Here is the video in question:
https://www.youtube.com/watch?v=CbwIJMz9PAQ#t=82
I've tried using Canny and Scharr edges to reproduce the effect, but the resulting live feed is very noisy, with hairs coming out of all the edges randomly. I've tried applying a Gaussian filter to the result, but the noise is no where near as low as is shown in the video.
Any ideas?
I am unfamiliar with this product, but it does look like Canny and also using contours. To get repeatable results an external lighting source (LEDs) should be used to light the paper, or just be in a well lite room. If the lighting is correct the filtering should only require thresholding, erodes and dilates. There also might be some calibration to help like a background subtraction removing any imperfections before starting.
Detecting edges is a "simple" task until you should have to make a "real" product for market.
The biggest problem in edge detection is pre-processing the image to make the edge detection reliable and fast. In the clip you link us, even if the egde detection is "in the wild" (no pre-model are loaded since the kids can put all the object he wants on the paper), it looks an "easy" situation because the background is uniform and the environment light is good. So i think you can use something like CLAHE to improve the contrast of the image and then try to use CANNY detector to detect edges. I'm pretty confident there's no "NASA-algorithm" in the clip you link, you'll be able do it with opencv.
Asked: 2014-09-17 11:55:41 -0600
Seen: 373 times
Last updated: Sep 18 '14
