SORTING OF MATERIALS BASED ON IDENTIFIED SIGNATURES
A system and method for classifying and sorting a plurality of materials utilizing a machine learning system in order to identify signatures of each of the materials corresponding to elemental compositions specific to each of the materials, which are then sorted into separate groups as a function of such identified signatures.
1 . A system for sorting a plurality of materials comprising:
a camera configured to capture image data associated with a plurality of pixels representing one or more characteristics of each of the materials;
a machine learning system configured to receive the captured image data as input on a per pixel basis to identify signatures for each of the materials, wherein each of the signatures corresponds to a specific elemental composition; and
a sorter configured to sort the materials as a function of the identified signatures.
2 . The system as recited in claim 1 , wherein the machine learning system processes a hierarchy of image features extracted from the image data in order to identify the signatures for each of the materials.
3 . The system as recited in claim 2 , wherein the hierarchy of image features extracted from the image data for each of the materials is processed in the machine learning system by a neural network in order to identify the signatures for each of the materials.
4 . The system as recited in claim 1 , wherein the plurality of materials contains a first material composed of a first elemental composition and a second material composed of a second elemental composition, wherein the first elemental composition is different than the second elemental composition, wherein the machine learning system is configured to identify a first signature for the first material as a function of the image data captured from the first material, and wherein the machine learning system is configured to identify a second signature for the second material as a function of the image data captured from the second material, and wherein the sorter is configured to sort the first material from the second material as a function of the first and second signatures.
5 . The system as recited in claim 4 , wherein the first material is composed of a first metal alloy, and the second material is composed of a second metal alloy, wherein the first metal alloy is composed of the first elemental composition, and wherein the second metal alloy is composed of the second elemental composition.
6 . The system as recited in claim 5 , wherein the first metal alloy is wrought aluminum and the second metal alloy is cast aluminum.
7 . The system as recited in claim 6 , wherein the system is configured to sort between the wrought aluminum and the cast aluminum based solely on the image data captured from the materials.
8 . The system as recited in claim 4 , wherein the first signature comprises a first hierarchy of image features that visually represent the first elemental composition of the first material, and wherein the second signature comprises a second hierarchy of image features that visually represent the second elemental composition of the second material, wherein the first hierarchy of image features is different than the second hierarchy of image features.
9 . The system as recited in claim 8 , wherein the first signature comprises a first hierarchy of image features extracted from a homogeneous set of samples of the first material, wherein the first hierarchy of image features is unique to the first elemental composition of the first material, and wherein the second signature comprises a second hierarchy of image features extracted from a homogeneous set of samples of the second material, wherein the second hierarchy of image features is unique to the second elemental composition of the second material.
10 . A method for sorting a plurality of materials comprising:
capturing image data associated with a plurality of pixels representing one or more characteristics of each of the materials;
receiving the captured image data on a per pixel basis as input to a machine learning system to identify signatures for each of the materials, wherein each of the signatures corresponds to a specific elemental composition; and
sorting the materials as a function of the identified signatures.
11 . The method as recited in claim 10 , wherein the machine learning system processes a hierarchy of image features extracted from the image data in order to identify the signatures for each of the materials.
12 . The method as recited in claim 11 , wherein the hierarchy of image features extracted from the image data for each of the materials is processed in the machine learning system by a neural network in order to identify the signatures for each of the materials.
13 . The method as recited in claim 10 , wherein the plurality of materials contains a first material composed of a first elemental composition and a second material composed of a second elemental composition, wherein the first elemental composition is different than the second elemental composition, wherein the machine learning system is configured to identify a first signature for the first material as a function of the image data captured from the first material, and wherein the machine learning system is configured to identify a second signature for the second material as a function of the image data captured from the second material, and wherein the first material is sorted from the second material as a function of the first and second signatures.
14 . The method as recited in claim 13 , wherein the first material is composed of a first metal alloy, and the second material is composed of a second metal alloy, wherein the first metal alloy is composed of the first elemental composition, and wherein the second metal alloy is composed of the second elemental composition.
15 . The method as recited in claim 14 , wherein the first metal alloy is wrought aluminum and the second metal alloy is cast aluminum, wherein the method is configured to sort between the wrought aluminum and the cast aluminum based solely on the image data captured from the materials.
16 . The method as recited in claim 13 , wherein the first signature comprises a first hierarchy of image features that visually represent the first elemental composition of the first material, and wherein the second signature comprises a second hierarchy of image features that visually represent the second elemental composition of the second material, wherein the first hierarchy of image features is different than the second hierarchy of image features.
17 . The method as recited in claim 16 , wherein the first signature comprises a first hierarchy of image features extracted from a homogeneous set of samples of the first material, wherein the first hierarchy of image features is unique to the first elemental composition of the first material, and wherein the second signature comprises a second hierarchy of image features extracted from a homogeneous set of samples of the second material, wherein the second hierarchy of image features is unique to the second elemental composition of the second material.
18 . A computer program product stored on a computer readable storage medium, which when executed performs a method for sorting a plurality of materials comprising:
receiving image data associated with a plurality of pixels representing one or more characteristics of each of the materials;
receiving the captured image data on a per pixel basis as input to a machine learning system to identify signatures for each of the materials, wherein each of the signatures corresponds to a specific elemental composition; and
sending to an automated sorting device information so that the automated sorting device can sort the materials as a function of the identified signatures.
19 . The computer program product as recited in claim 18 , wherein the machine learning system processes a hierarchy of image features extracted from the image data in order to identify the signatures for each of the materials.
20 . The computer program product as recited in claim 19 , wherein the hierarchy of image features extracted from the image data for each of the materials is processed in the machine learning system by a neural network in order to identify the signatures for each of the materials.
21 . The computer program product as recited in claim 18 , wherein the plurality of materials contains a first material composed of a first elemental composition and a second material composed of a second elemental composition, wherein the first elemental composition is different than the second elemental composition, wherein the machine learning system is configured to identify a first signature for the first material as a function of the image data captured from the first material, and wherein the machine learning system is configured to identify a second signature for the second material as a function of the image data captured from the second material, and wherein the first material is sorted from the second material as a function of the first and second signatures.
22 . The computer program product as recited in claim 21 , wherein the first signature comprises a first hierarchy of image features that visually represent the first elemental composition of the first material, and wherein the second signature comprises a second hierarchy of image features that visually represent the second elemental composition of the second material, wherein the first hierarchy of image features is different than the second hierarchy of image features.
23 . The computer program product as recited in claim 22 , wherein the first signature comprises a first hierarchy of image features extracted from a homogeneous set of samples of the first material, wherein the first hierarchy of image features is unique to the first elemental composition of the first material, and wherein the second signature comprises a second hierarchy of image features extracted from a homogeneous set of samples of the second material, wherein the second hierarchy of image features is unique to the second elemental composition of the second material.