Method for selecting representative endmember components from spectral data
View Patent ↗Described herein is a process for objectively and automatically determining spectral endmembers and transforming Spectral Mixture Analysis (SMA) from a widely used research technique into a user-friendly tool that can support the needs of all types of remote sensing. The process extracts endmembers from a spectral dataset using a knowledge-based approach. The process identifies a series of starting spectra that are consistent with a scene and its environment. The process then finds endmembers iteratively, selecting each new endmember based on a combination of physically and statistically-based tests. The tests combine spectral and spatial criteria and decision trees to ensure that the resulting endmembers are physically representative of the scene.
1. A process for determining at least one candidate spectral endmember within an image composed of multiple pixels of spectral data comprising:
unmixing the spectral data within each of the multiple pixels into matched spectral data and unmatched spectral data, the matched spectral data being determined based on spectral data that matches at least a first criteria determined by a first criteria threshold;
defining a metric value range, wherein the unmatched spectral data having first metric values within the metric value range are defined as candidate spectra;
ordering the candidate spectra from highest first metric value to lowest first metric value;
comparing each of the candidate spectra, beginning with the candidate spectra having the highest first metric value, to the unmatched spectral data, to determine the frequency with which each of the candidate spectra occurs within the unmatched spectral data; and
calculating a second metric value for each of the candidate spectra, wherein the second metric value combines the frequency of occurrence of each of the candidate spectra within the unmatched spectral data with a first metric value for each of the candidate spectra, wherein the candidate spectra having the largest second metric value is the at least one candidate endmember.
2. The process according to claim 1 , further comprising applying a second criteria threshold spectral mask to the unmatched spectral data within each of the multiple pixels to determine a subset of pixels whose unmatched spectral data is below the second criteria threshold;
defining a metric value range, wherein the unmatched spectral data from the subset of pixels having first metric values within the metric value range are defined as candidate spectra;
ordering the candidate spectra from highest first metric value to lowest first metric value;
comparing each of the candidate spectra, beginning with the candidate spectra having the highest first metric value, to the unmatched spectral data from the subset of pixels, to determine the frequency with which each of the candidate spectra occurs within the unmatched spectral data from the subset of pixels; and
calculating a second metric value for each of the candidate spectra, wherein the second metric value combines the frequency of occurrence of each of the candidate spectra within the unmatched spectral data from the subset of pixels with a first metric value for each of the candidate spectra, wherein the candidate spectra having the largest second metric value is the at least one candidate endinember.
3. The process according to claim 2 , wherein the second criteria is shade.
4. The process according to claim 1 , wherein the first criteria is first endmember spectra.
5. A process for determining at least one candidate spectral endmember within an image composed of multiple pixels of spectral data, comprising:
unmixing the spectral data within each of the multiple pixels into matched spectral data and unmatched spectral data, the matched spectral data being determined based on spectral data that matches at least a first criteria;
defining error values for the unmatched spectral data, wherein the error values are representative of portions of the unmatched spectral data within each of the multiple pixels;
comparing the error values for the unmatched spectral data to a predetermined error value range, wherein the unmatched spectral data having error values within the predetermined error value range are defined as candidate spectra;
comparing each of the candidate spectra, beginning with the candidate spectra having the highest error value, to each of the unmatched spectral data, to determine the frequency with which each of the candidate spectra occurs within the unmatched spectral data; and
calculating a metric for each of the candidate spectra, wherein the metric combines the frequency of cocurrence of each of the candidate spectra within the unmatched spectral data with an error value for each of the candidate spectra, wherein the candidate spectra having the largest metric is the at least one endmember.
6. The process according to claim 5 , wherein the first criteria is a first endmember spectra.
7. The process according to claim 5 , further comprising calculating a root mean square (RMS) error value for each of the multiple pixels by combining the error values for the unmatched spectral data.
8. The process according to claim 7 , wherein the step of comparing the error value for each of the unmatched spectral data to a predetermined error value range, wherein spectra having error values within the predetermined error value range are defined as candidate spectra includes:
determining an acceptable range of deviation from the mean RMS error;
comparing each of the RMS error values for each of the multiple pixels to the acceptable range of deviation from the mean RMS error and keeping a subset of multiple pixels that are within the acceptable range.
9. The process according to claim 6 , wherein the first endmember is a shade endmember and further wherein the shade endmember is determined by a pre-selected baseline percentage of reflectance.
10. The process according to claim 6 , wherein the first endmember is selected from the group consisting of a vegetation endrnember, a soil endmember and a nonphotosynthetic vegetation endmember, with the vegetation endmember being determined by comparing spectral data for the multiple pixels to a known set of vegetation spectra, the soil endmember being determined by comparing spectral data for multiple pixels to a known set of soil spectra, and the nonphoto synthetic vegetation spectra being determined by comparing spectral data for the multiple pixels to a known set of nonphotosynthetic vegetation spectra.
11. The process according to claim 10 , wherein the step of comparing the spectral data for the multiple pixels to a known spectral data utilizes at least one spectral mapping algorithm.
12. The process according to claim 10 , wherein the pre-selected baseline percentage of reflectance is at most 1 percent.
13. The process according to claim 5 , wherein the number of multiple pixein is at most 500.
14. The process according to claim 5 , wherein the spectral data is selected from the group consisting of hyperspectral data, multispectral data, and ultraspectral data.
15. The process according to claim 5 , wherein the unmixing is linear.
16. The process according to claim 5 , wherein the unmixing is non-linear.