Basics of object recognition and environments The contribution of informative features

annotation

The recognition of objects and environments by the signature of the radar signal is a difficult task. In the radar search for ground targets PSNR-5, SBR-3, SNAR-10, target classes are recognized by listening by listening to Doppler signals. It uses the instrument of the auditory organ and the human brain. Reliable recognition machines have not yet been created. This lecture provides an analysis of the signs of the signature.

With any method of recognizing the class of an object, it is necessary to ensure the corresponding signal-to-noise ratio     About _sp .    Otherwise, the informative features of the object will not be revealed against the background of informative signs of interference.

  Signs that can be used to recognize the class of the target can be attributed to both stochastic and deterministic. Deterministic features are stable and do not depend on the object's angle with respect to the radar, external conditions and are inherent only in this class. Stochastic - those that are random in nature, for example, the type of secondary radiation pattern. In the centimeter wavelength range, the diagram is so strongly incised that the width of the lobe and its power relative to the line of sight cannot be described analytically.

  Denote the set of deterministic A C   ( i ) and stochastic s _c ( j )

signs of purpose as J { A _c   ( i ), s _c ( j )}, and the interference, respectively, as H { B _n ( i ), s _n ( j )}.

  In the absence of resolution, the signal from the target and interference enters the input of the radar receiver simultaneously, and in the total volume.   Subject to the linearity of the system signs are preserved. In a non-linear system, the exchange or transformation of the informative part of the signs may occur. Then there will be a partial or complete degradation of areas, diffusion of adjacent areas and the formation of a zone of uncertainties, where the recognition of signs will be impossible.  

  Imagine the flow of signs of the target and interference in the following form.

                      W _c   { A _{c i} , s _c ( j )} = p _c   w _c { A _{c i} , s _c ( j )} - for the goal

                                                                                                                   (ten)

                       W _c   { A _{p i} , s _p ( j )} = p _p   w _p { A _{p i} , s _p ( j )} - for interference

         Where p _C , p _p - power signs of the target and interference, respectively; W , w is the complete distribution of the flow of signs and the probability density of the flow of signs.

  In the General case, the recognition is carried out on the total flux density of the signs of the useful signal from the target and interference w _CP { A _{c i}   , A _{c i}   ,    s (( j ), s (( j )}.

  The total component of the deterministic characteristics of the target A C _? can be calculated as a scalar sum of stable features.    The power of the signs is p _c as the square of the scalar sum.

The total part of the stochastic component of the signs s _{c ?} summed squared                                                

  The total power of all signs of the target and underlying surface

It is now interesting to consider the powers of the signs of the deterministic and stochastic parts of the target relative to the total power of the aggregate attributes of the mixture. The goal is the underlying surface.

Findings:

  Analyzing the form (15), we can draw several preliminary conclusions.

1. The proportion of signs inherent in a fixed target on the background of the underlying          the surface is higher, the greater the power of these signs on the general        field signs of the total signal from the target and interference.

1. The contribution of stable deterministic target features is much larger than stochastic. This means that it is necessary to conduct an active search for these signs in the signal reflected from a fixed target.

2. Despite the fact that stochastic signs make a smaller contribution, their price also depends on their power. If they have sufficient power, they can be successfully used for recognition tasks.

3. As a weighting function of the contribution of features, the power of the reflected signal can be considered. The higher the signal-to-noise ratio, the more tangible the contribution of the powers of the informative features.