Phase gradient auto-focus for SAR images

• US 6,046,695 A
• Filed: 07/09/1997
• Issued: 04/04/2000
• Est. Priority Date: 07/11/1996
• Status: Expired due to Term

• Current Assignee: Leidos, Inc. (Leidos Holdings, Inc.)
• Original Assignee: Science Applications International Corporation
• Inventors: Poehler, Paul L., Mansfield, Arthur W.
• Primary Examiner(s): Lobo, Ian J.
• Application Number:
  US08/890,598
• Time in Patent Office: 1,000 Days
• Field of Search: 342/25, 342/191
• US Class Current: 342/25.00A
• CPC Class Codes: G01S 13/9019 Auto-focussing of the SAR signals

Abstract

A SAR image auto-focus method uses statistics extracted from the image data to develop a phase pure hyperbolic correction function. The method processes an uncorrected azimuth line vector selected from an unfocused array into a corrected azimuth line vector. The method includes forming a bright spot list based on the selected uncorrected azimuth line vector where the bright spot list has at least one bright spot entry. Then, the uncorrected azimuth line vector is unpacked into an uncorrected segment set that has a first uncorrected segment corresponding to a first bright spot entry in the bright spot list. Then, the first uncorrected segment is processed into a first corrected segment based on an average phase slope corresponding to the first bright spot entry, the first corrected segment is packed into the corrected azimuth line vector, and the process repeated for all bright spots. The step of processing the first uncorrected segment into a first corrected segment includes processing the first uncorrected segment through a Fourier transform into an uncorrected spectrum vector, correcting the uncorrected spectrum vector to form a corrected spectrum vector, processing the corrected spectrum vector through an inverse Fourier transform into the first corrected segment. The step of correcting the uncorrected spectrum vector includes computing phase gradient φ_dot (f) at each frequency f in the uncorrected spectrum vector based on:

φ.sub.dot (f)=Im {G(f).sup.•

G'"'"'(f)}/(2π|G(f)|²),

then computing the average phase slope by averaging the phase gradient determined at each frequency in the uncorrected spectrum vector. Then a correction function vector corresponding to the average phase slope is generated, and the uncorrected spectrum vector is multiplied by the correction function vector, element by element, to form the corrected spectrum vector.

Claims

1. A method of processing an uncorrected azimuth line vector selected from an unfocused array into a corrected azimuth line vector comprising steps of:

forming a bright spot list based on the selected uncorrected azimuth line vector, the bright spot list including at least one bright spot entry;

unpacking the uncorrected azimuth line vector into an uncorrected segment set, the uncorrected segment set including a first uncorrected segment corresponding to a first bright spot entry in the bright spot list;

processing the first uncorrected segment into a first corrected segment based on an average phase slope corresponding to the first bright spot entry; and

packing the first corrected segment into the corrected azimuth line vector.

2. The method of claim 1, wherein the step of processing the first uncorrected segment comprises steps of:

preparing a filled uncorrected segment vector from the first uncorrected segment;

processing the filled uncorrected segment vector through a Fourier transform into an uncorrected spectrum vector;

correcting the uncorrected spectrum vector to form a corrected spectrum vector;

processing the corrected spectrum vector through an inverse Fourier transform into a filled corrected segment vector; and

culling the first corrected segment from the filled corrected segment vector.

3. The method of claim 2, wherein the step of correcting comprises steps of:

determining an average phase slope across the uncorrected spectrum vector;

generating a correction function vector corresponding to the average phase slope; and

multiplying the uncorrected spectrum vector by the correction function vector, element by element, to form the corrected spectrum vector.

4. The method of claim 3, wherein the step of determining comprises steps of:

computing a phase gradient at each frequency in the uncorrected spectrum vector; and

computing the average phase slope by averaging the phase gradient determined at each frequency in the uncorrected spectrum vector.

