Ah but you also need a validation method for assessing how well the citizens do! But I do like the idea of a 'Mechanical Turk' style of data crunching, https://www.mturk.com/mturk/welcome.
Thomas, I took a course in Computational Photography and what you are asking about is a very common task for motion tracking. I have not used R to do this and have never used ImageJ, but in the class we conducted a motion tracking exercise using OpenCV (in Python) to track people moving through security camera footage. Basically the image processing involves subtracting each frame at Tn from the 'background' frame (T0). Then you have to perform a few convolutions to remove noise and identify the thing that is moving as a blob that can be tracked. For your application, I imagine you are going to be dealing with motion and changes in scene illumination, which will make the otherwise simple subtraction a lot more difficult, possibly involving some histogram equalization or other steps. Pretty interesting though, I'd like to see how it turns out! Jonathan On Tue, Mar 27, 2012 at 10:46 AM, Sarah Frias-Torres < [email protected]> wrote: > Thomas, set up a citizen science project. Volunteers can sort through all > the images and classify the ones that have anything of interest. Those will > be the images you use in your study.Here's an example that can give you > some ideashttp://www.galaxyzoo.org/ > Sarah Frias-Torres, Ph.D. Schmidt Ocean Institute Postdoctoral FellowOcean > Research & Conservation Association (ORCA) 1420 Seaway Drive, Fort Pierce, > Florida 34949 USA Tel (772) 467-1600http://www.teamorca.orghttp:// > independent.academia.edu/SarahFriasTorres > > > > Date: Tue, 27 Mar 2012 09:06:07 -0400 > > From: [email protected] > > Subject: [ECOLOG-L] Image analysis/sorting packages? > > To: [email protected] > > > > I will be deploying digital wildlife cameras as part of my research. > These > > cameras will be out for extended periods and may collect thousands of > > pictures. Some of it will be of interest and some of it will be moving > > vegetation. > > > > 1)Does anyone have a suggestion for an [R] package that can sift through > > thousands of images and determine which ones are different from the > > "background." > > > > 2)Does anyone know if/Would anyone kindly explain how ImageJ can do this? > > > > Thanks, > > > > Thomas Parr > >
