Re: [Playerstage-users] BlobFinderProxy

[Jim S. <jsu...@cs...>]

Thanks to all who have replied to this topic.

As suggested I have started writing some code to disect the color from 
blobs.color.

During one of my expeditions through the playerclient.h file the 
following was found in the comments for the blobfinder proxy:
-----------------
The latest color blob data is stored in @p blobs, a dynamically 
allocated 2-D array, indexed by color channel.
-----------------
So, it got me to wondering, is this still a true statement?  If it is 
then possibly I am not accessing the array correctly?  Looking at the 
data it doesn't always seem so, but at times it does.  Which has brought 
me to the obvious conclusion that I do not understand what a color 
channel is.  Can someone please help me with that?

Also,  what is red or green?  I guess at what point does a color fall 
into a generic category, you know, white as opposed to eggshell / 
burgundy is red, etc..., or something like that?

Thanks,
--jim




Brian Gerkey wrote:
> Jim Summers wrote:
> 
>> Richard vaughan wrote:
>>
>>>
>>> A quick look in playerclient.h (which you included in your client 
>>> code) seems to suggest that BlobfinderProxy's member blob_count is 
>>> what you want.
>>
>>
>>
>> Yep, I found blob_count, but I can only use it as an integer.  In the 
>> player-1.4 libs/headers, the num_count is an array that was set up as 
>> different channels which corresponded to the basic color. Like red=0, 
>> green=1, blue=2, so forth.  Twas very nice.  I am sure that 
>> functionality is still there, I am just having trouble finding it.
> 
> 
> With Player 1.6, the blobfinder interface changed from a list of blobs 
> sorted by (color,size) to an unsorted list of blobs.  Also, instead of 
> the color being represented by an ambiguous index (what color table is 
> being used?), the color of each blob is now a packed 32-bit RGB: 
> 0x00RRGGBB.
> 
> The BlobfinderProxy reflects these changes.  So now, to get all the 
> "red" blobs, you need to iterate throught BlobfinderProxy::blobs, 
> looking for those blobs with Blob::color set to what you consider to be 
> "red."
> 
> As you point, the old interface with sorted blobs was useful for some 
> applications.  Unfortunately, it's a really unnatural way for many 
> vision system to present their output.  It'd be a cinch to post-process 
> BlobfinderProxy::blobs into a sorted list like it used to be.  Feel free 
> to submit a patch that does this.
> 
>     brian.
> 

-- 
Jim Summers
School of Computer Science-University of Oklahoma
-------------------------------------------------