Detection of textural defects using a one class support vector machine
View Patent ↗Method for detecting textural defects in an image. The image, which may have an irregular visual texture, may be received. The image may be decomposed into a plurality of subbands. The image may be portioned into a plurality of partitions. A plurality of grey-level co-occurrence matrices (GLCMs) may be determined for each partition. A plurality of second-order statistical attributes may be extracted for each GLCM. A feature vector may be constructed for each partition, where the feature vector includes the second order statistical attributes for each GLCM for the partition. Each partition may be classified based on the feature vector for the respective partition. Classification of the partitions may utilize a one-class support vector machine, and may determine if a defect is present in the image.
1. A method for detecting textural defects in an image, the method comprising:
receiving the image, wherein the image comprises an irregular visual texture;
decomposing the image into a plurality of subbands using wavelet frames;
constructing a plurality of feature vectors for the image, wherein the plurality of feature vectors comprise at least one feature vector from each of the subbands;
classifying the image based on the plurality of feature vectors for the image, wherein said classifying utilizes a one-class support vector machine (SVM), wherein said classifying determines if a textural defect is present in the image.
2. The method of claim 1 ,
wherein the one-class SVM is trained using only defect-free images.
3. The method of claim 1 , further comprising:
partitioning the image into a plurality of partitions;
wherein the plurality of feature vectors comprise a feature vector for each partition of the plurality of partitions.
4. The method of claim 3 ,
wherein the plurality of partitions do not overlap one another.
5. The method of claim 3 ,
wherein at least a subset of the plurality of partitions overlap one another.
6. The method of claim 3 , further comprising:
determining at least one grey-level co-occurrence matrix (GLCM) for each partition; and
extracting a plurality of second-order statistical attributes for each GLCM;
wherein the feature vector for each partition comprises the plurality of second order statistical attributes for each of the at least one GLCMs for the respective partition.
7. The method of claim 6 ,
wherein each of the at least one GLCMs is direction independent.
8. The method of claim 1 ,
wherein the wavelet frames are biorthogonal.
9. The method of claim 1 ,
wherein each of the plurality of subbands corresponds to a different resolution and/or orientation.
10. The method of claim 1 ,
wherein the image has a stochastic intensity distribution.
11. A computer readable memory medium comprising program instructions for detecting textural defects in an image, wherein the program instructions are executable to:
receive the image, wherein the image comprises an irregular visual texture;
decompose the image into a plurality of subbands using wavelet frames;
construct a plurality of feature vectors for the image, wherein the plurality of feature vectors comprise at least one feature vector from each of the subbands;
classify the image based on the plurality of feature vectors for the image, wherein said classifying utilizes a one-class support vector machine (SVM), wherein said classifying determines if a textural defect is present in the image.
12. The computer readable memory medium of claim 11 ,
wherein the one-class SVM is trained using only defect-free images.
13. The computer readable memory medium of claim 11 , wherein the program instructions are further executable to:
partition the image into a plurality of partitions;
wherein the plurality of feature vectors comprise a feature vector for each partition of the plurality of partitions.
14. The computer readable memory medium of claim 13 ,
wherein the plurality of partitions do not overlap one another.
15. The computer readable memory medium of claim 13 ,
wherein at least a subset of the plurality of partitions overlap one another.
16. The computer readable memory medium of claim 13 , wherein the program instructions are further executable to:
determine at least one grey-level co-occurrence matrix (GLCM) for each partition; and
extract a plurality of second-order statistical attributes for each GLCM;
wherein the feature vector for each partition comprises the plurality of second order statistical attributes for each of the at least one GLCMs for the respective partition.
17. The computer readable memory medium of claim 16 ,
wherein each of the at least one GLCMs is direction independent.
18. The computer readable memory medium of claim 11 ,
wherein the wavelets frames are biorthogonal.
19. The computer readable memory medium of claim 11 ,
wherein each of the plurality of subbands corresponds to a different resolution and/or orientation.
20. The computer readable memory medium of claim 11 ,
wherein the image has a stochastic intensity distribution.
21. A visual inspection system for detecting textural defects in an image, comprising:
an input for receiving images;
a processor coupled to the input for receiving images;
a memory medium coupled to the processor, wherein the memory medium comprises program instructions executable by the processor to:
receive the image, wherein the image comprises an irregular visual texture;
decompose the image into a plurality of subbands using wavelet frames;
construct a plurality of feature vectors for the image, wherein the plurality of feature vectors comprise at least one feature vector from each of the subbands;
classify the image based on the plurality of feature vectors for the image, wherein said classifying utilizes a one-class support vector machine (SVM), wherein said classifying determines if a textural defect is present in the image.
22. The visual inspection system of claim 21 ,
wherein the one-class SVM is trained using only defect-free images.
23. A method for detecting textural defects in an image, the method comprising:
receiving the image, wherein the image comprises an irregular visual texture;
decomposing the image into a plurality of subbands, wherein said decomposing comprises utilizing biorthogonal wavelet frames, wherein each of the plurality of subbands corresponds to a different resolution and/or orientation;
partitioning the image into a plurality of partitions;
determining a plurality of grey-level co-occurrence matrices (GLCMs) for each partition, wherein each GLCM of a respective partition corresponds to a respective subband of the plurality of subbands;
extracting a plurality of second-order statistical attributes for each GLCM, wherein the plurality of second-order statistical attributes comprise entropy, dissimilarity, contrast, homogeneity, and correlation;
constructing a feature vector for each partition, wherein the feature vector for a respective partition comprises the plurality of second-order statistical attributes for each subband for the respective partition;
classifying each partition based on the feature vector of the respective partition, wherein said classifying utilizes a one-class support vector machine (SVM), wherein said classifying determines if a textural defect is present in the image.
24. The method of claim 23 ,
wherein the one-class SVM is trained using only defect-free samples.
25. The method of claim 23 ,
wherein said receiving, decomposing, partitioning, determining, extracting, constructing, and classifying are performed substantially in real-time.
26. The method of claim 23 ,
wherein if there is a textural defect present in the image, said classifying each partition determines a location of the textural defect.