5. The method of claim 4, wherein the step of computing a phase gradient includes computing phase gradient φ

_dot (f) at each frequency f based on;

space="preserve" listing-type="equation">φ

.sub.dot (f)=Im{G(f).sup.•



G'"'"'(f)}/(2π

|G(f)|.sup.2), where

f is a frequency being evaluated,

G(f) is a value of the uncorrected spectrum vector at the frequency f,

G(f).sup.•

is a complex conjugate of G(f),

G'"'"'(f) is a value of a derivative with respect to frequency f of G(f).

6. The method of claim 3, wherein:

the correction function vector includes a correction phase at each frequency in the uncorrected spectrum vector; and

the step of generating the correction function vector includes computing the correction phase at each frequency in the uncorrected spectrum vector so that a slope of the correction phases of the correction function vector is a negative of the average phase slope.

7. The method of claim 2, wherein the step of preparing a filled uncorrected segment vector comprises steps of:

centering the first uncorrected segment in the filled uncorrected segment vector so that an element of the uncorrected segment that corresponds to the first bright spot entry is centered in the filled uncorrected segment vector;

multiplying the first uncorrected segment vector by a windowing vector, element by element, centered in the filled uncorrected segment vector, the windowing vector being weighted to suppress nearby target noise and the effects of asymmetry of the first uncorrected segment; and

filling elements of the filled uncorrected segment vector that are not filled by the first uncorrected segment with zeros.

8. The method of claim 2, wherein the step of culling comprises copying elements from the filled corrected segment vector into the first corrected segment that correspond to elements in the first uncorrected segment.

9. The method of claim 1, wherein the step of processing includes calculating the first corrected segment by determining the average phase slope of the first uncorrected segment.

10. The method of claim 1, wherein:

the bright spot list further includes a second bright spot entry;

the uncorrected segment set further includes a second uncorrected segment corresponding to the second bright spot entry;

the step of processing further processes the second uncorrected segment into a second corrected segment based on an average phase slope corresponding to the second bright spot entry, the average phase slope corresponding to the second bright spot entry being unequal to the average phase slope corresponding to the first bright spot entry.

11. The method of claim 1, wherein:

the bright spot list further includes a plurality of second bright spot entries;

the uncorrected segment set further includes a plurality of second uncorrected segments each corresponding to respective second bright spot entries;

each bright spot entry corresponds to an azimuth angle characterized by a unique Doppler phase;

the step of processing further processes each of the plurality of second uncorrected segments into corresponding second corrected segments based on an average phase slope corresponding to the respective second bright spot entries, the average phase slope corresponding to each of the second bright spot entries and the average phase slope corresponding to the first bright spot entry form a piece wise approximation of a theoretical hyperbolic distribution of Doppler phase slopes across azimuth angles corresponding to angles represented by the uncorrected azimuth line vector.

12. A computing machine comprising:

a processor;

a first module to form a bright spot list based on an uncorrected azimuth line vector selected from an unfocused array, the bright spot list including at least one bright spot entry;

a second module to unpack the uncorrected azimuth line vector into an uncorrected segment set, the uncorrected segment set including a first uncorrected segment corresponding to a first bright spot entry in the bright spot list;

a third module to process the first uncorrected segment into a first corrected segment based on an average phase slope corresponding to the first bright spot entry; and

a fourth module to pack the first corrected segment into a corrected azimuth line vector.

13. The machine of claim 12, wherein the third module comprises:

a first sub-module to prepare a filled uncorrected segment vector from the first uncorrected segment;

a second sub-module to process the filled uncorrected segment vector through a Fourier transform into an uncorrected spectrum vector;

a third sub-module to correct the uncorrected spectrum vector to form a corrected spectrum vector;

a fourth sub-module to process the corrected spectrum vector through an inverse Fourier transform into a filled corrected segment vector; and

a fifth sub-module to cull the first corrected segment from the filled corrected segment vector.

14. The machine of claim 13, wherein the third sub-module comprises:

a first module portion to determine an average phase slope across the uncorrected spectrum vector;

a second module portion to generate a correction function vector corresponding to the average phase slope; and

a third module portion to multiply the uncorrected spectrum vector by the correction function vector, element by element, to form the corrected spectrum vector.

