Surface Defect Identification and Grouping of Intermittent Leather Images using Linear Discriminant Model

Abstract

Scrutiny of intermittent leather is accepted through visual analysis on the natural material by an experienced individual based on many parameters which includes surface defects as a parameter. Such results comprising of base color, other than base color, share of regions, share of cutting area, share of cutting value, position wise length and position wise breadth will determine the value of the leather and surprisingly the result will vary form one experienced person to another. Hence, a new method for grouping of intermittent leather is proposed for a better or suitable decision making. Feature extraction technique, Grey Level Co-occurrence Matrix (GLCM) has been implemented to understand the features of color and area by extracting the texture features like Entropy, Energy, Contrast, Variance, Mean, Dissimilarity, Correlation and Homogeneity. A total of 428 intermittent leather imagesare used to understand the classification. The classifiers, Linear Discriminant Analysis (LDA) model and Support Vector Machine (SVM) are used to find out the accuracy. Further, linear discriminant model confirms 92% of accuracy over the support vector machine which is confirms 89.65% of accuracy. The proposed LDA model clearly shows that the approach is successful in classifying the variations among the defects and non-defects in intermittent leather images.

Region Wise Surface Level Defect Detection and Ranking of Crust Leather Images Based on Image Processing Techniques

“…Sornam, M. and Vasagam, S.N. 24 had suggested a model to group the Intermittent Leather Images using Linear Discriminant Model as per the category of normal leather and defective leather based on surface level features such as base color, other than base color, share of regions, share of cutting area, share of cutting value, position wise length and position wise breadth. Xie, X., et al 25 has proposed an improved image matching algorithm of defects by describing spatial signals on navel orange surface based on compressed sensing by combining of wavelet transform (WT) and speeded up robust features (SURF).…”