Does this project spark your interest?

Become a member to follow this project and don't miss any updates

Oculus Rift featured Crane control

Mount 2 cameras on a few servos, glue them to a crane. Read the position of the crane and display it into the field of view of the Rift.

977 0 9 5

This project was created on 06/30/2014 and last updated 4 months ago.

After I tried the Oculus Rift the first time in 2013 I immediately wanted to build something up with it. Relatively fast I decided to develop some kind of camera system which is mounted on 3 servos. The head movement of the user should be followed by the cameras.

So I started to build up a first prototype which was using a single camera. After it worked way better than I had imagined I bought 2 wide angle webcams which are now giving me also a stereoscopic view.

As I wanted to develop this project a bit further I made it an university project. For this we needed a possible industrial application. We decided to mount the system on top of a model toy crane. The crane can be controlled using a Joystick or gamepad and the current positions are displayed in the field of view of the user.

I also implemented a color based object recognition which can be used to estimate the distance to an object, also additional data is displayed once an object is recognized.

The Software is written in C#. I use a PSoC5 and some other controllers to establish a USB-UART connection to the crane/servos.

As the Oculus Rift is distorting the image I also had to distort it to compensate it. It would use way to many resources to do that on the CPU. Therefore I implemented a pixel-shader to distort the images on the GPU.

  • 1 × Dickie Mega Crane

Project logs
  • The final video

    4 months ago • 0 comments

    Check to see the final version.

  • Putting it together

    4 months ago • 0 comments

    Today we put everything together and gave it a try. 

    Its so much fun to use this thing and everything works quiet well.

    We have just a few issues.
    - As the cameras of course can not cross-eye (dont know how to say that in english), you will see very close object double.

    - We have some problems with the performance. With Object Recognition turned on, we have ~ 70-80% CPU load on a quad-core machine. I had to rearrange some thread priorities to stop the GUI from freezing. Its working but I want to optimize it. I maybe need to rethink how to process the images, maybe I also will try another framework for the cameras.

View all 2 project logs


Similar projects