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 computer readable medium encoded with a computer program for aiding a user in the selection of a candidate spectral endmember for a scene composed of multiple pixels comprising:
receiving scene reflectance data comprised of multiple unknown spectra;
receiving reference spectral data comprised of multiple known spectra;
comparing each unknown spectra to the known spectra, each being represented as a vector in n-dimension space, by measuring the angle of separation between the each unknown spectra vector and each known spectra vector;
selecting unknown spectra with smallest measured angles of vector separation from known spectra as candidate spectral endmembers;
unmixing each of the multiple pixels to determine percentage contribution of each of the candidate spectral endmembers to each of the multiple pixels;
applying a shade mask to the unmixed pixels, resulting in X number of pixels which have less than a threshold amount of shade determined by the shade mask;
generating a root mean square (RMS) error image utilizing the X number of pixels;
setting an RMS minimum and maximum range, wherein pixels falling with the range represent candidate spectral endmembers; and
comparing the X number of pixels to the range to determine candidate spectral endmembers; and
selecting a candidate spectral endmember.
2. The program according to claim 1 , further comprising varying at least one of the reference spectral data, the shade mask and the RMS minimum and maximum range.
3. The program according to claim 1 , further comprising varying at least two of the reference spectral data, the shade mask and the RMS minimum and maximum range.
4. The program according to claim 1 , wherein the shade mask is set at 80 percent and pixels containing 80 percent or more of shade are no longer considered to be candidate spectral endmembers.
5. The program according to claim 2 , wherein varying at least one of the reference spectral data, the shade mask and the RMS minimum and maximum range is performed by the user.
6. The program according to claim 2 , wherein varying at least one of the reference spectral data, the shade mask and the RMS minimum and maximum range is performed by the program utilizing decision tree algorithms.
7. The program according to claim 1 , wherein the scene reflectance data is selected from the group consisting of hyperspectral data, multispectral data, and ultraspectral data.
8. The program according to claim 1 , wherein the unmixing is linear.
9. The process according to claim 1 , wherein the unmixing is non-linear.