Instead of using a camera gimbal servo based solution, would you use a wide-angle camera or a head-tracking system with multiple cameras?

[wiehahuo]

I'm working on a head tracking system.
There are three systems currently known
1. Drone synchronized rotation. dji Avata
2. gimbal servo.
3. Wide-angle or multi-camera cropping
Currently searching for information, I don’t see keywords, construction methods and directions.

I learned on reddit that these 3. Wide-angle or multi-camera cropping theory can lower latency and reduce the risk of physical components and damage~
So I want to ask here if there are keywords, construction methods and directions.

 

[JasonK]

the problem with option3 and the camera cropping thing is that you would need to process the camera data to find the subset of the data that should currently be transmitted... this is going to __add__ latency, not remove it.

as far as the servo solution, unless you can move your head really fast it is unlikely that your going to notice the servo latency for moving the camera (at least if you use a link that has a high update rate).

So, the link consists of reading your head movement, converting that into signals to sent to the drone, then either converting that to gimbal movement in servos or using it to change the video processing parameters.

if your using analog cameras, processing the image before sending it, is going to add significant processing power needs on board and unless (and maybe even with) an asic specific designed to do this, your going to be adding latency (given that analog can stream straight from camera to the VTX, I believe the least amount of latency that you could add doing this would be ~17ms NTSC or ~20ms PAL as the processor would need to get a full frame of video to pick from).

if your thinking about how 360 cameras work - normally they record the whole 360, then sample at play/process time - a time were a bit of latency is going to go unnoticed.

Unless you plan on getting into designing a complicated image meshing [ to mesh multiple camera images together and deal with various bits of distortion/etc that cameras create] along with a sampling system to pick out the part to send [IE DSP/ASIC/etc design work], option3 is unlikely to be an option unless you found an "out of the box" solution. Given that even camera drones use Gimbals, I suspect the Gimbal is an easier/faster/cheaper solution then your option 3.