Frank,
I saw your question to me on JU but I can't get there to reply to it now. I see you have a discussion going here too so I'll respond here.
The heart of my system is micromachined accellerometers:
http://products.analog.com/products/...roduct=ADXL202
These accellerometers can be used to measure tilt. I am interfacing them with a microcontroller. In essence, it is polishing the ultimate turd accessory, the Inclin-O-Meter. Except I get to do a lot more things with the data and have more Inclin-O-Meters in strange places; namely, I can get data on any part of my vehicle that can exist in a plane, and can orient itself in up to 2 dimensions. The frame, axles, driveshafts, control arms, steering components, wheels, et al. I can also calculate how these different planes orient with respect to each other, which is really the most important part of the system. By using the parameters of my vehicle, I can basically reconstruct graphically what it's doing in real-time.
Then I can take it further. I could allow the driver to orient the graphical representation with respect to his own angle of inclination, or orient it with respect to the earth (flat). I plan to add infrared distance sensors to the system, so that when correctly positioned, they can notify me on my display when an object lies infront of or in back of a tire, or infront of a skid or other part, and display to me how close that object is. In addition I could add accellerometers to the wheels to determine wheel spin (and not just if wheel spin exists, but also to what degree it exists, and what forces are involved).
I could combine this data with USGS Topo data and a GPS module, and even a previous record of the trail using this system, and graphically superimpose this data on my display. To that I could add an RF module and have someone else in another vehicle some distance away see exactly where the vehicle is and what it is doing.
Ultimately, I could make such a system smart. It could determine that the vehicle is about to do an endo, and when wired into an electronic throttle control system, throttle out of a rollover if the driver doesn't respond in time. Or if one was travelling uphill and was about to roll backwards the system could shift an electronically controlled automatic transmission into neutral and let the vehicle roll back downhill instead of letting the driver use the stupid pedal to try and get it up and over when it is hopeless to even try such a move. Obviously such situations would have massive computational complexity to even work right (and work safely), but it could be taught to do that. Hopefully you are able to see where else this could go as there are an infinite number of possibilities with such a system that can do this kind of data acquisition.
Now, to take a step back into reality, I am currently only at the stage where I only have one accellerometer interfaced, and I am simply trying to filter out unwanted noise from that system (mainly vibration). That alone has proved time consuming and I hardly have time to work on this as it is. I don't have a graphical interface yet and have only really started to theorize about how I might develop that. I have several ideas thought out, ranging from just using a cheap lcd panel and nothing more than Pong-level graphics, to using an old Palm Pilot, to using an actual laptop and creating a rendered OpenGL display. Using OpenGL gives me the most freedom and capability but it would also be the most bulky and complex system. I hope to find some sort of happy medium.
Your system is far more practical and far less prone to failure than this one. I may never even get the system I have described off the ground at all. It is different in that I am less concerned about this actually being practical than seeing what I can actually learn from developing it. I infact expect NONE of it to be practical or useful, but I'm also not doing it for the gee-whiz factor. If something practical came of it, then I will probably stumble across that anyways.
As I said, this system would probably not be able to exist in place of a system like yours. Cameras provide way more visual information to the driver than my system ever could. Like I said, though, I think it might complement a system like yours nicely. That said, I don't think either of these systems could ever replace a human spotter--no way. However that doesn't seem to be the real goal with either of our projects; moreso, it is to provide the driver with more information than they would normally have.
Hope that makes sense.