Method and apparatus for analyzing an image to detect and identify patterns

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.

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.

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 ψa,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.

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 ψa,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.