Is the Nightdriver A-11 camera a double resolution Raytheon BST?

[Fraser]

After a lot of searching I have failed to come up with a definitive statement on the actual resolution provided by the Cadillac de Ville camera beyond the statement that it uses a 320 x 240 pixel BST sensor.

I even found a General Motors paper on the system and attach a small extract, describing the camera, here.
It states that the camera uses a 320 x 240 pixel sensor but that is as detailed as it gets on the resolution front. There was also some interesting comment on the playoff between wide angle lenses and telephoto lenses when it comes to depth of field which is important in vehicle borne pedestrian safety systems.

The system was designed for distance observation beyond the coverage of the vehicles headlamps so a relatively narrow 11 Degrees HFOV by 4 Degrees VFOV was presented to the driver.

Fraser

[Ultrapurple]

Thanks Fraser - that's very interesting. I hadn't known that BST had a critical temperature around 35°C; I'd always assumed the temperature stabilisation was to remove (reduce) one of the variables in the image processing calculations.

I can see why they went for the FoV they chose. I've often wondered whether it's possible (or useful)  to have a non-linear optical system rather like a modified fisheye that projects a distorted view onto the sensor. My idea is that the central (say) 50% of pixels form a hi-res 'fovea centralis' that provides high resolution coverage of the straight-ahead view, with the angle of view-per-pixel becoming increasingly wide as one moves off-axis. The idea is that you get a really good view in the middle but also a (much lower resolution) wide angle view at the same time from the same sensor and optical system.

I'm not sure how you'd display this - I start thinking of non-linear scan lines on a CRT, but I'm old-fashioned that way - but it would surely be possible to take a leaf out of the virtual reality headset research and come up with something suitable. Display pixels are cheap as (potato) chips compared to microbolometer pixels!

Perhaps it would work best with a wide horizontal FoV and narrow vertical range. One can envisage (say) a widescreen-type 640x240 sensor that simply didn't bother with the sky and stuff that'd be illuminated by headlights.

I am aware that many years ago at least one flight simulator used eye tracking to work out where the pilot was looking and then only render the outside scene in detail for their central field of vision. It saved an enormous amount of computer processing in a the early era of high resolution graphics. It also looked really weird to anyone else looking at the video feed because most of the image was a blur. I think there was also a system that rendered two spots - one each for pilot and co-pilot.

[Bill W]

Adding here as it is the most sensible place.

Having run the Raytheon software (liberated and linked in another topic from another forum member) to fix a couple of cores, I see it has an option for an automotive 320 x 120 image.

Images online show the Hummer H2 system (Raytheon BST) as an extreme widescreen - which now makes sense.
https://images-na.ssl-images-amazon.com/images/I/41BWNVHGK1L._SX425_.jpg
https://www.ebay.com/itm/Original-Hummer-H2-Raytheon-Nightdriver-Sales-Catalogue/303665683959?hash=item46b3e29df7:g:1WQAAOSwSB1axvJz

This also explains the different chopper wheel patterns noted by Fraser in the thread here which compares them:
https://www.eevblog.com/forum/thermal-imaging/thermal-camera-teardown-raytheon-cadillac-deville-thermal-nightvision-camera/msg1001000/#msg1001000

Might the vehicular mode be a higher speed at the sensor ?

Bill