The Research On Image Processing Of Fabric Pilling

The evaluation of fabric pilling has always been an important content in the field of textile testing,since pills not only spoil the fabric appearance but also reduce the wearing comfort.Traditionally,the fabric pilling grade is judged according to the standard sample.But it is greatly subjective and will produce different results among different people.To enhance the objectivity of fabric pilling evaluation,using digital image processing technique to evaluate the grade of fabric automatically through the research on image processing of fabric pilling is of great significance.The main contents are as follows:1.For the fabric pilling image obtained by circular locus method which is commonly used in our country,an improved Hough transform algorithm is proposed to detect the circle in the acquisition of fabric pilling image,which get only the fabric pilling area.Compared with manual intercept method,the proposed approach improves the automaticity of pilling evaluation.2.Multi-resolution analysis of wavelet is used to decompose the pilling image,and sub-images containing different frequency information is obtained.In order to remove non-texture noise,the appropriate layer is selected,and then reconstruct pilling image after low and high frequency decomposition coefficients is processed.With the further analysis of fabric image,wavelet energy indicator is constructed to study the variation of energy and locate fabric texture layer,and then fabric roughness is determined.3.Fabric pilling image noise not only includes non-texture noise generated by the uneven illumination and other reasons,but also includes inherent fabric texture noise.Relationship between wavelet multi-resolution decomposition and noise is studied to realize the removal of non-texture noise.By analyzing the characteristic of Gabor filter,different Gabor filter is constructed to smooth fine texture and detect the edge of rough texture respectively.4.Relationship between wavelet transform and the choice of Gabor filter parameters such as center frequency and scale is established.According to the correlation of image information and entropy,entropy curve constructed by the filtered image in the typical direction is analyzed,and then the fabric texture direction which is used to determine the direction of Gabor filter is judged.Inaddition,the 2D Gabor filtering algorithm is optimized to improve execution speed of evaluation system.5.Three kinds of segmentation methods is applied to segment the de-noised image.Through comparison of the results from the three segmentation methods,the appropriate method is selected.According to the pearson correlation coefficient which is used as the indicator to measure the relationship between pilling feature parameters and pilling grades,pill density,pill number,and standard deviation of pill areas is chosen as the standard feature parameters finally.The experimental results show that,the fabric image noise including both texture and non-texture noise can be filtered effectively.Fabric pilling information can be extracted from pilling image accurately,which is important in pilling evaluation.The fabric pilling rating system proposed in this paper through the research and analysis of the fabric pilling image can carry out evaluation of fabric pilling accurately.