Towards a Theory of Perception for Radar Targets

Abstract

Visual perception depends on observation of three-dimensional objects, based upon a set of given exposures of the objects relative to the observer. Hence there is in the data base a built-in observer bias which has to be eliminated if the target is to be recognized as a physical object, independently of the observer.

Radar observation is built on the same principle, except that due to polarization dependence of the target a “polarizationbias” is introduced, which corrupts the target data base. Elimination of the polarization bias can be achieved if the target scattering matrix is known for the monostatic radar case. This approach was called phenomenological in previous work by this author.

The report reviews the methods of data presentation of objects based upon fields and upon power. For the polarized scattered return this amounts to a coherent wave-field or a completely polarized (cp) stokes vector power-presentation. For targets this amounts to a presentation based upon the scattering matrix or the stokes-matrix (Mueller-matrix). A novel vector formulation is introduced which relates to the cognitive requirements for a string of target features.

The power-presentations are useful to consider object mixtures which lead to a new class of partially polarized (pp) waves and of distributed radar targets. From the object mixtures we wish to formulate the concept of ‘the average object’.

This can be accomplished through the so-called ‘N-target decomposition theorem’, introduced by the author in 1970.

Concept formulation of general object structures is shown to be diagrammically related to hierarchical object tree structures, widely used in AI work. The work object (n=2), the radar-target (n=3) and for the general object of index n. The generally ‘useful’ attributes are linearly related to the index: (2n-1)2. Hence there is great economy in preserving those meaningful parameters by which the idealized object is recognized.

These novel perceptually related data presentations should be useful as a guide to the radar-analyst in his search for meaningful target designators.