Method and apparatus for analyzing an image to detect and identify patterns
First Claim
1. An apparatus for analyzing a 2-D representation of an object, said apparatus comprising:
- at least one sensor disposed to capture a 2-D representation of at least a portion of an object;
a memory that stores at least a portion of said 2-D representation received from said sensor;
a processor containing a program module operative to;
receive said stored portion of said 2-D representation;
derive a plurality of features from said stored portion of said representation;
provide said features to a multi-dimensional wavelet neural network, said multi-dimensional wavelet neural network incorporating a learning technique from the following group consisting of structure learning, parameter learning, or combined parameter and structure learning for classification performance, wherein said features are compared to a predetermined fault pattern to determine if the features represent a defect; and
produce a classification output indicative of whether said stored portion of said representation comprises a defect; and
a decision logic unit which receives said features and said classification output and determines if comparing said features to said predetermined fault patterns results in a classification output which is potentially indicative of a predetermined fault pattern, wherein said decision logic unit resolves said classification output by referencing said multi-dimensional wavelet neural network to provide a final declaration as to whether the classification output should be classified as a defect.
1 Assignment
0 Petitions
Accused Products
Abstract
A method and apparatus is provided which analyzes an image of an object to detect and identify defects in the object utilizing multi-dimensional wavelet neural networks. “The present invention generates a signal representing part of the object, then extracts certain features of the signal. These features are then provided to a multidimensional neural network for classification, which indicates if the features correlate with a predetermined pattern. This process of analyzing the features to detect and identify predetermined patterns results in a robust fault detection and identification system which is computationally efficient and economical because of the learning element contained therein which lessens the need for human assistance.”
246 Citations
42 Claims
-
1. An apparatus for analyzing a 2-D representation of an object, said apparatus comprising:
-
at least one sensor disposed to capture a 2-D representation of at least a portion of an object;
a memory that stores at least a portion of said 2-D representation received from said sensor;
a processor containing a program module operative to;
receive said stored portion of said 2-D representation;
derive a plurality of features from said stored portion of said representation;
provide said features to a multi-dimensional wavelet neural network, said multi-dimensional wavelet neural network incorporating a learning technique from the following group consisting of structure learning, parameter learning, or combined parameter and structure learning for classification performance, wherein said features are compared to a predetermined fault pattern to determine if the features represent a defect; and
produce a classification output indicative of whether said stored portion of said representation comprises a defect; and
a decision logic unit which receives said features and said classification output and determines if comparing said features to said predetermined fault patterns results in a classification output which is potentially indicative of a predetermined fault pattern, wherein said decision logic unit resolves said classification output by referencing said multi-dimensional wavelet neural network to provide a final declaration as to whether the classification output should be classified as a defect. - View Dependent Claims (2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15)
-
-
16. A method for recognizing defects in an object, comprising:
-
generating a 2-D digital representation of at least part of an object, the digital representation comprising a plurality of pixels;
extracting a plurality of features from said 2-D digital representation based on classification characteristics;
providing said features to a multi-dimensional wavelet neural network, said multidimensional wavelet neural network incorporating a learning technique from the following group consisting of structure learning, parameter learning, or combined parameter and structure learning for classification performance;
comparing said features to a predetermined fault pattern to determine if the feature represents a defect, and providing said features and said classification output from said multi-dimensional wavelet neural network to a logic unit to determine if comparing said features to said predetermined fault pattern results in an uncertain classification output, wherein said logic unit resolves uncertainty in said classification output by referencing said multi-dimensional wavelet neural network to determine whether the uncertain classification output should be classified as a defect. - View Dependent Claims (17, 18, 19, 20, 21, 22, 23, 24)
deriving at least one feature of said plurality of feature values from said 1-D signal;
convolving said 1-D signal with a plurality of wavelet functions utilizing fast wavelet transforms to produce a plurality of wavelet coefficients corresponding to at least one feature of said plurality of features; and
arranging said plurality of features into a feature vector.
-
-
22. The method of claim 16 further comprising a step of determining whether at least one of said plurality of features comprise a value indicative of a predetermined pattern, and if said value is not indicative of a predetermined pattern not providing said features to said multi-dimensional wavelet neural network.
-
23. The method of claim 16 wherein said object comprises a textile material.
-
24. The method of claim 16, wherein extracting a feature comprises a learning element wherein extraction is based on features present in the multi-dimensional wavelet network which are updated based on the resolution of conflicts in the classification output concerning prior derived signals from said logic unit.
-
25. A computer readable medium containing instructions for a computer comprising:
-
means for instructing the computer to read at least a portion of a 2-D digital image, said digital image comprising a plurality of pixels;
means for instructing the computer to generate a feature vector from said digital image based on classification characteristics;
means for instructing the computer to provide said feature vector to a multidimensional wavelet neural network;
means for instructing the computer to provide a classification output indicative of whether said feature vector corresponds to a predetermined pattern; and
means for resolving any conflicts arising from providing such classification output by referencing said multi-dimensional wavelet neural network to determine which one of a plurality of potentially identified classification outputs should be classified as a defect. - View Dependent Claims (26, 27, 28, 29, 30, 31, 32)
-
-
33. An apparatus for pattern recognition comprising:
-
an input that receives a 2-D representation of at least part of an object;
a memory that stores at least a portion of said 2-D representation; and
a processor that generates a plurality of feature values representing features of said at least one signal and that provides each of said feature values to a perceptron neural network comprising a plurality of neurons each defined by the function ψ
a,b={square root over (diag(a))}(diag(a)(x−
b)) where x is a vector comprising said feature values, a is a squashing matrix for the neuron and b is the translation vector for that neuron, said perceptron neural network providing a classification output indicative of whether said representation contains a predetermined pattern.- View Dependent Claims (34, 35, 36, 37, 38, 39, 40, 41)
-
-
42. An apparatus for analyzing a 2-D representation of an object, said apparatus comprising
at least one sensor disposed to capture a 2-D representation of at least a portion of an object; -
a memory that stores at least a portion of the 2-D representation; and
a processor that derives at least one signal from said 2-D representation that generates a feature representing a characteristic fault signature of at least one signal and that provides said feature to a multi-dimensional wavelet neural network which provides a classification output indicative of whether said representation comprises a predetermined pattern;
wherein said multi-dimensional wavelet neural network comprises a plurality of wavelet neurons each defined by ψ
a,b={square root over (diag(a)|)}ψ
(diag(a)(x−
b)where x is a vector comprising said feature, a is a squashing matrix for that neuron and b is the translation vector for that neuron.
-
Specification