Saturday, 11 December, 2010 § Leave a comment
So i just finished a major code overhaul for the Optosensor, i now have pixel mosaic, interpolation, gaussian filtering and
tracking (correction blob detection working on tracking at the moment) all done via OpenCV. The image below, shows the raw, scaled output in RGB, the next image is Grey Scale (Not really needed since the the IR sensors only detect a narrow band or IR only), Next is the tracked output, and finally is the reference image used for background subtraction (going clockwise from top left).
Thursday, 9 December, 2010 § Leave a comment
So i’ve continued my work on the optosensor, when ever i’ve had time; i’m now at a point where i thought i’d talk about what i’ve been doing during the past few weeks. The past few weeks i’ve mainly been working on is the clean up the code on pc side i.e. updating the tracker for the most part, i’ve cleaned up the code separated it into various classes each representing a specific portion tracker i.e. pixel mosiac, serial communication, tracking, and tuio. There have been various improvements to the back end of code where i’m getting the data, how i’m processing it, etc. Not all that interesting what is interesting is that i’ve managed scale up the raw 16×8 pixel image via standard bicubic interpolation and i’ve further smoothed out the data by applying a Gaussian filter to the interpolated data. Below is what is looks like when i have 5 test points.
To explain it a little better these are the raw pixels that are active pixels (1,1), (1,16), (1,8), (16,8), (4,4), (4,5), (5,4) and (5,5) but anyway below you can see both the interpolated scaled output as well the non interpolated scaled output. Here in this test case I’ve interpolated the data set by a factor of 100 using standard bicubic interpolation, and applied a Gaussian kernel of 10 x 10. However in the final release the interpolation factor will be turned down to 10 and the Gaussian filter will be optional.