US6571228B1
Expired / Lapsed
US
⏱️ 0.0 years remaining
Hybrid neural networks for color identification
Assignee: Individual
•
Filing: 2000-08-09
•
Granted: 2003-05-27
•
Expires: 2020-08-09
Abstract
Abstract A method uses a hybrid neural network including a self organizing mapping neural network (SOM NN) and a, back-propagation neural network (BP NN) for color identification. In the method the red, green and blue (RGB) of color samples are input as features of training samples and are automatically classified by way of SOM NN. Afterwards, the outcomes of SOM NN are respectively delivered to various BP NN for further learning; and the map relationship of the input and the output defines the X,Y, Z corresponding the x, y and z values of a coordinate system of the standard color samples of RGB and IT8. By way of the above learning structure, a non-linear model of color identification can be set up. After color samples are self organized and classified by SOM NN network, data can be categorized in clusters as a result of characteristic difference thereof. Then the data are respectively sent to BP NN for learning whereby-the learning system not only can be quickly converged but also lower error discrepancy in operation effectively.
Inventors
- 👤 Po-Tong Wang
- 👤 Ching-Han Chen
Bibliographic Details
Application Number
—
Jurisdiction
US
Data Source
google_patents
Last Synchronized
2026-10-25 00:38
CPC Classifications
G; G06; G06V; G06V10/00; G06V10/40; G06V10/56; G06N; G06N3/00; G06N3/02; G06N3/04; G06N3/045; G06N3/0499; G06N3/08; G06N3/09
Claims Analysis
Total Claims: 13
Claim 1
Independent Claim
185 words
A method of color identification by way of neural networks, wherein: a color identification system comprises a color image input device, a self organizing mapping neural network(SOM NN) and a back-propagation neural network (BP NN) and a color output device; said color image input device is connected to said SOM NN and said BP NN network is connected to said color output device; said method includes the following steps: Step 1: digitalizing color images by way of said image input device transforming said color images into a form acceptable by a computer; then extracting color features thereof by two sets of different region description devices for automatic classification by means of said SOM NN; Step 2: transmitting respectively output neuron samples of an identical kind from SOM NN to said BP NN, according to a sorting result of said SOM NN for supervision type learning; and mapping outputs of said learning being X, Y, Z of a (x,y,z) coordinate of standard color samples of IT8; after the above learning procedures, a final learning procedure is reached only when a preset learning termination condition is met.
Claim 2
Independent Claim
185 words
A method of color identification by way of neural networks, wherein: a color identification system comprises a color image input device, a self organizing mapping neural network(SOM NN) and a back-propagation neural network (BP NN) and a color output device; said color image input device is connected to said SOM NN and said BP NN network is connected to said color output device; said method includes the following steps: Step 1: digitalizing color images by way of said image input device transforming said color images into a form acceptable by a computer; then extracting color features thereof by two sets of different region description devices for automatic classification by means of said SOM NN; Step 2: transmitting respectively output neuron samples of an identical kind from SOM NN to said BP NN, according to a sorting result of said SOM NN for supervision type learning; and mapping outputs of said learning being X, Y, Z of a (x,y,z) coordinate of standard color samples of IT8; after the above learning procedures, a final learning procedure is reached only when a preset learning termination condition is met.
Claim 3
Dependent Claim
19 words
The method of color identification as claimed in claim 1 wherein said color image input device is a scanner.
Claim 4
Dependent Claim
22 words
The method of color identification as claimed in claim 1 wherein said color image input device is a CCD (charge-coupled device) camera.
Claim 5
Dependent Claim
20 words
The method of color identification as claimed in claim 1 wherein said color image input device is a color sensor.
Claim 6
Dependent Claim
173 words
The method of color identification as claimed in claim 1 wherein: Color features are extracted by way of two region description devices of different functions which measures on color character values in following steps: A. A first set of said region description device adopts 4-stage previous statistical momentum: 1-stage: A mean value for evaluation of intensity of mean gray level of an image area block; 2-stage: A variance value for description of a degree of dispersion of gray level of an area block; 3-stage: A skewness value for measurement of symmetry of a linear square diagram of gray level; 4-stage: A kurtosis value for measurement of a drawing mode of a gravy level distribution; B. Another set of region description device makes use of 5 specific statistical distribution values of said linear square diagram of gray level: (1) A non-zero maximum gray level value on said linear square diagram; (2) A non-zero minimum gray level value on said linear square diagram; (3) A middle position number; (4) Multiple numbers; (5) A full distance.
Claim 7
Dependent Claim
43 words
The method of color identification as claimed in claim 1 wherein said SOM NN extracts features of color samples as input vectors and sets numbers of mapping output layers and system learning parameters so as to allow automatic classification to be performed thereby.
Claim 8
Dependent Claim
49 words
The method of color identification as claimed in claim 1 wherein said BP NN can take tri-stimulating X,Y, Z values of a xyz coordinate obtained by way of a spectrometer for color identification of a sample instead of choice of IT8 in a mapping output of a learning process.
Claim 9
Dependent Claim
119 words
The method of color identification as claimed in claim 1 wherein said SOM NN has a learning rule set as below: Setting a number of output layers, initial neighboring region N c (0) and a node value W i (0): Seeking a winning neuron c, having a smallest distance from an input feature vector X, i.e.; ∥ X−W c ∥=min{∥ X−W i ∥}Wi ( t )+α( t )[ X ( t )− Wi ( t )] if iεN c ( t ) and Wi ( t+ 1)= Wi ( t ) if i∉N c ( t ) Wherein α(t) is learning rate, and α(t) and neighboring region N c (t) are decreasing function with time in a learning process.
Claim 10
Dependent Claim
104 words
The method of color identification as claimed in claim 1 wherein said BP NN has a learning rule set as below: taking a sigmoid function as a conversion function and O = 1 1 + - ( ∑ wX + bias ) wherein O is an output value of said conversion function, w is a weight, X is an input unit; w and bias sector has a randomly given initial value; said weight has a renewal rule as below: Δ w ( n )=−ε∂ E/∂w+αΔw ( n− 1) wherein said α is an inertial factor, E is an MSE (sum of errors' square).
Claim 11
Dependent Claim
27 words
The method of color identification as claimed in claim 1 wherein output Xn′, Yn′,Zn′ of said BP NN after a learning process are outcomes of color identification.
Claim 12
Dependent Claim
49 words
The method of color identification as claimed in claim 1 wherein competition type learning principles including a clustering algorithm such as K-means, fuzzy c-means or vector quantization are used for automatic classification; and then said BP NN is integrated for supervising type learning to carry out color space transformation.
Claim 13
Dependent Claim
49 words
The method of color identification as claimed in claim 1 wherein said the number of output layer neurons of said SOM NN and learning parameters are adjusted to repeat a learning process when an outcome of said color identification system is not acceptable until a satisfactory outcome is obtained.
References Cited (Backward Citations)
USPTO Annuity & Maintenance Windows
| Annuity Stage | Window Opens | Payment Deadline | Surcharge Period Until | Estimated Fee | Status |
|---|---|---|---|---|---|
| 3.5 Years (Year 4) | 2006-05-26 | 2006-11-25 | 2007-05-27 | $2000.00 | Window Closed / Lapsed |
| 7.5 Years (Year 8) | 2010-05-26 | 2010-11-25 | 2011-05-27 | $3760.00 | Window Closed / Lapsed |
| 11.5 Years (Year 12) | 2014-05-26 | 2014-11-25 | 2015-05-27 | $7700.00 | Window Closed / Lapsed |