MATCHING FEATURES USING FOURIER DESCRIPTORS
Gabor Barsai · 2006
A major problem for today's cartographer is in the realm of cartographic automation, where paper, or analog, maps are scanned into digital form and edited into a final cartographic product. Of the many hurdles in this endeavor, are the problems in raster to vector conversion. This problem results in an in-exact fit from the source map to the final digital product, producing line work that does not match exactly. This inexactness causes obstacles later on during the object/feature matching phase of the automation process. To deal with the problems of inexact matches of objects, a method must be used to determine shape of objects, with enough flexibility to increase matches with poor quality input. The inexactness may be represented in scale, translation and rotation and noise of true map objects. The ideal tool for studying stationary signals is the Fourier transform. In other words, stationary signals decompose canonically into linear combinations of waves (sines and cosines). In the same way, signals that are not stationary decompose into linear combinations of wavelets. (Meyer, 1993)