This chapter considers techniques and approaches that are not oriented towards investigations in basic or scientific study. They also do not directly compute the classification decisions or rankings that constitute discrimination, per se (see Chapter 8); instead, they feed essential quantities into the classification processing and provide a basis for it. They are designed to deal specifically with data, to extract essential or helpful things, and, ideally, to perform efficiently. As such, they constitute vital connective and supportive links. The first two techniques are fast enough to operate in real time during surveying (at least during cued interrogation). Joint diagonalization provides a basis for quickly estimating the number of subsurface sources, along with the time or frequency patterns of distinct response components. It exploits the potential of multi-static response data, that is, data in which an Rx unit may be at a different location from a Tx unit, as in an array. The HAP (“H-A-C”) method also exploits the data forms provided by contemporary and emerging instruments, quickly estimating the location and orientation of a single object. The remaining sections of the chapter treat two extremely useful computational approaches. The ONVMS, when combined with DE, absorbs elements of the material treated in the preceding chapters to produce an effective rendering of the cycle depicted in Fig. 4.5. The clustering methods in Section 7.3 constitute a means for analyzing the structure of parameter information. They identify groupings, guide labeling, and provide a means for generalization to broader decision making. These computational systems are thus closely linked to the calculations that ultimately determine classification ranking (Chapter 8).
Online access to SPIE eBooks is limited to subscribing institutions.