Spatial-temporal-structure processor for multi-sensor, multi scan data fusion

• US 5,293,455 A
• Filed: 02/13/1991
• Issued: 03/08/1994
• Est. Priority Date: 02/13/1991
• Status: Expired due to Term

• Current Assignee: Hughes Electronics Corporation (Rtx Corporation)
• Original Assignee: Hughes Aircraft Company (Rtx Corporation)
• Inventors: Castelaz, Patrick F.
• Primary Examiner(s): MacDonald, Allen R.
• Application Number:
  US07/660,362
• Time in Patent Office: 1,119 Days
• Field of Search: 395/1, 395/23, 395/24
• US Class Current: 706/25
• CPC Class Codes: G06F 15/8038 Associative processors

Abstract

A system is provided for associating a set of data values where each data value belongs to one of a plurality of classes and also to one of a plurality of types. The system includes an array of memory cells arranged in rows and columns, each memory cell storing one data value. An energy computation cell for each column is coupled for computing an energy value proportional to the likelihood that the data values in each column comprise one of the desired associations. A system control module to the energy computation cells for transferring the data values of a given class and type from one memory cell to another memory cell of the same class and type, within a given row of memory cells when the replacement lowers to the energy value for the column. The energy values computed by all the energy computation cells are summed to determine a total energy value. Each time data values are transferred, the total energy value of all of the energy cells is computed. The lowest total energy value after successive replacements represents a state in which each column contains the desired association of data values.

Claims

1. A system for forming desired associations among a set of data values, each of said data values belonging to one of a plurality of classes and one of a plurality of types, said system comprising:

an array of memory cells arranged in rows and columns, each memory cell storing one of said data values, said data values arranged in said memory cells, so that each column contains one data value of each possible combination of said classes and types;

data bus means for transmitting said data values into and out of said memory cells;

an energy computation cell for each column coupled to said data bus means for computing an energy value for each column, wherein said energy value is proportional to the likelihood that the data values in said column together comprise one of said desired associations;

system control means connected to said data bus means and said energy computation cells for transferring said data values of a given class and type from one memory cell to another memory cell of the same class and type within a given row of memory cells, when said energy value for said column is lowered by said transfer;

said system control means including a summing means for adding up all the energy values computed by said energy computation cells to determine a total energy value proportional to the likelihood that all columns contain desired associations of data values, and

wherein each time said system control means transfers data values, said summing means determines the total energy value of all of said energy computation cells, whereby, the total energy value having a lowest total energy after successive transfers represents a state in which each column contains said desired associations of data values.

2. The system of claim 1 wherein said energy computation cells comprise a correlation means for computing a first correlation value between data values having the same type but different classes within a column, and also for computing a second correlation value between data values having the same class but different types, within a column;

and

wherein said energy value is equal to the total of said first and second correlation values.

3. The system of claim 2 wherein said data values are azimuth and elevation values from detection of objects by sensors, at a plurality of time sequences, each of said classes of data comprises data values from one of said time sequences, and each of said types of data values originate from one of said sensors;

wherein said correlation means computes a first correlation value by measuring the difference in distance between two lines defined by the sensor position and the azimuth and elevation of the data values at the point of closest approach of the two lines for the same time sequence from different sensors, and said correlation means computes a second correlation value which is proportional to the difference in velocity of said object in successive time sequences of data from a single sensor.

4. The system of claim 1 further comprising a host computer coupled to said data bus means for storing two-dimensional data values and for transmitting said data to said data bus means.

5. The system of claim 4 further comprising host interface and control means coupled to said host computer and data bus means for transmitting said data values to said memory cells through said data bus means in a random distribution.

6. The system of claim 1 wherein said system control means further comprises a means for transmitting a mobility value to each memory cell, and mobility check means for comparing the mobility value of said memory cells with a predetermined threshold and for transferring data values from only memory cells having mobility values which exceed a predetermined threshold.

7. The system of claim 6 wherein said mobility checking means multiplies said predetermined threshold by a randomly varying number each time it makes said comparison.

8. The system of claim 6 wherein said system control means further comprises a mobility changing means for lowering the mobility value of the memory cell receiving the transferred data value and for raising the mobility value of the memory cell from which the transferred data value came.

