UnCanny: Exploiting Reversed Edge Detection as a Basis for Object Tracking in Video

UnCanny edge detector paper

My second paper from my postdoctoral work at the Biosurvey has been published titled: UnCanny: Exploiting Reversed Edge Detection as a Basis for Object Tracking in Video.  This paper, published in the Journal of Imaging discusses the novel edge detection algorithm I made for the LunAero project.  You can view this open access at the publisher page or email me for a preprint.  If you’d like to read how we collected the video in this paper, see my previous LunAero post.  This paper goes over the steps used to create the UnCanny edge detector.  Why is it called the UnCanny edge detector?  Simple, it executes the steps from the famous Canny edge detector in reverse!  The big change here is that the detector runs across sequential images.  When it is run this way, it is capable of detecting very small objects that other filters may miss. This was very advantageous to the LunAero project, as we needed a way to spot the super small birds in our video footage.  Many other methods had trouble accomplishing this.  I run this method on more than just bird-brained video in the paper.  Snapshots in the paper show how this can be applied to small object tracking in other fields.

It’s pretty neat, and now my inbox is filled with requests to review Computer Vision papers!  Yippee!


Graphical Abstract for the UnCanny Edge DetectorFew object detection methods exist which can resolve small objects (<20 pixels) from complex static backgrounds without significant computational expense. A framework capable of meeting these needs which reverses the steps in classic edge detection methods using the Canny filter for edge detection is presented here. Sample images taken from sequential frames of video footage were processed by subtraction, thresholding, Sobel edge detection, Gaussian blurring, and Zhang–Suen edge thinning to identify objects which have moved between the two frames. The results of this method show distinct contours applicable to object tracking algorithms with minimal “false positive” noise. This framework may be used with other edge detection methods to produce robust, low-overhead object tracking methods.