Explicit probabilistic target object selection and engagement
First Claim
1. A method for selecting an object of interest in a cloud of objects of lesser interest, said method comprising the steps of:
- sensing objects of said cloud with a radio-frequency (RF) sensor at a location, to thereby generate sensed RF signals for at least some of said objects of said cloud;
discriminating said sensed RF signals by computing the probability that each RF tracked object is one of a predetermined number (X) of possible object types, to thereby form an RF confusion matrix;
sensing objects of said cloud with an optical sensor at a location different from the location of said RF sensor, to thereby generate sensed optical signals for at least some of said objects of said cloud;
discriminating said sensed optical signals by computing the probability of each optical tracked object is one of a predetermined number (X) of possible object types to thereby form an optical confusion matrix;
calculating the correlations between all pairs of objects consisting of one RF object and one IR object to thereby form an RF/IR correlation matrix;
calculating joint probabilities for all pairs of RE and optical signals and all objects to produce a matrix set of joint probabilities;
normalizing said joint probabilities over all object types to produce a matrix set of normalized joint probabilities;
calculating marginal probabilities of the joint RF/IR discrimination results to produce a vector set of marginal optical probabilities;
normalizing said vector set of marginal optical probabilities over all object types to thereby produce a vector set of normalized marginal optical probabilities; and
selecting a guide-to object as the IR object of said vector set of normalized marginal optical probabilities with the highest probability of being the object type of interest.
1 Assignment
0 Petitions
Accused Products
Abstract
An object of interest in a cloud of objects is identified by RF and IR sensing. The RF and IR signals are separately discriminated to determine the probability that the RF tracked object is one of a predetermined number of possible object types, and the IR tracked object is one of the possible object types. Joint probabilities are calculated for all pairs of RF and IR signals and all objects, and the joint probabilities are normalized. Marginal probabilities of the joint RF/IR discrimination results are calculated to produce a vector set of marginal optical probabilities. The vector set is normalized over all object types to thereby produce a vector set of normalized marginal optical probabilities. The object of interest is selected to be the IR object of said vector set of normalized joint optical probabilities with the highest probability of being the object type of interest.
46 Citations
17 Claims
-
1. A method for selecting an object of interest in a cloud of objects of lesser interest, said method comprising the steps of:
-
sensing objects of said cloud with a radio-frequency (RF) sensor at a location, to thereby generate sensed RF signals for at least some of said objects of said cloud; discriminating said sensed RF signals by computing the probability that each RF tracked object is one of a predetermined number (X) of possible object types, to thereby form an RF confusion matrix; sensing objects of said cloud with an optical sensor at a location different from the location of said RF sensor, to thereby generate sensed optical signals for at least some of said objects of said cloud; discriminating said sensed optical signals by computing the probability of each optical tracked object is one of a predetermined number (X) of possible object types to thereby form an optical confusion matrix; calculating the correlations between all pairs of objects consisting of one RF object and one IR object to thereby form an RF/IR correlation matrix; calculating joint probabilities for all pairs of RE and optical signals and all objects to produce a matrix set of joint probabilities; normalizing said joint probabilities over all object types to produce a matrix set of normalized joint probabilities; calculating marginal probabilities of the joint RF/IR discrimination results to produce a vector set of marginal optical probabilities; normalizing said vector set of marginal optical probabilities over all object types to thereby produce a vector set of normalized marginal optical probabilities; and selecting a guide-to object as the IR object of said vector set of normalized marginal optical probabilities with the highest probability of being the object type of interest. - View Dependent Claims (2, 3, 4, 5, 6, 7, 8)
-
-
9. A system for identifying an object of interest in a cloud of remote objects of different types, where the number of object types is no greater than X, said system comprising:
-
an electromagnetic sensor observing said cloud, and configured to generate RF signals representing at least some of the objects of said cloud; an electromagnetic discriminator coupled to said electromagnetic sensor and configured to compute the probability of each RF tracked object being one of the X possible object types; an optical sensor observing at least portions of said cloud, and configured to generate optical signals representing at least some objects of said cloud; an optical discriminator configured to compute the probability of each IR tracked object being one of the X possible object types; an electromagnetic/optical correlator coupled to said electromagnetic discriminator and to said optical sensor, and configured to determine the probability that the ith electromagnetic object is correlated or matched with the jth optical object; a processor coupled to said electromagnetic discriminator, to said optical discriminator, and to said electromagnetic/optical correlator, for (a) generating an electromagnetic/optical correlation matrix, (b) calculating joint probabilities for all pairs of electromagnetic and optical signals and all types of objects, (c) normalizing said joint probabilities over all object types to produce a matrix set of normalized joint probabilities, (d) calculating marginal probabilities for each optical object, (e) normalizing the marginal probabilities for each optical object and over all object types, and (f) selecting as the object of interest the optical object with the highest normalized probability. - View Dependent Claims (10, 11, 12, 13, 14, 15)
-
-
16. A method comprising the steps of:
-
receiving RF probability data indicative of a probability that an object being tracked by an RF sensor is one of a predetermined number (X) of possible object types; receiving IR probability data indicative of a probability that an object being tracked by an optical sensor is one of said predetermined number (X) of possible object types; receiving RF and IR correlation data indicative of correlations between pairs of objects consisting of one said RF object being tracked and one said IR object being tracked, corresponding to a probability that the ith RF object being tracked is matched with the jth IR object being j tracked; determining via a computer processor, based on said received RF probability data, IR probability data, and RF and IR correlation data, data indicative of joint probabilities for all pairs of said RF objects being tracked and said IR objects being tracked; normalizing said joint probability data over all object types to produce data sets of normalized joint probabilities; determining marginal probabilities based on the normalized joint probabilities data sets to produce a data set of marginal optical probabilities representative of the probability that a given IR object is of a given object type, independent of which RF object the given IR object corresponds to; normalizing said data set of marginal optical probabilities over all object types to thereby produce a data set of normalized marginal optical probabilities; and determining the IR object of said data set of normalized marginal optical probabilities having the highest probability for the object type of interest to be the guide-to object. - View Dependent Claims (17)
-
Specification