Systems and methods for color agnostic material face detection

Abstract

Various examples are provided related to face identification of material. An image can be captured by a vision system and feature parameters determined and compared to a material feature database to determine which face of the material is being presented. The vision system can employ a set of parameters for configuration to acquire the image. The system can communicate the identification to downstream processes in real time. A near-universal, color agnostic, angular orientation independent identification of material faces can be determined without the need for physical manipulation of the material.

Therefore, at least the following is claimed: 
     
         1 . A method for material face identification, comprising:
 capturing an image of at least a portion of a side of a piece of material in a work area, the image captured by a vision system;   generating a confidence score based upon feature data determined from the image of the first side of the piece of material; and   determining whether a correct face of the piece of material is correctly presented in the work area based upon the confidence score.   
     
     
         2 . The method of  claim 1 , wherein the vision system is configured based upon imaging parameters associated with the piece of material. 
     
     
         3 . The method of  claim 2 , wherein the imaging parameters are identified by a material ID associated with the piece of material. 
     
     
         4 . The method of  claim 3 , wherein the imaging parameters are retrieved from an imaging parameter database based upon the material ID. 
     
     
         5 . The method of  claim 3 , wherein the imaging parameters are obtained in response to identification of the material ID. 
     
     
         6 . The method of  claim 3 , wherein a machine learning model is selected based upon the material ID, the machine learning model trained to generate the confidence score based upon the feature data determined from the image of the piece of material. 
     
     
         7 . The method of  claim 1 , wherein a machine learning model associated with the piece of material generates the confidence score based upon the feature data extracted from the image of the first side of the piece of material. 
     
     
         8 . The method of  claim 1 , wherein the feature data comprises a compact feature set generated from material features extracted from the image of the first side of the piece of material. 
     
     
         9 . The method of  claim 8 , wherein the material features comprise filtered channel response that are devoid of color information. 
     
     
         10 . The method of  claim 1 , comprising flipping the piece of material to expose a second side of the piece of material in the work area. 
     
     
         11 . The method of  claim 10 , wherein the piece of material is flipped in response to the correct face not being correctly presented. 
     
     
         12 . The method of  claim 10 , comprising:
 capturing an image of at least a portion of the second side of the piece of material; and   generating a second confidence score based upon feature data determined from the image of the second side of the piece of material.   
     
     
         13 . The method of  claim 12 , comprising determining whether the correct face of the piece of material is correctly presented in the work area based upon the two confidence scores. 
     
     
         14 . A method for material face identification, comprising:
 obtaining a plurality of sample images of one or more portions of a face of a sample of material, the plurality of sample images captured by a vision system configured based upon imaging parameters associated with the sample of material;   training a machine learning model to generate a confidence score corresponding to the face of the sample of material, the machine learning model trained using feature data determined from at least a portion of the plurality of sample images; and   storing the trained machine learning module in a model data base, the trained machine learning model identified in the model data base by a material ID associated with the material.   
     
     
         15 . The method of  claim 14 , wherein the imaging parameters are obtained from an imaging parameter database based upon the material ID. 
     
     
         16 . The method of  claim 15 , wherein the imaging parameters are determined based upon quality scores generated for a plurality of test images of one or more pieces of the material, the plurality of test images captured with different combinations of imaging parameters. 
     
     
         17 . The method of  claim 16 , wherein the imaging parameters corresponding to the quality score that is highest are used to capture the plurality of sample images of the face of the sample of material. 
     
     
         18 . The method of  claim 16 , comprising:
 capturing the plurality of test images of at least a portion of a face of at least one piece of material in a work area, the plurality of test images captured by the vision system with the different combinations of imaging parameters;   generating a quality score for each test image of the plurality of test images based upon feature data determined from that test image; and   selecting the imaging parameters associated with the material based upon a comparison of the quality scores.   
     
     
         19 . The method of  claim 14 , wherein a compact feature set generated from the material features is used to train the machine learning model corresponding to the sample of material. 
     
     
         20 . The method of  claim 14 , wherein the machine learning model is trained for both faces of the sample of material.