Working on an embedded systems project. I’ve purchased a FLIR style thermal camera off an older cadillac. It outputs grayscale video at 320x240 (NTSC). Ultimately, I’d like to add night vision with image recognition to my car outputting to either a display or a display that’s reflected onto a HUDWAY type translucent display.
I intitially looked at using some sort of ADC to convert the analog NTSC feed to digital and feeding it to the Pi but everything I read says a generic USB easycap will be the way to go. Real-time is important - will the easycap introduce much of a lag? Are there any simple chips that do this better?
My second question has to do with image recognition - really not very familiar with this at all but I’ve seen OpenCV projects with the Pi in the past. Performance wise, most examples I’ve seen have relatively low frame rates (under 1FPS in some cases), but these all used much higher resolution cameras. Whereas I’m trying to do just very basic shape recognition on a low resolution feed, will the Pi have enough resources to produce a reasonable (20+fps) output? I’ve got a Pi 3B+. If not, is there a board more suited for this application?
Am I missing anything obvious? I believe this should be fairly straightforward. Basically trying to do what’s done in the link below but with boxed outlines around detected objects like modern auto night vision systems. Thanks in advance!