47657Re: [SeattleRobotics] Re: Robomagellan cone color(s) and cmvision
- Feb 23, 2013Orange is a unique color?
Agreed, meant to be easily seen.
Not used much in nature perhaps?
Not in nature but used by people's clothing.
From: KM6VV <KM6VV@...>
Sent: Friday, February 22, 2013 12:17 AM
Subject: Re: [SeattleRobotics] Re: Robomagellan cone color(s) and cmvision
Thanks very much for the tips. Hadn't heard of BGR. I did get my cmvision running, but not great results from my first experiments. I will try to add some of your steps.
Orange is a unique color? Not used much in nature perhaps?
Thanks and 73!
On 2/21/2013 9:38 PM, Will Smith wrote:Hi Alan,I use OpenCV for my cone finder implementation. OpenCV is an included component of ROS as well. I demoed it at the SRS meeting last month.I wrote the prototype in python but will probably port it over to C++ for easier integration with libv4l (using a webcam.)A brief rundown of what I do. Typing on my tablet so please excuse brevity!- Acquire frame and perform lens correction. OpenCV uses BGR instead of RGB. Convert to HSV. In my app I display the HSV value at the cursor position. A color picker... in this case I choose cone color. I also have sliders to choose tolerances for the next stage.- Using the color and three tolerance bands I call the opencv inRange function. This is a thresholding operation that will isolate the cone(s) color. Note that red wraps around the H axis at 255. I then threshold again to a 1 bit image.- Perform a dilation to make unified, measurable blobs.- Acquire contours. Throw out anything too small or large for expected range. I also check center of mass for the contour to isolate cone shape, along with size ratio checks.- measure blob height and calculate estimated range.Hope that helps. This method is pretty robust and flexible, thanks to the cone's unique color!Will
- << Previous post in topic Next post in topic >>