Ok, the squirrel cam is running at www.squirrelCam.org I detect a squirrel not by frame differences, ie. motion detection but by analyzing the spatial frequency content of a small region of interest (ROI) of each image in the place where the squirrel would be present. Because the camera is zoomed in tight and the resulting aperture is wide open, the focus depth of field is very small resulting in the far background being out of focus. The spatial frequencies present in that defocused background are low and when the squirrel is present the spacial frequencies of the well focused rodent are high. Taking the FFT of the ROI reveals the frequency spectrum of the ROI. Taking the absolute value of the FFT tells me what the relative magnitude of the spatial frequencies are. Low frequencies means bacground and high frequecies means foreground focused object. I set a mangitude threshold for the presence of a squirrel and presto, when he is there I save the image and post it. And that's my technical rant of squirrel detection :)
Somehow I've got to get inside the mind of the squirrel.. what are his habits, what makes him tick. Squirrelentology, the science of squirrel study, will hopefully be advanced by this webcam.
Making progress on the squirrel cam. Still testing the squirrel detection software.
-Mike
No comments:
Post a Comment