pic: Tracking in Depth with the Microsoft Kinect

[faust1706]

loading photo ...

[bvisness]

I assume you're going to connect this to your code for vision tracking somehow. How are you going to do that? (Genuinely curious!)

[faust1706]

haha. funny how the left ball is an excercise ball. anyways. Yeah, we plan on finding velocity of a ball, and autonomously going to where it will go so we can retrieve it in the least amount of time. (this is will be used in autonomous to find and shoot teammate's balls in case they dont shoot.)

Tracking the hotzones and the targets will be pretty easy. There is plenty of documentation on tracking reflective tape and yellow leds.

Next, we are going to track a friendly robot, calculate their velocity, and then lead them in a pass so they will intercept the ball and they dont have to do anything but keep moving at their velocity.

Lastly, we are hoping for autonomous catching, but that is pushed back on our to do list because of the complexity and because a driver can do it fairly easily.

[nightpool]

Any write up for this? Our team has looked into doing this in the past, but we were stymied by a requirement to have a laptop computer on the robot, to deal with the Kinect USB drivers.

[faust1706]

Originally Posted by nightpool

And write up for this? Our team has looked into doing this in the past, but we were stymied by a requirement to have a laptop computer on the robot, to deal with the Kinect USB drivers.

We used a single board computer, the O-Droid X2 in 2013, and this year we are using at least one O-Droid XU. They are very small and lightweight, but can manage about 27 fps if you write your code efficiently. Last year we were averaging 28-29 fps during matches. Our whole vision system will weigh less than a pound this year.

[nightpool]

Originally Posted by faust1706

We used a single board computer, the O-Droid X2 in 2013, and this year we are using at least one O-Droid XU. They are very small and lightweight, but can manage about 27 fps if you write your code efficiently. Last year we were averaging 28-29 fps during matches. Our whole vision system will weigh less than a pound this year.

Awesome! I assume you're using the OpenKinect libfreenect? Are you doing any image processing on top of libfreenect? Would you mind sharing your code, for other teams to possibly use?

[yash101]

Originally Posted by nightpool

Awesome! I assume you're using the OpenKinect libfreenect? Are you doing any image processing on top of libfreenect? Would you mind sharing your code, for other teams to possibly use?

He's probably using OpenNI, with OpenCV. That's what he typically posts about

[Abhishek R]

For reference, in 2012 987 used the Kinect quite well.

http://www.team987.com/wp-content/up...dTheKinect.pdf

[faust1706]

Originally Posted by yash101

He's probably using OpenNI, with OpenCV. That's what he typically posts about

this is just opencv with libfreenect A problem that we have discovered tonight is the plated metal on the wall of the targets. It will give an inaccurate reading of distance to the wall with a range finder or a depth map.

[drakesword]

Originally Posted by Abhishek R

For reference, in 2012 987 used the Kinect quite well.

http://www.team987.com/wp-content/up...dTheKinect.pdf

Google detected malware on your host ... are you running ads on your site maybe?

http://www.google.com/safebrowsing/d...987.com/&hl=en