Hi,
I started to play around with computer vision - recognizing some patterns is great. Currently do it on my laptop (MS VB + integrated webcam)
I would like to hear your comments about using different hardware for CV.
I plan to use a good quality USB or IP webcam (with hardware compression) that is connected to Banana Pi or similar PC what is going to do some real time image processing.
Can You please suggest which ARM based PC I should get, there are so many of them out there after Raspberry Pi release. I understand that higher clock rates are my friends in this type of applications. I probably would feel most comfortable with Android or Linux based environment.
Best Regards,
Raimo
Replies
I will second the Odroid. I have used it for this exact purpose.
Basically, I had anpu Odroid communicating with the APM (Pixhawk still works) and getting data from a USB Cam (Logitech C920). The Odroid ran Ubuntu with ROS. OpenCV was used to process vision and vehicle control was then passed through ROS to a node that I made called ROSCopter. This node allowed direct control of the APM through either waypoints or PWM control.
To add, I specifically used the ODroid-U3.
Hello Interested in Quads ,
I am very interested in the project you are describing. I am working on an engineering senior capstone project, and we plan on doing obstacle avoidance with the odriod xu3 and that same camera, I would love to hear about your experience or even see the code (as rough as it might be).
I don't know if the algorithms you're going to use can be accelerated using the GPU (via OpenCL), but here are a couple of boards which came out recently which allow for a great deal of parallelism (great for certain types of machine vision). Both of these are linux / android based.
The Odroid-XU3 is an 8-core + GPU board which also has a bunch of USB ports & GPIO.
The Parallella is a dual-core board with 16 Auxillary processors for use w/ OpenCL/OpenMP. Only one USB on this one, and GPIO has an odd connector.