Fabric Defect Detection And Classification Based On Extreme Learning Machine And Wavelet

Textile quality control is an important part of the fabric defect detection.However,most depends on manual nowadays, with the progress of the technology and increasingly high in the market of fabric, Traditional artificial detection method which the highlabor intensity, visual fatigue,subjectivity and other problems more and more can not meet the market requirements.Therefore, Automatic machine detection based on machine learning has become a hot research in recent years. The theme of this paper: defect defection and classification.And the paper selects 8 kinds of plain weave fabric defect images,the main work is as follows:1 Summarize the research status of fabric defect detection at home and abroad,and makes a classification on analysis method.2 Fabric defect automatic inspection system overview.Describing the fabric defect automatic inspection system and related components of the parameeters.3 Fabric image preprocessing. The homomorphic filtering method eliminate the fabric of an image of the object light is not uniform,then histogram equalization,contrast enhancement,highligt the fabric defect.4 Fabric defect detection based on the wavelet transform. Mechanism of this paper according to the principle and defects of woven fabric produced is proposed based on multi-scale fusion and adaptive threshold of wavelet modulus maxima fabric defect edge detection. The experimental results were compared to a variety of detection effect is more obvious.5 Research on multi feature fusion on the feature extraction method of fabric defect. On the basis of the above,the uniform Local Binary Pattern(median) is proposed;Then the texture features extracted from the operator and the texture features(Energy,Entropy,Contrast and Correlation) ofthe fabric are analyzed,and then the feature fusion is carried out by means of an adaptive weight.6 Introduction and improvement of the Extreme Learning Machine. Extreme Machine ELM(Learning) is an effective single hidden layer feedforward neural network SLFNS learning algorithm.The obviues feature is that it only needs to set the number of hidden nodes in the network. In the process of implementation of the algorithm,it does not need to adjust the input weights and the hidden layer bias,and can produce the unique optimal solution, that is to say,it has the characteristics of fast speed and high generalization ability. In order to make the ELM is better applied to fabric defect detection,First of all,Because the fabric texture is complex,the amount of data large,online ELM algorithm must be used to solve the problems in the treatment of massive data overall treatment of ELM;Then in order to improve the generalization ability of the algorithm and reduce the performance of the algorithm the depenence of nodes in the hidden layer,the sensitivity analysis based on pruning the hidden layer nodes. Finally,in order to ensure that the algorithm has good structural risk and the empirical risk,the improved algorithm will be regularization.