I recall work from the 1980s doing something similar.
It turns out that a Fast Fourier Transform of an object
can be used to recognize the object in an scene.<p>The result of convoluting the FFT of the "interesting object" and the scene is a "spot" that highlights all of the places the object of interest occupies in the scene.<p>I played with this, literally, in the last century.<p>Some day a bright-spot engineer at Google will re-invent it, patent it, and give it a spiffy name. Then everyone will have to be able to explain it during the Google whiteboard interview by just-graduated college students.<p>Sigh.