Surveyor Robotics Journal
email:
support@surveyor.com
web:
Surveyor Corporation
rss:
Subscribe
Archives
March 2010
February 2010
January 2010
December 2009
November 2009
October 2009
September 2009
August 2009
July 2009
June 2009
May 2009
April 2009
March 2009
February 2009
January 2009
December 2008
November 2008
October 2008
September 2008
August 2008
September 2008
August 2008
July 2008
June 2008
May 2008
April 2008
March 2008
February 2008
January 2008
December 2007
November 2007
October 2007
September 2007
August 2007
July 2007
June 2007
May 2007
April 2007
March 2007
February 2007
January 2007
December 2006
November 2006
October 2006
September 2006
August 2006
July 2006
June 2006
May 2006
April 2006
March 2006
Februray 2006
January 2006
Sun, 19 Oct 2008
combining color segmentation with edge detection
Continuing the exploration of edge detection algorithms for the SRV-1 / SVS firmware, we finally had a chance to explore an earlier idea of combining color segmentation with edge detection as a possibly richer approach to feature extraction. Here are some preliminary results -
basic 16-color segmentation:
16-color segmentation combined with edge detection:
There results are very encouraging - scene features stand out much more strongly with the addition of edges. Note that for edge detection, we are still using just one horizontal and one vertical comparison (1x3) rather than a full 3x3 convolution kernel. The 3x3 approach didn't actually provide significantly better results and computation was a lot slower.
The discussion of this effort continues here -
"Surveyor Robotics Forum: new firmware with edge detect: srv-blackfin-101708"
Next step is to use this approach to begin building depth maps with the SVS.
Posted Sun, 19 Oct 2008 17:40 | HTML Link | see additional stories ...
Fri, 17 Oct 2008edge detection for Surveyor SRV-1 / SVS
A new test version of firmware with edge detection has been posted. Download link and discussion are found here -
"Surveyor Robotics Forum: new firmware with edge detect: srv-blackfin-101708"
This is just a first pass using a somewhat modified Sobel edge detector. Note that edges which are primarily horizontal will appear as cyan, while edges which are primarily vertical will appear as orange. We could take an even more aggressive approach to classification along the lines of a Canny edge detector, as we have the gradient information.
At the moment, this code only makes one horizontal and one vertical comparison (1x3) as opposed to the standard 3x3 convolution kernel, but it is comparing Y, U and V values as opposed to the standard Y-only approach. Changing to a 3x3 convolution on all 3 components will be tested next. Then we will be using this along with our color segmentation code to begin building depth maps for the SVS.
Posted Fri, 17 Oct 2008 17:54 | HTML Link | see additional stories ...