Multi-agent training of a color identification neural network

• US 10,643,099 B2
• Filed: 05/23/2019
• Issued: 05/05/2020
• Est. Priority Date: 09/29/2016
• Status: Active Grant

• Current Assignee: Datacolor, Inc.
• Original Assignee: Datacolor, Inc.
• Inventors: Xu, Zhiling
• Primary Examiner(s): Strege, John B
• Application Number:
  US16/421,060
• Publication Number:
  US 20190279023A1
• Time in Patent Office: 348 Days
• CPC Class Codes: G06F 18/24133 Distances to prototypes
• CPC Class Codes: G06N 3/02 Neural networks
• CPC Class Codes: G06V 10/56 relating to colour
• CPC Class Codes: G06V 10/764 using classification, e.g. of video objects
• CPC Class Codes: G06V 10/82 using neural networks

Abstract

Various embodiments of the systems and methods described herein are directed towards training an artificial neural network to identify color values of a sample by providing image data obtained through multiple image capture devices under a plurality of lighting conditions. The present invention also includes using a pre-trained neural network to identify the color values of a sample having an unknown color value by capturing an image of an unknown color sample and known color reference samples under any illumination or hardware configuration.

Claims

1. A color identification system for identifying the color of a sample object comprising:

a color reference chart having a center portion, and one or more color reference element groups, each color reference group comprised of a plurality of color reference elements having a known color value wherein the color reference element groups are arranged to be rotationally symmetric about the center portion;

an image capture device configured to acquire an image of the color image reference chart and the sample, where the sample is visible to the image capture device through the center portion of the image reference chart;

a processor having memory configured by program code executed thereby to;

extract RGB values corresponding to the pixels of the center portion as sample color values,

extract RGB values for each of the plurality of color reference elements as reference color values,

apply each one of the extracted sample color values and reference color values to one of a plurality of input nodes in an input layer of a pre-trained neural network, said neural network having;

i. an output layer containing output nodes corresponding to CIE color values;

ii. at least one intermediate layer containing a plurality of nodes connectable between the input and output in layers;

generate, at the respective output nodes, CIE color values characterizing the one or more color values of the sample;

compare the output CIE color values to a database of standard illuminant color values; and

identify the standard illuminant color value matching the output CIE color values.

2. The color identification system of claim 1, wherein the output CIE color values are CIELAB values.

3. The color identification system of claim 1, wherein color reference groups have fourfold rotational symmetry about the center portion.

4. The color identification system of claim 3, wherein each color reference element group comprises identical colors reference elements.

5. The color identification system of claim 4, wherein the processor is configured determine an average measured color value for each identical color reference element found across the one or more color reference groups.

6. The color identification system of claim 1, wherein the pre-trained neural network is trained by:

receiving, by the processor, color values for a plurality of known color samples under a plurality of illuminants;

setting, by the processor, the output values of the output layer to the known color value; and

setting, by the processor, the input values of the input layer to the color values of the color sample and the color values of the reference elements.

7. The color identification system of claim 1, wherein the known color values are obtained by measuring the sample color using a spectrophotometer.

8. The color identification system of claim 1, wherein the pre-trained neural network is trained by:

receiving by the processor one or more data objects from one or more remote users, each data object containing at least a remote color dataset, a sample color dataset and configuration data detaining information regarding the image capture device used to capture the remote color dataset;

setting by the processor the output values of the output layer to the values of the data in the sample color dataset;

setting by the processor the input values of the input layer to the color values of the remote color dataset;

generating additional input layer nodes corresponding to the configuration data;

training, by the processor, the ANN.

9. A color identification method for identifying the CIE color values of a sample object, the method comprising:

capturing, using an image device, an image of a color reference chart having a center portion, and a plurality of identical color reference element groups, the color reference element groups arranged so as to be rotationally symmetric about the center portion;

each color reference group comprised of a plurality of color reference elements having a known wavelength-dependent reflectance value;

extracting, using a processor having memory configured by program code executed thereby, the RGB values of the pixel depicting the center portion as sample color values;

extracting, using the processor, the RGB values for the pixels representing one or more of the plurality of color reference elements reference color values,

applying, using the processor, each one of the extracted sample RGB color values and reference RGB color values to one of a plurality of input nodes in an input layer of a pre-trained neural network, said neural network having;

i. an output layer containing output nodes configured to output CIE color values;

ii. at least one intermediate layer containing a plurality of nodes connectable between the input and output in layers;

generating, at the respective output nodes, CIE color values characterizing the one or more color values of the sample and;

comparing the output CIE color values to a database of standard illuminant color values;

identifying the standard illuminant color value matching the output CIE color values; and

outputting the color value to a display device.