I am tying to make a route finding robot. In specific if it is left in a room it can autonomously find where is the door and come out of the room. At a distance using some wireless source we tell the robot where to reach(the destination). So it has to be able to reach there by its own using some image/video processing technics. I am thinking to do like this, when it is in the room it takes a picture and using object detection it can find where the doorstep, in a similar way detects the obstacles. Now my doubt is can I do such image/video processing using PIC33F microcontroller? Does it have enough processing ability to do this? if not, what can be done to improve?
Microcontrollers don’t have GPU in them so they are kind of out-of-scope for this forum.
Jetson TK1 has 2GB RAM, 2.3GHz quad core CPU and 192 core GPU for visual computing and it can do face recognition with OpenVC with 640x480 resolution at 5 FPS.
So based on that I would say it would take an eternity to use one frame for object recognition. Not to mention that you can’t fit any of the needed code to the microcontroller’s memory. Or that nothing of the needed code exists for microcontrollers.
So, use Jetson TK1 as the brains of the robot and pic33f to control the motors through PWM and GPIO.