ارزیابی عملکرد توصیفگرهای بافت رنگ نرم برای درجه بندی سطح با استفاده از طرح آزمایشی و رگرسیون لجستیک

This paper presents a novel approach to the question of surface grading, the soft color texture descriptors method. This method is extracted from an extensive evaluation process of several factors based on the use of two well established statistical tools: experimental design and logistic regression. The utility of different combinations of factors is evaluated in regard to the problem of automatic classification of materials such as ceramic tiles that need to be grouped according to homogeneous visual appearance, that is, the surface grading application. The set of factors includes the number of neighbors in the k-NN classifier (several values of k parameter), color space representation schemes (CIE Lab, CIE Luv, RGB, and grayscale), and color texture features (mean, standard deviation, 2nd–5th histogram moments). A factorial experimental design is performed testing all combinations of the above factors on a large image database of ceramic tiles. Accuracy estimates are computed using logistic regression to determine the best combinations of factors. From the point of view of machine learning the overall process conforms a wrapper approach able to select significant design choices (k parameter in k-NN classifier and color space) and carry out a feature selection within the set of color texture features at the same time. Experiments were repeated with alternate color texture schemes from the literature: color histograms and centile-LBP. Comparisons of methods are presented describing both accuracy estimates and runtimes.

There are many industries currently manufacturing flat surface materials that need to split their production into homogeneous series grouped by the global appearance of the final product. These kinds of products are used as wall and floor coverings. Some of them are natural products such as marble, granite or wooden boards, and others are artificial stuff such as ceramic tiles. In these industries the quality control stage is crucial in remaining competitive, and one of the most important quality problems is the non-uniformity of the visual aspect of the product within the same lot of a specific model. As the final product is used to form areas which are supposed to be uniform in appearance, the presence of pieces which look different or even slightly different is considered a serious quality fault. Nowadays, industries rely on human operators to perform the task of surface grading. However, human grading is subjective and often inconsistent between different graders [1]. Thus, automatic and reliable systems are needed. Also, capacity to inspect overall production at on-line rates is important. In this paper we approach the use of experimental design and logistic regression [2], [3] and [4], two well established statistical tools, to carry out an extensive evaluation of different factors that will lead to define a new method for the issue of surface grading. These statistical tools in conjunction provide a methodology for finding the best combinations of factors in a set of experiments by seeking to maximize the classification accuracy. The studied factors include, k parameter in k-NN classifier, color space, and a set of color texture descriptors. All these factors are evaluated at the same time using the statistical methodology in a wrapper manner [5] (see Section 5). After the evaluation two design choices are fixed (k parameter and color space), and a feature selection is performed within the color texture descriptors. The mentioned evaluation procedure was carried out successfully in previous works where multivariate projection models were used for the detection of defects in orange fruits and ceramics [6]. The new approach to surface grading, resulting from the previous evaluation study, is called the soft color texture descriptors method, and is also compared with two other methods coming from the literature, color histograms [7] and centile-LBP [8]. Experimental design and logistic regression are also used to tune design choices in these literature methods. The presented work demonstrates that a simple set of global statistics softly describing color and texture properties, together with the well-known classifier k-NN and perceptually uniform color spaces, are powerful enough to meet stringent factory performance requirements, 95% surface grading accuracy. The paper is developed as follows. Section 2 presents an overview of the literature on the surface grading issue. The proposed method of soft color texture descriptors is described in Section 3. Literature methods chosen for comparison purposes are explained in Section 4. Section 5 presents the statistical procedure for the evaluation of factors based on experimental design and logistic regression. Section 6 deals with the experimental work and results. And finally, Section 7 summarizes the paper.