Automatic carpet wear classification based on support vector machine and Haralick descriptors

The Flemish carpet industry starts to look for a quality label to attest the resistance-to-wear of their product. Currently, this quality label is determined through visual assessment by human experts. However, this approach has some disadvantages, mainly related to problems generated by the subjectivity of human evaluators. The idea of using computers for automatic quality labelling is not new, but until now most results were not satisfying; the human expert still seems to be the best choice. This paper presents an approach to this problem using 3D lasers for scanning the carpets. Resampling the 3D data on different grid sizes, a 2D image is obtained and then a technique based on Haralick descriptors is applied to detect the carpet features. These features represent the inputs to a classifier system which is based on support vector machine (SVM). The performance of our proposed technique gives an average of about 92% correct labelling.