15. The machine of claim 14, wherein the first module portion comprises:

a first module sub-portion to compute a phase gradient at each frequency in the uncorrected spectrum vector; and

a second module sub-portion to compute the average phase slope by averaging the phase gradient determined at each frequency in the uncorrected spectrum vector.

16. The machine of claim 15, wherein the first module sub-portion computes phase gradient φ

_dot (f) at each frequency f based on;

space="preserve" listing-type="equation">φ

.sub.dot (f)=Im{G(f).sup.•



G'"'"'(f)}/(2π

|G(f)|.sup.2), where

f is a frequency being evaluated,

G(f) is a value of the uncorrected spectrum vector at the frequency f,

G(f).sup.•

is a complex conjugate of G(f),

G'"'"'(f) is a value of a derivative with respect to frequency f of G(f).

17. The machine of claim 14, wherein:

the correction function vector includes a correction phase at each frequency in the uncorrected spectrum vector; and

the second module portion computes the correction phase at each frequency in the uncorrected spectrum vector so that a slope of the correction phases of the correction function vector is a negative of the average phase slope.

18. The machine of claim 13, wherein the first sub-module comprises:

a first module portion to center the first uncorrected segment in the filled uncorrected segment vector so that an element of the uncorrected segment that corresponds to the first bright spot entry is centered in the filled uncorrected segment vector;

a second module portion to multiply the first uncorrected segment vector by a windowing vector, element by element, centered in the filled uncorrected segment vector, the windowing vector being weighted to suppress nearby target noise and the effects of asymmetry of the first uncorrected segment; and

a third module portion to fill elements of the filled uncorrected segment vector that are not filled by the first uncorrected segment with zeros.

19. The machine of claim 13, wherein the fifth sub-module comprises a module portion to copy elements from the filled corrected segment vector into the first corrected segment that correspond to elements in the first uncorrected segment.

20. The machine of claim 12, wherein the third module includes a sub-module to calculate the first corrected segment by determining the average phase slope of the first uncorrected segment.

21. The machine of claim 12, wherein:

the bright spot list further includes a second bright spot entry;

the uncorrected segment set further includes a second uncorrected segment corresponding to the second bright spot entry;

the third module includes a sub-module to processes the second uncorrected segment into a second corrected segment based on an average phase slope corresponding to the second bright spot entry, the average phase slope corresponding to the second bright spot entry being unequal to the average phase slope corresponding to the first bright spot entry.

22. The machine of claim 12, wherein:

the bright spot list further includes a plurality of second bright spot entries;

the uncorrected segment set further includes a plurality of second uncorrected segments each corresponding to respective second bright spot entries;

each bright spot entry corresponds to an azimuth angle characterized by a unique Doppler phase;

the third module includes a sub-module to process each of the plurality of second uncorrected segments into corresponding second corrected segments based on an average phase slope corresponding to the respective second bright spot entries, the average phase slope corresponding to each of the second bright spot entries and the average phase slope corresponding to the first bright spot entry form a piece wise approximation of a theoretical hyperbolic distribution of Doppler phase slopes across azimuth angles corresponding to angles represented by the uncorrected azimuth line vector.

23. A computer readable medium having stored thereon a plurality of modules for controlling a processor, the plurality of modules comprising:

a first module to control the processor to form a bright spot list based on an uncorrected azimuth line vector selected from an unfocused array, the bright spot list including at least one bright spot entry;

a second module to control the processor to unpack the uncorrected azimuth line vector into an uncorrected segment set, the uncorrected segment set including a first uncorrected segment corresponding to a first bright spot entry in the bright spot list;

a third module to control the processor to process the first uncorrected segment into a first corrected segment based on an average phase slope corresponding to the first bright spot entry; and

a fourth module to control the processor to pack the first corrected segment into a corrected azimuth line vector.

