Tool Wear Classification Using LBP-based Descriptors Combined with LOSIB-based Enhancers.

In this paper, an automatic process to determine tool wear in machining procedures has been developed using computer vision and texture recognition techniques. Two different methods based on Local Binary Pattern (LBP) were evaluated combined with the LOSIB texture booster (Local Oriented Statistical Information Booster). The dataset used is composed by 577 images representing different wear of inserts. Two classifications were carried out: (i) a binary classification with Low-High discrimination and (ii) a ternary classification with Low-Medium-High discrimination. The results show that when combining LBP with LOSIB, all the other methods are outperformed featuring an 80.58% of accuracy in the binary classification and a 67.76% in the ternary classification. These results are very interesting for industry due to the possible savings in terms of cost and time if it is applied in a tool condition monitoring system (TCMS). [ABSTRACT FROM AUTHOR]