Objective Evaluation Of Fabric Pilling Based On Image Processing

Pilling is an important indicator of evaluating fabric performance and quality. Generally,for the assessment of fabric pilling grades, pilled fabric specimens are compared with a set ofstandard pilling images by an expert to determine the grade of pilling. However, thissubjective evaluation might be inconsistent and inaccurate in different evaluation personnel.For objective pilling evaluating process, a number of automated image analysis systems havebeen developed and reported. All of these methods employ either expensive and complicatedequipment or complex image processing algorithms involving multiple stages. This paper isfocus on the more reliable and accurate method for objective fabric pilling evaluation.A new objective fabric pilling evaluation method based on wavelet transform and thelocal binary pattern (LBP) is developed. The surface pills are identified from thehigh-frequency noise, fabric texture, and illuminative variation of a pilled fabric image by thetwo-dimensional discrete wavelet transform (2DDWT). The energies of each detailedsub-image at scales4-6in three orientations (horizontal, vertical and diagonal) and the LBPfeatures of the reconstructed detail image from scales3-6are calculated as elements of thepilling feature vector to characterize the pilling intensity. These feature values are normalizedand the vector dimensions are reduced by principal component analysis (PCA). Then thesupport vector machine (SVM), a kind of data mining tool, is used as a classifier to classifythe pilling grades.At the same time, the other algorithm was used for feature extraction in the experiment.As an improvement of the traditional Gabor wavelet, the Log-Gabor filter has been widelyused in the filed of machine vision. This paper propose a new method of rating the fabricpilling objectively using the characteristics of Log-Gabor filter combined with PrincipalComponent Analysis (PCA) and Support Vector Machine(SVM). For the high dimensionalproblem of Log-Gabor features, this paper has improved the traditional feature extractionmethods, which could improve the computation efficiency and reduce the running time.The result suggests that the proposed method could effectively capture the pillinginformation and successfully evaluate the pilling intensity of knitted fabrics. The highestcorrect recognition rate is95%, so the method could be applicable to practical objective pilling evaluation.