Digital correlation in photogrammetric instruments

Information on the image power spectrum gives a basis for rational design of digital image correlation systems in that the sampling interval, scanner spot size, and expected sharpness of the correlation function can be estimated. The latter estimate is possible because the correlation function and power spectrum are Fourier transforms of each other. Furthermore, since accuracy is of primary concern in photogrammetry, the scanning mechanism must be sufficiently accurate, and the data used in individual correlations must be limited to the smallest possible area capable of supporting a specified accuracy level. It is shown that multi-level correlation is far superior to polarity correlation, that image data shaping is of great importance, and that the correlation system must be highly adaptive. A successful implementation embodying these principles is briefly described.