초록 ▼

Apparatus and methods for target detection in hyperspectral images are disclosed. In one embodiment, a method of detecting a target in a hyperspectral image includes spectrally unmixing the hyperspectral image into segments, each segment having at least one of similar spectral composition, similar t

Apparatus and methods for target detection in hyperspectral images are disclosed. In one embodiment, a method of detecting a target in a hyperspectral image includes spectrally unmixing the hyperspectral image into segments, each segment having at least one of similar spectral composition, similar textural composition, and similar variation, and spectrally unmixing at least one of the segments. The method further includes creating a clutter rejection filter for at least one segment, filtering at least one segment, and calculating target abundances in at least one segment. In alternate embodiments, channel reduction can be performed on the hyperspectral image and also on at least one segment. In further embodiments, data associated with the location of possible targets in the segments may be compiled. In yet another embodiment, this data may be compressed by cross referencing data from all segments and eliminating redundancies.

대표청구항 ▼

What is claimed is: 1. A method of detecting a target or anomaly in a hyperspectral image, comprising: spectrally unmixing the hyperspectral image; segmenting the hyperspectral image into a plurality of segments, each segment having at least one of a similar spectral composition, a similar textural

What is claimed is: 1. A method of detecting a target or anomaly in a hyperspectral image, comprising: spectrally unmixing the hyperspectral image; segmenting the hyperspectral image into a plurality of segments, each segment having at least one of a similar spectral composition, a similar textural composition, and a similar variation; spectrally unmixing at least one of the segments; creating a clutter rejection filter for at least one of the unmixed segments; filtering at least one of the unmixed segments, wherein the clutter rejection filter is oriented orthogonal to background end members within the at least one unmixed segment while being not oriented orthogonal to the target or anomaly; and calculating a target abundance in at least one of the unmixed segment. 2. The method of claim 1, further comprising performing channel reduction on the hyperspectral image. 3. The method of claim 1, further comprising performing channel reduction on at least one segment. 4. The method of claim 1, wherein creating a clutter rejection filter for at least one segment comprises creating clutter rejection filters for a plurality of segments, and further comprising compiling data associated with the location of possible targets in the plurality of segments. 5. The method of claim 4, wherein the data compiled comprises at least one of a channel index, a target index, a target abundance, an end member index, and an endmember abundance. 6. The method of claim 4, further spatially compressing the compiled data by cross referencing data from all segments and eliminating redundancies. 7. The method of claim 6, wherein cross referencing data from all segments and eliminating redundancies comprises referencing near identical endmembers between two segments to a single endmember spectrum. 8. The method of claim 1, wherein segmenting the hyperspectral image into segments variation includes creating a semivariogram map of the image. 9. A computer readable storage medium having computer executable instructions for processing a hyperspectral image, comprising: a first computer program portion adapted to receive the hyperspectral image; a second computer program portion adapted to spectrally unmix the hyperspectral image; a third computer program portion adapted to segment the hyperspectral image into a plurality of segments, each segment having at least one of similar spectral composition, similar textural composition, and similar variation; a fourth computer program portion adapted to spectrally unmix at least one of the segments; a fifth computer program portion adapted to create a clutter rejection filter for at least one segment and to use the clutter rejection filter to filter that segment, wherein the clutter rejection filter is oriented orthogonal to background endmembers within the segment while being not oriented orthogonal to a target within the segment; and a sixth computer program portion adapted to calculate an abundance of targets in at least one segment. 10. The computer readable storage medium of claim 9, wherein the first computer program portion is adapted to perform channel reduction on the hyperspectral image. 11. The computer readable storage medium of claim 9, wherein the third computer program portion is adapted to perform channel reduction on at least one segment. 12. The computer readable storage medium of claim 9, wherein the fifth computer program portion is adapted to create a clutter rejection filter for a plurality of segments, and further wherein the fifth computer program portion is adapted to compile data associated with the location of possible targets in the plurality of segments. 13. The computer readable storage medium of claim 12, wherein the fifth computer program portion is adapted to compile data comprising at least one of a channel index, a target index, a target abundance, an endmember index, and an endmember abundance. 14. The computer readable storage medium of claim 12, wherein the fifth computer program portion is adapted to compress the data by cross referencing data from all segments and eliminating redundancies. 15. The computer readable storage medium of claim 9, wherein the third computer program portion is adapted to create a semivariogram map using the maximum contrast channel between target and background. 16. A sensor system for processing a scene, comprising: a sensor adapted to receive an hyperspectral image of the scene; a data reduction system operatively coupled to the sensor and including a processor and a memory device operatively coupled to the processor, wherein the data reduction system includes: a first portion adapted to receive the hyperspectral image of the scene from the sensor; a second portion adapted to spectrally unmix the hyperspectral image; a third portion adapted to segment the hyperspectral image into a plurality of segments, each segment having at least one of similar spectral composition, similar textural composition, and similar variation; a fourth portion adapted to spectrally unmix at least one of the segments; a fifth portion adapted to create a clutter rejection filter for the at least one segment and filter the at least one segment, wherein the clutter rejection filter is oriented orthogonal to background endmembers within the at least one segment while being not oriented orthogonal to a target within the at least one segment; and a sixth portion adapted to calculate an abundance of targets in the at least one segment. 17. The sensor system of claim 16, wherein the first portion is adapted to perform channel reduction on the hyperspectral image. 18. The sensor system of claim 16, wherein the third portion is adapted to perform channel reduction on at least one segment. 19. The sensor system of claim 16, wherein the fifth portion is adapted to create a clutter rejection filter for a plurality of segments, and further wherein the fifth portion is adapted to compile data associated with the location of possible targets in the plurality of segments. 20. The sensor system of claim 19, wherein the fifth portion is adapted to compile data comprising at least one of a channel index, a target index, a target abundance, an endmember index, and an endmember abundance, and further wherein the fifth portion is adapted to compress the data by cross referencing data from all segments and eliminating both spatial and spectral redundancies.