Moving vehicle classifier with nonlinear mapper

• US 5,337,257 A
• Filed: 05/17/1993
• Issued: 08/09/1994
• Est. Priority Date: 05/09/1991
• Status: Expired due to Term

• Current Assignee: The Boeing Co.
• Original Assignee: Rockwell International Corporation
• Inventors: White, Stanley A.
• Primary Examiner(s): Harvey, Jack B.
• Assistant Examiner(s): MILLER, CRAIG S
• Application Number:
  US08/061,879
• Time in Patent Office: 449 Days
• Field of Search: 364/554, 364/571.04, 364/553, 364/571.07, 364/577, 364/436, 364/555, 455/63, 340/928, 340/933
• US Class Current: 702/73
• CPC Class Codes: G01S 7/415 Identification of targets based on measurements of movement associated with the target
• CPC Class Codes: G01S 7/539 using analysis of echo signal for target characterisation; Target signature; Target cross-section

Abstract

The moving vehicle classifier with nonlinear mapper provides a method and apparatus for estimating the presence of a target moving vehicle in a sensor environment through estimation and classification of the sensor data. Values are extracted from the sensor data and a target function of extracted values is generated. Integrated function values representative of known vehicles in various physical states are prestored in a reference library. An integral function for the target function data is generated and mapped into a number of the prestored reference integral functions which provides a typically nonlinear mapping function. A MMSE curve fit is applied to this nonlinear mapping function to form a scaling factor for use in generating a scaled function of the data. The scaled data function is then compared to scaled reference functions in the library and the most similar, or least dissimilar, reference is chosen as an estimation of what known vehicle the sensor data represents. The output of the apparatus can be presented as a classification or likelihood of classification, as preferred by the application of interest.

Claims

1. A method of classifying a target moving vehicle, the method comprising the steps of:

(a) storing, in a memory, a plurality of known functions representative of a plurality of known moving vehicles;

(b) sensing, with at least one sensor, a signature of the target vehicle;

(c) extracting at least one signal of interest from the signature of the target vehicle;

(d) generating a function for the extracted signals;

(e) adding a bias to the generated function, thereby producing a biased function;

(f) scaling the biased function so as to limit the integral thereof to a maximum value of unity, thereby producing a target first scaled function;

(g) integrating the target first scaled function, thereby producing a target integrated function;

(h) mapping the target integrated function to at least one of the known functions in the memory, thereby producing a mapping function;

(i) performing a minimum mean squared error linear fit on the mapping function, thereby producing a relative scaling factor;

(j) scaling the target first scaled function by multiplying it by the relative scaling factor, thereby producing a target second scaled function;

(k) determining a measure of similarity between the target second scaled function and each of the known functions; and

(l) indicating the known moving vehicles, the known functions of which produce a measure of similarity which exceeds a predetermined threshold.

2. The method of claim 1, wherein each known function comprises a function selected from the group consisting of time signature, time varying waveform, power spectrum, amplitude histogram, amplitude distribution, and cepstrum.

3. The method of claim 2, wherein the step of storing comprises the step of storing reference data in a random access memory using values for preselected function variables as storage addresses.

4. The method of claim 1, wherein the function generated for the extracted signals values comprises a function selected from the group consisting of time signature, time varying waveform, power spectrum, amplitude histogram, amplitude distribution, and cepstrum.

5. The method of claim 1, wherein the bias added to the generated function is determined by the steps of:

(a) determining a difference between a maximum data value and a minimum data value for a predetermined segment of the generated function; and

(b) multiplying the difference by a predetermined percentage.

6. The method of claim 5, wherein the predetermined percentage is about five percent.

7. The method of claim 1, wherein the step of scaling the biased function comprises the steps of:

(a) integrating the biased function over a preselected domain, thereby producing an integral; and

(b) dividing the biased function by the integral.

8. The method of claim 1, wherein the step of mapping the target integrated function to a known function comprises the step of generating a ratio between the target integrated function and the known function.

9. The method of claim 8, wherein the step of mapping comprises the steps of:

(a) storing target integrated function data in a random access memory using values for preselected function variables as storage addresses; and

(b) addressing the memory by applying functional variable values from the integrated functions.

10. The method of claim 1, wherein the step of performing a minimum mean squared error fit comprises the step of performing a linear regression of mapping function data so as to form a substantially linear scaling factor.

11. Apparatus for classifying a target moving vehicle, the apparatus comprising:

(a) a memory containing a plurality of known functions representative of a plurality of known moving vehicles;

(b) at least one sensor connected for sensing a signature of the target vehicle;

(c) an extractor connected to the sensor for extracting at least one signal of interest from the signature of the target vehicle;

(d) a generator connected to the extractor for generating a function for the extracted signals;

(e) a biaser connected to the generator for adding a bias to the function, thereby producing a biased function;

(f) a first scaler connected to the biaser for scaling the biased function so as to limit the integral thereof to a maximum value of unity, thereby producing a target first scaled function;

(g) an integrator connected to the scaler for integrating the target first scaled function, thereby producing a target integrated function;

(h) a mapper connected to the integrator and to the memory, for mapping the target integrated function to at least one of the known functions in the memory, thereby producing a mapping function;

(i) a fitter connected to the mapper for performing a minimum mean squared error linear fit on the mapping function, thereby producing a relative scaling factor;

(j) a second scaler connected to the fitter and to the first scaler, for applying the scaling factor to the target first scaled function, thereby producing a target second scaled function;

(k) a comparator connected to the second scaler and to the memory, for comparing the target second scaled function with each of the known functions; and

(l) an indicator connected to the comparator for indicating the known moving vehicles, the known functions of which produce a measure of similarity which exceeds a predetermined threshold.

12. The apparatus of claim 11, wherein the memory comprises a random access memory capable of storing data representative of a plurality of known and target second scaled functions.

13. The apparatus of claim 12, wherein the target second scaled functions and known functions comprise functions selected from the group consisting of time signature, time varying waveform, power spectrum, amplitude histogram, amplitude distribution, and cepstrum.

14. The apparatus of claim 11, wherein the memory comprises a random access memory capable of using values for function variables as storage addresses.

15. The apparatus of claim 11, wherein the generator comprises a random access memory having predetermined values stored therein at address locations corresponding to variable values.

16. The apparatus of claim 11, wherein the biaser comprises:

(a) a differencer connected to the extractor for determining a difference in maximum and minimum values of the signals of interest;

(b) a data storage device in which is stored a predetermined percentage value by which the bias is to offset data values; and

(c) a multiplier connected to the differencer and the data storage device for providing the product of the difference and the percentage value, the product being the bias to be added to the function.

17. The apparatus of claim 11, wherein the first scaler comprises:

(a) an integrator connected to the biaser for integrating the biased function over a preselected value range;

(b) an inverter connected to the integrator for generating the multiplicative inverse of the integral; and

(c) a multiplier connected to the integrator and the inverter for multiplying the function by the inverse integral.