2011 Camera Code on Driver Station

[drakesword]

So I have been working on camera code all day. Currently I am able to track targets based off of color and determine weather or not they are in a line.

Currently the most limiting factor is network speed which is keeping the FPS to about 15.

So if I were to release this to you guys how would you want the data returned to the cRio?

Example of ways data could be returned

1) Raw centroids and sizes
2) Pole positions (calculated based off the difference in x position)
3) Raw frame difference
4)??

I plan on (once it works well enough) releasing the application and the source code.

General rules on the usage of the source:

1) You may modify and redistribute at will.
2) You should be nice and credit me for the use
3) If you make improvements that helps tracking send the modifications to me so other teams can use it as well (Remember Gracious Professionalism)

[drakesword]

PS here is a low-res teaser of what I have at the moment

[kinganu123]

What language would this be in?

[drakesword]

The core is java over processing which would be able to send data to the cRIO via udp

[MagiChau]

I plan on using reflection from infrared if possible, if not LED off of the reflection tape on the scoring peg. The Camera will detect this. I have set the camera to grey-scale since full color is unnecessary and take up more space. Maybe use the reflection detected to also detect a circle with some help from the occlusion feature. This can further be used possibly to calculate distance from that circle or help align the robot to the column.

[drakesword]

Originally Posted by ZonChau

I plan on using reflection from infrared if possible, if not LED off of the reflection tape on the scoring peg. The Camera will detect this. I have set the camera to grey-scale since full color is unnecessary and take up more space. Maybe use the reflection detected to also detect a circle with some help from the occlusion feature. This can further be used possibly to calculate distance from that circle or help align the robot to the column.

This software runs on the driver station with the camera plugged into the netgear router. The program does all of the calculations to determine where the reflective tape is and then subtracts the background (the green and red area) then the frame is run through openCV to detect the blobs and do those calculations. Finally the blobs are compared to each other to determine which ones are in a column. Other calculations coming would be angle and distance to targets.

The software then returns the target information (not the image) to the cRio where you would then do your own processing.

[Greg McKaskle]

The latest LabVIEW update also contains an example under the vision category that shows how to statistically locate the columns. It runs on the development computer and uses a folder of images as inputs, so even if you are going to code in Java or C++, you can use it to tune colors, adjust lighting, etc.

Post any questions you have about how it operates.

Greg McKaskle

[drakesword]

I will be releasing the alpha before mid-day tomorrow. Keep your eyes pealed.

[mwtidd]

Originally Posted by ZonChau

I plan on using reflection from infrared if possible, if not LED off of the reflection tape on the scoring peg. The Camera will detect this. I have set the camera to grey-scale since full color is unnecessary and take up more space. Maybe use the reflection detected to also detect a circle with some help from the occlusion feature. This can further be used possibly to calculate distance from that circle or help align the robot to the column.

also note that the tape on the circle peg is not actually circular,
on the kickoff field it was square. So personally I would not plan on using the circle in your processing, because it very well may not be a circle at competition

[divixsoft]

Could you make the output the pole positions? Also, I admire your effort, the camera code is incredibly complex and hard. When and where will you post this code?

[drakesword]

Ok everyone.

The alpha can be found on my site at http://www.satanicspider.com/frc.php

You will need to do the following to get it running

I have not tested the networking end very much (lack of time) if you run into problems let me know

*************
**IMPORTANT**
*************
Do this in this order or it wont be installed properly

1 Install openCV version 1.0 (not 1.1 pre)
2 Install openCV version 2.1
3 Make sure that .\lib\ is added to the PATH system environment variable

**For people running vista or 7**
4 copy all of the DLL files from the bin directories of BOTH openCV directories into the lib subfolder in the Axis Tracking folder.

Known issues:The pole tracking isn't finished yet but it works as it is. As you approach the poles the upper points will converge towards the middle of the image. I did not compensate for that but it is on my list of things to do.

Let me know if you have issues.

[drakesword]

Originally Posted by divixsoft

Could you make the output the pole positions? Also, I admire your effort, the camera code is incredibly complex and hard. When and where will you post this code?

I will consider that for the next release. At the moment it reports each point in each pole (x, y, size) the poles are divided by 2 bytes with a value of 255

[divixsoft]

Where is the source code?

[drakesword]

Originally Posted by divixsoft

Where is the source code?

Good observation. It will be available in the same place after I clean it up a bit ^_^

[EDIT]

Its up

[divixsoft]

what do you guys think of this code? and how do you transfer the processing from the crio to the classmate using the classes from frc?