9. The system of claim 8 wherein said mobility changing means further lowers the mobility value of selected memory cells in the same column as said memory cell receiving the transferred data, said selected memory cells comprising those cells having the same class, or the same type, as the cell receiving the transferred data.

10. The system of claim 8 wherein said mobility changing means comprises a structure determining means for lowering the mobility value an amount that is proportional to a computation based on the surrounding data values.

11. The system of claim 10 wherein said structure determining calculates a centroid of said data values in a specific class.

12. The system of claim 1 wherein said system control means further comprises energy minimum checking means for determining the rate of change of said summed energy, and mobility reset means for resetting the mobility of said memory cells to higher values if said rate of change of said energy determined by energy minimum checking means is below an energy change threshold.

13. The system of claim 12 wherein said energy minimum checking means further comprises a means for determining the second derivative of said summed energy, for determining when said second derivative is less than a second predetermined energy change threshold, wherein convergence of said processor is detected by minimal changes in said summed energy.

14. The system of claim 1 wherein said data values each comprise two dimensional data values.

15. A method for forming desired associations among a set of data values, said data values each belonging to one of a plurality of classes and one of a plurality of types, said method comprising:

providing an array of memory cells arranged in rows and columns, each memory sell storing one of said data values, wherein said data values are arranged in said memory cells so that each column contains one data value of each possible combination of said classes and types;

transmitting along a data bus, said data values into said memory cells;

computing an energy value for each column of memory cells, wherein said energy value is proportional to the likelihood that the data values in said column together comprise one of said desired associations;

transferring said data values of a given class and type from one memory cell to another memory sell of the same class and type, within a given row of memory cells;

determining if the energy for the column of the cell receiving the transferred data values is increased due to said transfer, and putting said transferred data values back in their original positions when said energy value is increased due to aid transfer;

summing all the energy values for each column to determine a total energy value proportional to the likelihood that all columns contain desired associations of data values; and

repeating said step of replacing said data values, until a lowest total energy value is reached based on a predetermined criteria.

16. The method of claim 15 wherein said step of computing an energy value for each column further comprises the step of computing a first correlation value between data values having the same type but different classes, and computing a second correlation value between data values having the same class but different types within a column, wherein said energy value is equal to the total of said first and second correlation values.

17. The method of claim 16 wherein said data values are azimuth and elevation values from detection of objects by sensors, at a plurality of time sequences, each of said classes of data comprise data values from one of said time sequences, and each of said types of data values originate from one of said sensors, wherein the step of computing an energy value for each column further comprises the step of computing a first correlation value by measuring the difference in distance between two lines defined by the sensor position and the azimuth and elevation data value at the point of closest approach of the two lines for the same time sequence from different sensors, and said step of computing a second correlation value computes a second correlation value which is proportional to the difference in velocity of said object in successive time sequences of data from a single sensor.

18. The method of claim 15 further comprising the step of transmitting a mobility value to each memory cell and comparing the mobility value of said memory cells with a predetermined threshold and replacing only memory cells having mobility values which exceed a predetermined threshold.

19. The method of claim 18 wherein said step of comparing the mobility further comprises a step of multiplying the predetermined threshold by a randomly varying number each time it makes said comparison.

20. A method of claim 18 further comprising the step of lowering the mobility value of the memory cell receiving the transferred data, and raising the mobility value of the memory cell from which the transferred data value came.

21. The method of claim 20 further comprising the step of lowering the mobility value of selected memory cells in the same column as said memory cell receiving the transferred data, said selected memory cells comprising those cells having the same class, or the same type, as the cell receiving the transferred data.

22. The method of claim 20 further comprising the step of lowering the mobility value an amount that is proportional to a computation based on surrounding data values.

23. The method of claim 22 further comprising the step of calculating a centroid of said data values in a given class to determine the amount to lower mobility.

24. The method of claim 18 further comprising the step of checking for a minimum energy state by determining the rate of change of said summed energy, and the step of resetting all mobility values to higher values if the rate of change of said energy is below a predetermined energy change threshold.

25. The method of claim 15 further comprising the step of determining the second derivative of said summed energy and determining when said second derivative is less than a second predetermined energy change threshold, wherein convergence is detected by minimal changes in said summed energy.

26. The method of claim 15 wherein said data values each comprise two dimensional data values.