Evaluation of Goat Leather Quality Based on Computational Vision Techniques

• Published in: Circuits, systems, and signal processing systems, and signal processing, 2020-02, Vol.39 (2), p.651-673
• Main Authors: Santos Filho, Edmilson Q., de Sousa, Pedro Henrique Feijó, Rebouças Filho, Pedro P., Barreto, Guilherme A., de Albuquerque, Victor Hugo C.
• Format: Article
• Language: English
• Subjects: Algorithms; Bayesian analysis; Chemical treatment; Circuits and Systems; Classification; Classifiers; Clusters; Computation; Correlation analysis; Correlation coefficients; Electrical Engineering; Electronics and Microelectronics; Engineering; Industrial development; Instrumentation; Leather; Machine learning; Organic chemistry; Sheep; Signal,Image and Speech Processing; Support vector machines; Vision systems
• ISSN: 0278-081X; 1531-5878
• DOI: 10.1007/s00034-019-01180-4

The evaluation of finished (i.e., chemically treated) goat/sheep leather can be highly subjective, resulting in disagreements that can eventually lead to the interruption of production programs in the tannery and leather industry. As a result, much research has been carried out in the leather industry aiming at developing an automated system to evaluate goat/sheep leather. In this paper, a computational vision system is proposed in order to classify the quality of the leather automatically. Initially, three filtering steps are used to segment the region of interest (ROI). After highlighting the ROI, the Haralick texture obtained from the gray level co-occurrence matrix is extracted. Some descriptions (e.g., Haralick) are used in this work, namely energy, homogeneity, contrast, cluster tendency, cluster shade, correlation, information measures of correlation, and maximal correlation coefficient. After that, the performances of machine learning algorithms, such as the Naive Bayes classifier, classifier optimum-path forest and support vector machines are compared. The hit rate of the results to automatically classify goat leather quality was similar to other approaches in the literature. However, the proposed system used ten attributes, six less than the best approach found in the literature and in shorter processing time. In summary, the proposed methodology was considered reliable to automatically classify goat leather quality, as it had an accuracy of 93.22% and total processing time was 3.78 s.