Material classification

Patent metadata

Field	Value
Publication number	US-9989463-B2
Application number	US-201514635848-A
Country	US
Kind code	B2
Filing date	Mar 2, 2015
Priority date	Jul 2, 2013
Publication date	Jun 5, 2018
Grant date	Jun 5, 2018

Abstract

Material classification of an object is provided. Parameters for classification are accessed. The parameters include a selection to select a subset of angles for classification, a selection to select a subset of spectral bands for classification, a selection to capture texture features, and a selection to compute image-level features. The object is illuminated and a feature vector is computed based on the parameters. The material from which the object is fabricated is classified using the feature vector.

First claim

The invention claimed is:

1. A method for material classification of an object, comprising:
accessing parameters for classification by a first controller, wherein the parameters include a selection to select a subset of angles for classification, a selection to select a subset of spectral bands for classification, a selection to capture texture features, and a selection to compute image-level features;
illuminating the object by a light source;
capturing an image of the illuminated object by an image capturing device;
computing a feature vector of the captured image of the illuminated object by a second controller based on the parameters; and
classifying the material from which the object is fabricated by the second controller using the feature vector.

2. The method according to claim 1 , wherein responsive to a selection to select a subset of angles, an incident illumination angle of each light element cluster illuminating the object is selected based on a mathematical clustering analysis of labeled training data captured under a superset of a second number of light element clusters from different incident angles, so as to select a subset of incident illumination angles by a first number of light element clusters from the superset of the second number of light element clusters, the first number being smaller than the second number, wherein each light clement cluster comprises a plurality of light elements.

3. The method according to claim 2 , further comprising:
calculating a feature vector representation for training data in a database of labeled training data captured under the superset of the second number of light element clusters from different incident angles;
performing mathematical clustering on the feature vector representations so as to identify a subset of mathematically significant clusters of data for a corresponding the first number of incident illumination angles; and
selecting incident illumination angles for the light element clusters based on the mathematical clusters.

4. The method according to claim 3 , wherein the training data is captured from a superset of a relatively large number of exitant angles, and wherein the method further comprises selecting a subset of a relatively small number of mathematically significant clusters of data for a corresponding small number of exitant angles by using mathematical clustering.

5. The method according to claim 3 , wherein the number of mathematically significant clusters is selected automatically using a mathematical clustering algorithm which includes convex clustering.

6. The method according to claim 1 , wherein responsive to a selection to select a subset of spectral bands, multiple spectral bands of light reflected from the illuminated object are measured,
wherein with respect to the measured multiple spectral bands, wavelengths for the multiple spectral bands are selected by analysis of a database of labeled training material samples within a multi-class classification framework, captured using a relatively large number of spectral bands, so as to select a subset of a relatively fewer number of spectral bands, wherein the selected spectral bands in the subset have a significant aptitude for distinguishing between different classifications of materials in the database.

7. The method according to claim 6 , wherein the multiple spectral bands are selected by steps which include:
computing feature vector representations of the materials in the database;
learning a set of weights representing the importance of the spectral bands on the features in each binary classification task;
converting the weights from the binary classification tasks to a set of weights for the multi-class classification framework using a mapping function; and
selecting spectral bands of highest weight as the selected spectral bands.

8. The method of claim 7 , wherein learning a set of weights representing the importance of the spectral bands comprises application of an Adaboost algorithm.

9. The method according to claim 7 , wherein the feature vector is computed from the measured spectral bands.

10. The method according to claim 1 , wherein responsive to selection to capture texture features, a local binary pattern feature vector is calculated based on images of the illuminated object.

11. The method according to claim 1 , wherein responsive to a selection to compute image-level features, BRDF image slices are captured for the object, low-level features of each image slice are clustered into at least two clusters, and an image-level feature vector representation is computed for each image slice with entries that are weighted means of the clusters.

12. The method according to claim 11 , wherein the low-level features of each image slice are clustered into at least three clusters including a first cluster for specular reflections, a second cluster for diffuse reflections, and a third cluster for dark reflections.

13. The method according to claim 11 , further comprising:
computing feature vector representations of each slice by sorting all entries of the intermediate feature vector representations by the mean of the corresponding clusters.

14. The method according to claim 11 , wherein clustering includes application of K-means clustering on the low-level features for each image slice.

15. An apparatus for selecting incident illumination angles for illumination of an object by respective light element clusters, comprising:
a non-transitory memory for storing computer-executable process steps and for storing labeled training data; and
one or more processors for executing the computer-executable process step stored in the non-transitory memory;
wherein the computer-executable process steps include steps wherein the incident illumination angle of each light element cluster is selected based on a mathematical clustering analysis of labeled training data captured under a superset of a second number of light element clusters from different incident angles, so as to select a subset of incident illumination angles by a first number of light element clusters from the superset of the second number of light element clusters, the first number being smaller than the second number, wherein each light clement cluster comprises a plurality of light elements.

16. An apparatus for material classification of the object, comprising:
a first controller configured to access parameters for classification, wherein the parameters include a selection to select a subset of angles for classification, a selection to select a subset of spectral bands for classification, a selection to capture texture features, and a selection to compute image-level features;
a light source configured to illuminate the object;
an image capturing device configured to capture an image of the illuminated object; and
a second controller configured to compute a feature vector of the captured image of the illuminated object based on the parameters and to classify the material from which the object is fabricated by using the feature vector.

Assignees

• Canon Kk

Inventors

• Skaff Sandra
• Liu Chao

Classifications

• G01N21/55 Primary

  Specular reflectivity · CPC title

• G01N2201/0627

  Use of several LED's for spectral resolution · CPC title

• G01N2021/845

  Objects on a conveyor · CPC title

• B07C5/342

  according to optical properties, e.g. colour {(according to radiation transmittivity B07C5/3416)} · CPC title

• G01N2201/12

  Circuits of general importance; Signal processing · CPC title