioned)
Yes, you can build all this now, and yes, you're correct, it really isn't cost effective.
And thanks to the press nowadays it's even less cost effective as any systems retro-fitted to DL or SS stations would be required to be fitted to all of them, the, "TfL chose not to fit a lifesaving peice of kit to the one station where someone jumped" is too harder hit in the press.
Last sentence is exactly my own thoughts, aside from keeping the driver in the loop, fitting something to a train is much easer (and cheaper) than fitting it to a station...
Everything we've already discussed gets much cheaper when you fit it to a train, reduced install costs, the only 'issue' is finding the space.
See...
It's not just Jay_g's posts I pick to pieces...
PS: If anyone can think of a better system I'd love to hear about it, honestly, I'm not even being sarcastic here. (So long as you're prepared to discuss it rather than just say, "Oh, this will work." for the sake of gaining virtual brownie points in a forum then not be able to stand up to scrutiny of that idea.)