24. The medium of claim 23, wherein the third module comprises:

a first sub-module to control the processor to prepare a filled uncorrected segment vector from the first uncorrected segment;

a second sub-module to control the processor to process the filled uncorrected segment vector through a Fourier transform into an uncorrected spectrum vector;

a third sub-module to control the processor to correct the uncorrected spectrum vector to form a corrected spectrum vector;

a fourth sub-module to control the processor to process the corrected spectrum vector through an inverse Fourier transform into a filled corrected segment vector; and

a fifth sub-module to control the processor to cull the first corrected segment from the filled corrected segment vector.

25. The medium of claim 24, wherein the third sub-module comprises:

a first module portion to control the processor to determine an average phase slope across the uncorrected spectrum vector;

a second module portion to control the processor to generate a correction function vector corresponding to the average phase slope; and

a third module portion to control the processor to multiply the uncorrected spectrum vector by the correction function vector, element by element, to form the corrected spectrum vector.

26. The medium of claim 25, wherein the first module portion comprises:

a first module sub-portion to control the processor to compute a phase gradient at each frequency in the uncorrected spectrum vector; and

a second module sub-portion to control the processor to compute the average phase slope by averaging the phase gradient determined at each frequency in the uncorrected spectrum vector.

27. The medium of claim 26, wherein the first module sub-portion controls the processor to compute phase gradient φ

_dot (f) at each frequency f based on;

space="preserve" listing-type="equation">φ

.sub.dot (f)=Im{G(f).sup.500 

G'"'"'(f)}/(2π

|G(f)/.sup.2), where

f is a frequency being evaluated,

G(f) is a value of the uncorrected spectrum vector at the frequency f,

G(f).sup.•

is a complex conjugate of G(f),

G'"'"'(f) is a value of a derivative with respect to frequency f of G(f).

28. The medium of claim 25, wherein:

the correction function vector includes a correction phase at each frequency in the uncorrected spectrum vector; and

the second module portion control the processor to compute the correction phase at each frequency in the uncorrected spectrum vector so that a slope of the correction phases of the correction function vector is a negative of the average phase slope.

29. The medium of claim 24, wherein the first sub-module comprises:

a first module portion to control the processor to center the first uncorrected segment in the filled uncorrected segment vector so that an element of the uncorrected segment that corresponds to the first bright spot entry is centered in the filled uncorrected segment vector;

a second module portion to control the processor to multiply the first uncorrected segment vector by a windowing vector, element by element, centered in the filled uncorrected segment vector, the windowing vector being weighted to suppress nearby target noise and the effects of asymmetry of the first uncorrected segment; and

a third module portion to control the processor to fill elements of the filled uncorrected segment vector that are not filled by the first uncorrected segment with zeros.

30. The medium of claim 24, wherein the fifth sub-module comprises a module portion to control the processor to copy elements from the filled corrected segment vector into the first corrected segment that correspond to elements in the first uncorrected segment.

31. The medium of claim 23, wherein the third module includes a sub-module to control the processor to calculate the first corrected segment by determining the average phase slope of the first uncorrected segment.

32. The medium of claim 23, wherein:

the bright spot list further includes a second bright spot entry;

the uncorrected segment set further includes a second uncorrected segment corresponding to the second bright spot entry;

the third module includes a sub-module to control the processor to processes the second uncorrected segment into a second corrected segment based on an average phase slope corresponding to the second bright spot entry, the average phase slope corresponding to the second bright spot entry being unequal to the average phase slope corresponding to the first bright spot entry.

33. The medium of claim 23, wherein:

the bright spot list further includes a plurality of second bright spot entries;

the uncorrected segment set further includes a plurality of second uncorrected segments each corresponding to respective second bright spot entries;

each bright spot entry corresponds to an azimuth angle characterized by a unique Doppler phase;

the third module includes a sub-module to control the processor to process each of the plurality of second uncorrected segments into corresponding second corrected segments based on an average phase slope corresponding to the respective second bright spot entries, the average phase slope corresponding to each of the second bright spot entries and the average phase slope corresponding to the first bright spot entry form a piece wise approximation of a theoretical hyperbolic distribution of Doppler phase slopes across azimuth angles corresponding to angles represented by the uncorrected azimuth line vector.