Hi, currently my group has 6x Oak-1 Fixed Focus cameras that we use in robotics. When we used them previously, our pipeline was Camera (USB-C 3.1) -> Radxa Zero (GPIO UART) -> UART to RS485 conversion board -> Robot Brain.
This year, we have developed a board for our sensors and other external devices that cannot directly interface with the Robot Brain. This board uses a Teensy 4.0 as its microcontroller. We are looking into using the Oak cameras again, but having a pipeline of camera -> radxa zero -> new board -> robot brain seems like there is a lot of room for error and unnecessary latency. We were hoping to cut out the radxa zero and interface the camera directly with the teensy 4.0. I remember reading that the cameras have debugging SPI pins, however those are small surface mounts and due to the conditions the robots will be in I'm not too trusting of a stable connection. Additionally, we would need to add external memory to our board to store the model and upload it to the camera over these SPI lines, which is a lot of extra effort. I'm sure this method is possible, but there is little to no documentation on this and time is not on our side.
Therefore, we were looking into using a POE model camera (and likely upgrading our OAK-1s to OAK-D S2s), as the POE cameras allow us to store the models on the camera itself, removing our need for external storage. I was wondering if there was existing documentation or examples of people interfacing the M12 or M8 connector on the camera directly with a microcontroller, or some documentation of the protocol used by the M12/M8 connector if we need to create everything ourselves.
Thanks for your help.
~ Ben