Material classification fused with spatio-spectral edge detection in spectral imagery

A method comprising:

• acquiring data of an image;
• processing the data including identifying a candidate location of an edge between two materials in the image, the data having neighboring pixels defining a portion of the candidate location;
• applying a classifier to the neighboring pixels to determine, based on material changes, if the neighboring pixels are correlated to two different materials with respect to the candidate location, providing a modifier for a spatio-spectral mask;
• applying the spatio-spectral mask using the modifier to accept or reject the candidate locations as an edge;
• and
• generating an image to a device, the candidate location in the image correlated to the application of the spatio-spectral mask using the modifier, wherein the method includes:
• identifying a plurality of candidate locations of a plurality of edges of a plurality of materials by defining a plurality of edge signatures between the plurality of materials on a pairwise basis using a ratio of spectral bands that maximize spectral contrast between the materials of each pair;
• defining a plurality of spatio-spectral masks equal in number to the number of edge signatures;
• and
• applying the classifier to neighboring pixels of each candidate location.

A non-transitory machine-readable storage device having instructions stored thereon, which when executed by a processor, cause a machine to perform operations comprising:

• acquiring data of an image;
• processing the data including identifying a candidate location of an edge between two materials in the image, the data having neighboring pixels defining a portion of the candidate location, wherein processing the data including identifying the candidate location of an edge includes defining an edge signature between the two materials using a ratio of spectral bands that maximize spectral contrast between the two materials in the data of the image;
• applying a classifier to the neighboring pixels to determine, based on material changes, if the neighboring pixels are correlated to two different materials with respect to the candidate location, providing a modifier for a spatio-spectral mask;
• applying the spatio-spectral mask using the modifier to accept or reject the candidate location as an edge;
• and
• generating an image to a device, the candidate location in the image correlated to the application of the spatio-spectral mask using the modifier, wherein applying the spatio-spectral mask using the modifier includes using the modifier in an indicator matrix, each element of the indicator matrix correlated to a comparison of a difference of the spatio-spectral mask and the edge signature for a respective neighborhood of the candidate location with respect to a product of a respective modifier and a tolerance parameter.

A non-transitory machine-readable storage device having instructions stored thereon, which when executed by a processor, cause a machine to perform operations comprising:

• acquiring data of an image;
• processing the data including identifying a candidate location of an edge between two materials in the image, the data having neighboring pixels defining a portion of the candidate location;
• applying a classifier to the neighboring pixels to determine, based on material changes, if the neighboring pixels are correlated to two different materials with respect to the candidate location, providing a modifier for a spatio-spectral mask;
• applying the spatio-spectral mask using the modifier to accept or reject the candidate location as an edge;
• and
• generating an image to a device, the candidate location in the image correlated to the application of the spatio-spectral mask using the modifier, wherein the operations include:
• identifying a plurality of candidate locations of a plurality of edges of a plurality of materials by defining a plurality of edge signatures between the plurality of materials on a pairwise basis using a ratio of spectral bands that maximize spectral contrast between the materials of each pair;
• defining a plurality of spatio-spectral masks equal in number to the number of edge signatures;
• and
• applying the classifier to neighboring pixels of each candidate location.

A system comprising:

• an interface to receive data of image;
• and
• a processor operatively coupled to acquire the data received at the interface, the processor arranged to:
• process the data including identification of a candidate location of an edge between two materials in the image, the data having neighboring pixels defining a portion of the candidate location, wherein the processor arranged to process the data including identification of the candidate location includes the processor arranged to define an edge signature between the two materials using a ratio of spectral bands that maximize spectral contrast between the two materials in the data of the image;
• apply a classifier to the neighboring pixels to determine, based on material changes, if the neighboring pixels are correlated to two different materials with respect to the candidate location, providing a modifier for a spatio-spectral mask;
• apply the spatio-spectral mask using the modifier to accept or reject the candidate location as an edge;
• and
• generate an image to a device, the candidate location in the image correlated to a result of the application of the spatio-spectral mask using the modifier, wherein application of the spatio-spectral mask using the modifier includes use of the modifier in an indicator matrix, each element of the indicator matrix correlated to a comparison of a difference of the spatio-spectral mask and the edge signature for a respective neighborhood of the candidate location with respect to a product of a respective modifier and a tolerance parameter.

A system comprising:

• an interface to receive data of image;
• and
• a processor operatively coupled to acquire the data received at the interface, the processor arranged to:
• process the data including identification of a candidate location of an edge between two materials in the image, the data having neighboring pixels defining a portion of the candidate location;
• apply a classifier to the neighboring pixels to determine, based on material changes, if the neighboring pixels are correlated to two different materials with respect to the candidate location, providing a modifier for a spatio-spectral mask;
• apply the spatio-spectral mask using the modifier to accept or reject the candidate location as an edge;
• and
• generate an image to a device, the candidate location in the image correlated to a result of the application of the spatio-spectral mask using the modifier, wherein the processor is arranged to:
• identify a plurality of candidate locations of a plurality of edges of a plurality of materials by defining a plurality of edge signatures between the plurality of materials on a pairwise basis by use of ratios of spectral bands that maximize spectral contrast between the materials of each pair;
• define a plurality of spatio-spectral masks equal in number to the number of edge signatures;
• and
• apply the classifier to neighboring pixels of each candidate location.