System, apparatus and method for hierarchical identification, multi-tier target library processing, and two-stage identification
AssigneeGeospatial Technology Associates LLC
InventorWilliam F. Basener
Life of this patent
Estimating the term…
Abstract
Embodiments of a system, method and apparatus incorporate two-stage identification of targets in hyperspectral image analysis. In various embodiments, unmixing is employed that integrates F-test and model averaging approaches. Further, a multi-tier target library process provides an improvement in the spectra that can be used to detect target materials and spectra that can be used for unmixing in identification. Additionally, the hierarchical identification of the present disclosure combines probabilities from model averaging to generate target identifications simultaneously at multiple levels of specificity.
Claims20
A method for detection and identification of target material in spectral imagery, comprising:
- storing a first set of spectra for a plurality of target materials in a target detection library;
- receiving a spectral image captured by and transmitted from an image capture sensor;
- determining, by an apparatus using the target detection library, a target detection score representing the likelihood of presence of at least one of the plurality of target materials in each of a plurality of pixels in the spectral image;
- storing a second set of spectra for the plurality of target materials in a target identification library;
- determining, by the apparatus using the target identification library, a material identification score representing a likely identification of one or more of the plurality of target materials for each of the plurality of pixels, wherein the material identification score is determined as a set of probabilities with a probability for each class label at each level in a multi-level hierarchical target material class designation;
- and
- based on the material identification score, identifying and displaying a target material on a display.
An apparatus for detecting and identifying target material in spectral imagery, comprising:
- an image capture sensor;
- at least one processor;
- at least one memory device storing a plurality of instructions which, when executed by the at least one processor, cause the at least one processor to:
- determine, using a target identification library, a material identification score representing a likely identification of one or more of a plurality of target materials for each of a plurality of pixels in a spectral image obtained from the image capture sensor, wherein the material identification score is determined as a set of probabilities with a probability for each class label at each level in a multi-level hierarchical target material class designation;
- and
- based on the material identification score, identifying and displaying a target material on a display.
A system for detection and identification of target material in spectral imagery, comprising:
- an image capture sensor;
- an apparatus comprising at least one processor and at least one memory device storing a plurality of instructions which, when executed by the at least one processor, cause the at least one processor to:
- establish a plurality of multi-level hierarchical target material class designations for a plurality of target materials;
- receive at least one spectral image from the image capture sensor, wherein the at least one spectral image comprises a plurality of pixels;
- determine, using a target identification library, a material identification score representing a likely identification of one or more of the plurality of target materials for the at least one pixel, wherein the material identification score is determined as a set of probabilities with a probability for each class label at each level in a multi-level hierarchical target material class designation;
- and
- an interface in communication with the at least one processor representing the material identification score and an identification of one or more of the plurality of target materials for the at least one pixel.
Dependencies are inferred from the claim text.